Maternal and Perinatal Health of Refugees in Ontario: A Population Based Approach

Transcription

1 Maternal and Perinatal Health of Refugees in Ontario: A Population Based Approach by Susitha Wanigaratne A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Dalla Lana School of Public Health University of Toronto Copyright by Susitha Wanigaratne (2015)

2 Maternal and Perinatal Health of Refugees in Ontario: A Population Based Approach Susitha Wanigaratne Doctor of Philosophy (Epidemiology) Dalla Lana School of Public Health University of Toronto 2015 ABSTRACT Refugee women and their newborns are suspected to experience greater risk of adverse maternal and perinatal health outcomes. However, little systematic refugee specific research has been done. This dissertation poses the research question are refugee immigrant women and their newborns at higher risk of adverse maternal and perinatal outcomes compared to 1) non refugee immigrant and 2) non immigrant women? Large population based immigration ( ) and hospitalizations databases ( ) from Ontario, Canada were used (all women eligible for health insurance). The first manuscript examined severe maternal morbidity (SMM) and SMM subtypes among refugee immigrants compared to both non refugee immigrants and non immigrants. The rate of any SMM was elevated among refugees (n=30,420) compared to non refugee immigrants (n=235,540) (adjusted rate ratios (ARR)=1.22, 95% confidence interval: ) and non immigrants (n=878,709) (ARR=1.34 ( )). HIV (SMM subtype) among refugees was 8 and 17 times that of non refugee immigrants and non immigrants, respectively. SMM rate among refugees was no longer elevated when deliveries with HIV were excluded. ii

3 The second manuscript examined the risk of severe neonatal morbidity (SNM) among neonates born to refugees compared to those of non refugee immigrants and non immigrants. SNM risk among non sponsored refugees (i.e., asylum seekers) was also compared to that of sponsored refugees (i.e., government, privately sponsored). The risk of severe neonatal morbidity was significantly higher among newborns of refugee compared to non refugee immigrants (ARR=1.10 ( )) but lower in comparison to non immigrants (ARR=0.94 ( )). There was no difference by sponsorship status. The objective of the third manuscript was to determine if the relationship between refugee status and risk of preterm birth (PTB) was modified by secondary (migration to another country prior to Canada) or primary (direct to Canada) migration. A secondary objective was to examine whether this relationship varied by maternal region of birth. Secondary refugees and primary refugees experienced a significantly higher cumulative probability of PTB (22 31, 32 36, 37 weeks) compared to their secondary and primary non refugee counterparts (adjusted cumulative odds ratio (ACOR) =1.59 ( ) and ACOR=1.12 ( )), respectively. These associations were also significant for Asian immigrants. This dissertation makes a substantial contribution towards understanding the health and the determinants of health of refugee immigrant mothers and their infants in Canada. iii

4 ACKNOWLEDGEMENTS This work would not be possible without my thesis advisors and committee members. My biggest thanks goes to Dr. Marcelo Urquia for providing me with the opportunity to work on this project. With his own work in immigrant perinatal health in Canada, knowledge of and contacts in the field, methodological approaches, practical advice on scope and general enthusiasm, my dissertation experience was more enjoyable, rewarding and productive than I imagined it would be. I owe much of my pursuit of a PhD to Dr. Donald Cole who has been a selfless mentor since I first expressed an interest in epidemiology. Without his encouragement, guidance and interest in health inequity, I would not have pursued this topic. I am also grateful to Dr. Kate Bassil for engaging me in the field of perinatal epidemiology and for her guidance and support throughout my degree. I thank Dr. Ilene Hyman for encouraging me to think critically about the social determinants of immigrant health and Dr. Rahim Moineddin for his patient guidance with statistical and methodological matters. I thank Dr. Anita Gagnon for publishing early work identifying a knowledge gap with respect to refugee maternal and perinatal health in Canada; Dr. Yogendra Shakya for his thoughts on the refugee experience; Dr. K.S. Joseph for clarifying details about the severe maternal morbidity indicator; Dr. Darrell Tan for a conversation discussing the practical and policy implications of manuscript one; Dr. Joel Ray for assisting me in identifying appropriate codes for the severe neonatal morbidity indicator; Dr. Samantha Lain for sharing her thoughts on the comparisons between the Australian and Canadian severe neonatal morbidity indicator; and Dr. Jonathan Hellmann for providing feedback on severe neonatal morbidity frequencies. I also thank my thesis examiners Drs. Jason Pole, Mohammad Agha, Shiliang Liu and Laura Rosella for their insights. I am thankful to Dr. Katie McIsaac for her thorough review of a draft of this dissertation and her comments and questions. I thank my fellow students Meghan, Alana, Jordan, Sarah, Brendan, Sarai, Jess and friends Kate and Michael for providing feedback on various aspects of this work. I also thank my colleagues in Research at Cancer Care Ontario for their support and encouragement. This work would not be possible without the financial support of the University of Toronto Open Fellowship. I thank my family for giving me first hand insight into the hardships, determination and benefits associated with the Canadian immigrant experience this thesis is dedicated to them. Finally, my heartfelt thanks goes to my supportive husband, Michael. iv

5 CANDIDATE STATEMENT Susitha Wanigaratne identified and refined the research questions examined in this dissertation, proposed the study designs and methods to address the questions, conducted all statistical analyses and interpretation of the results. Susitha Wanigaratne wrote all chapters of this dissertation. Data for this dissertation was made available through a separately funded grant obtained by Dr. Urquia. Drs. Urquia, Cole, Basil, Moineddin and Hyman provided guidance and oversight on all aspects of the thesis. v

6 TABLE OF CONTENTS Table of Contents vi List of Tables ix List of Figures xv List of Appendices xvi Abbreviations xvii Chapter 1: Introduction Background Filling the gap Conceptual framework Manuscripts & study objectives Research ethics approvals Dissertation layout 11 References 13 Chapter 1 Appendices 20 Appendix 1.1: Descriptive information on refugees around the globe 20 Appendix 1.2: Descriptive information on refugees in Canada 22 Appendix 1.3: Details on literature search and summary of studies included in literature review 22 Appendix 1.4: Determinants of immigrant health conceptual framework (Hyman et al, 2007) 26 Chapter 2: Methodology Study design, settings and participants Data sources, variables, measurement and potential biases Statistical methods and data analysis 43 References 49 Chapter 2 Appendix 53 vi

7 Chapter 3: Severe maternal morbidity among refugee immigrants in Ontario 61 Abstract Introduction Methods Results Interpretation Conclusion 69 References 71 Tables 74 Chapter 3 Appendix 78 Section A3.1: Supplementary Files 78 Section A3.2: Sensitivity Analyses 87 Chapter 4: Severe neonatal morbidity among births to refugee immigrants in Ontario 89 Abstract Introduction Methods Results Discussion Conclusion 100 References 101 Tables 105 Chapter 4 Appendix 109 Section A4.1: Supplementary Files 109 Section A4.2: Sensitivity Analyses 120 Chapter 5: The migration journey of refugee immigrant women to Ontario, Canada and preterm birth in offspring 122 Abstract 123 vii

8 5.1 Introduction Methods Results Discussion Conclusion 137 References 139 Tables 144 Figures 148 Chapter 5 Appendix 149 Section A5.1: Supplementary Files 149 Section A5.2: Sensitivity Analyses 155 Chapter 6: Conclusion What have I learned? Contributions, implications & recommendations for research and practice A summary of common strengths & limitations Additional research questions and future studies Final thoughts 171 References 172 Chapter 6 Appendix 174 Integrated Bibliography 177 viii

9 LIST OF TABLES CHAPTER 1: Introduction Appendices Table A1.2.1: Distribution of Immigrants and Refugees, Ontario vs. Canada (both sexes) from Source Regions 22 Table A1.3.1: Search strategy summary: epidemiological literature examining refugee maternal and infant health 22 Table A1.3.2: Summary of literature related to refugee maternal and perinatal health 23 CHAPTER 2: Methodology Table 2.1: Sample sizes by exposure status included in each manuscript 31 Appendix Table A2.1: Methodological overview summary of manuscripts included in this dissertation 53 Table A2.2: List of exposures, outcomes and covariates 54 A2.2a: Exposures 54 A2.2b: Outcomes 55 A2.2c: Covariates 55 Table A2.3a: Proportion of CIC females having at least one birth between 1988 and Table A2.3b: Estimated proportion of NON IMMIGRANT group that are unmatched CIC female immigrants that remained in Ontario and have at least one birth between Table A2.3c. Estimated proportion of NON IMMIGRANT group that are interprovincial immigrants have at least one birth between 1988 and Table A2.3d. Estimated proportion of immigrant mothers (arriving before 1985, unlinked CIC immigrants and interprovincial immigrants) among NON IMMIGRANT mothers giving birth between 1988 and ix

10 Table A2.4. Assuming 10% of non immigrant group are truly immigrants. Potential misclassification of counts (N, deliveries), rate and rate ratios (RR) reported for refugee immigrant vs. non immigrant model reported in manuscript 1 for Severe Maternal Morbidity (SMM) 60 Table A2.5 Assuming 10% of non immigrant group are truly immigrants. Potential misclassification of counts (N, births), risk and risk ratios (RR) reported for refugee immigrant vs. non immigrant model reported in manuscript 2 for Severe Neonatal Morbidity (SNM) 60 CHAPTER 3: Severe maternal morbidity among refugee immigrants in Ontario Table 3.1: Characteristics of deliveries ( ) to refugee immigrant, non refugee immigrant mothers and non immigrants 74 Table 3.2: Severe maternal morbidity (SMM) and select SMM subtypes, number of cases, rate (/1000 deliveries) and unadjusted rate ratios (95% CI) by immigration status 76 Table 3.3: Any Severe maternal morbidity (SMM), SMM excluding mothers with HIV disease and SMM excluding mothers with significant subtypes (other than HIV disease) number of cases, rate (/1000 deliveries) and rate ratios (95% CI) comparing refugee immigrants to nonimmigrants (model 1) and refugee immigrants to non refugee immigrants (model 2) for deliveries from Appendix Section A3.1: Supplementary Files Table A3.1.1: Severe maternal morbidity (SMM) indicator ICD 10 CA diagnosis and CCI codes Joseph et al, Table A3.1.2: SMM Model 1 (refugee immigrants vs. non immigrants) coefficients (n=1,144,674) 80 Table A3.1.3: SMM without HIV Model 1 (refugee immigrants vs. non immigrants) coefficients (n=1,144,263) 81 Table A3.1.4: SMM Model 2 (refugee immigrants vs. non refugee immigrants) coefficients (n=265,960) 82 x

11 Table A3.1.5: SMM without HIV Model 2 (refugee immigrants vs. non refugee immigrants) coefficients (n=265,733) 83 Table A3.1.6: Stratification by arrival cohort any SMM and SMM excluding mothers with HIV, number of cases, rate (/1000 deliveries) and rate ratios (95% CI) comparing refugee immigrants to non immigrants (model 1) and refugee to non refugee immigrants (model 2) for deliveries from Table A3.1.7: Stratification by world region, sub regions and major refugee contributing countries of birth SMM and SMM excluding mothers with HIV disease, rate (/1000 deliveries) and rate ratios (95% CI) comparing refugee vs. non refugee immigrant for deliveries from Table A3.1.8: Coefficient estimates (SE) comparing risk of SMM among refugee immigrants to non refugee immigrants, fully adjusted models with specified interaction 86 Section A3.2: Sensitivity Analyses Table A3.2.1: Comparing rates of SMM among refugee immigrants (vs. non immigrants and vs. non refugee immigrants) stratified by age at arrival categories (<15, 15 years old and BOTH age groups) 87 Table A3.2.2: Comparing rates of SMM for refugee immigrants compared to non refugee immigrants, with and without Ontario Health Insurance Plan (OHIP) eligibility at the time of delivery 88 Table A3.2.3: Comparing nesting structures: random effect for country of birth or for mother. Coefficient estimates and unadjusted rate ratios with 95% confidence intervals comparing rates of SMM among refugee immigrants to non refugee immigrants 88 CHAPTER 4: Severe neonatal morbidity among births to refugee immigrants in Ontario Table 4.1: Characteristics of deliveries ( ) to refugee immigrant, non refugee immigrant and non immigrant women 105 xi

12 Table 4.2: Severe neonatal morbidity (SNM) and SNM subtypes (in descending order of risk among refugee immigrants), number of cases, % of births with SNM and unadjusted risk ratios (95% CI) by refugee status 107 Table 4.3: Table 3: Any Severe neonatal morbidity (SNM) number of cases, proportion SNM (%) and risk ratios (95% CI) comparing refugee immigrants to non immigrants (model 1), refugee immigrants to non refugee immigrants (model 2), unsponsored refugees to sponsored refugees (model 3) for deliveries from Appendix Section A4.1: Supplementary Files Table A4.1.1: ICD 10 CA diagnosis and CCI codes for severe neonatal morbidity (SNM) 109 Table A4.1.2: Concurrent validity comparing Severe Neonatal Morbidity (SNM) subtypes (Canada) to Neonatal Adverse Outcome Indicator (NAOI) subtypes (Australia) among all live births (singletons and multiples) with gestational age 24 weeks GA 113 Table A4.1.3: Comparing predictive validity of all live births with SNM in Canada and Australia: neonatal and infant mortality 114 Table A4.1.4: SNM Model 1 (refugee immigrants vs. non immigrants) coefficients (n=1,113,367) 115 Table A4.1.5: SNM Model 1 (refugee immigrants vs. non refugee immigrants) coefficients (n=259,719) 116 Table A4.1.6: SNM Model 3 (non sponsored refugee immigrants vs. sponsored refugee immigrants) coefficients (n=25,564) 117 Table A4.1.7: Stratification by world region and sub regions of maternal birth SNM risk ratios (RR) (95% CI) comparing refugee vs. non refugee immigrant for deliveries from Table A4.1.8: Coefficient estimates (SE) comparing risk of SNM among refugee immigrants to non refugee immigrants, fully adjusted models with specified interaction 119 Section A4.2: Sensitivity Analyses xii

13 Table A4.2.1: Comparing risk of SNM among refugee immigrants (vs. non immigrants and vs. non refugee immigrants) stratified by age at arrival categories (<15, 15 years old and BOTH age groups) 120 Table A4.2.2: Comparing risk of SNM for refugee immigrants compared to non refugee immigrants, with and without Ontario Health Insurance Plan (OHIP) eligibility at the time of birth 121 Table A4.2.3: Comparing nesting structures: random effect for country of birth or for mother. Coefficient estimates and unadjusted risk ratios with 95% confidence intervals comparing risk of SNM among refugee immigrants to non refugee immigrants 121 CHAPTER 5: The migration journey of refugee immigrant women to Ontario, Canada and preterm birth in offspring Table 5.1: Study characteristics for primary non refugee immigrants, primary refugee immigrants, secondary non refugee immigrants and secondary refugee immigrants 144 Table 5.2: Cumulative odds coefficient estimates (standard errors) and cumulative odds ratios (95% confidence intervals) for preterm birth subtypes (22 31 weeks, weeks), all immigrants multivariable model 146 Table 5.3: Adjusted cumulative odds ratios and 95% confidence intervals for preterm birth subtypes (22 31 weeks, weeks), Sub Saharan African immigrants (n=9,223) and Asian immigrants (n=86,304) 147 Appendix Section A5.1: Supplementary Files Table A5.1.1: Sub Saharan African (n=9,223) model coefficients for the cumulative probability of preterm birth among secondary and primary refugee immigrants and non refugee immigrants 149 Table A5.1.2: Asian (n=86,304) model coefficients for the cumulative probability of preterm birth among secondary and primary refugee immigrants and non refugee immigrants 150 Table A5.1.3: List of ICD 10 codes for maternal co morbidities 151 xiii

14 Table A5.1.4: Risk and risk ratios with 95% confidence intervals (95% CI) for maternal comorbidities among secondary and primary immigrant mothers with preterm births 152 Table A5.1.5: Secondary refugee and non refugee shared and unshared top 10 countries of birth and countries of last permanent residence 153 Table A5.1.6: Coefficient estimates (SE) comparing cumulative odds of PTB among secondary refugee immigrants to secondary non refugee immigrants, fully adjusted models with specified interaction 154 Table A5.1.7: Proportions of preterm birth among singleton live births to refugee immigrants and non refugee immigrants by education level at arrival in the first five years of residence 154 Section 5.2: Sensitivity Analyses Table A5.2.1: Comparing risk of PTB among refugee immigrants (vs. non refugee immigrants) stratified by secondary migration status and age at arrival categories (<15, 15 years old and BOTH age groups) 155 Table A5.2.2: Comparing risk of PTB for refugee immigrants compared to non refugee immigrants, with and without Ontario Health Insurance Plan (OHIP) eligibility at the time of birth, stratified by secondary migration status 155 Table A5.2.3: Comparing nesting structures: random effect for country of birth or for mother. Coefficient estimates and unadjusted risk ratios with 95% confidence intervals comparing risk of PTB (dichotomous) among refugee immigrants to nonrefugee immigrants, stratified by secondary migration status 156 xiv

15 LIST OF FIGURES Chapter 1: Introduction Figure 1.1: Conceptual diagram highlighting (in bold type), the determinants of refugee maternal and perinatal health explored in this dissertation 10 Appendices Figure A1.1.1: Global distribution of refugees countries of origin and countries of destination in Figure A1.1.2: Global protracted refugee situations as of 2009 (US Department of State, 2011) 21 Figure A1.4.1: Determinants of Immigrant Health Conceptual Framework (Hyman et al, 2007) 26 Chapter 5: The migration journey of refugee immigrant women to Ontario, Canada and preterm birth in offspring Figure 5.1: Illustrative examples of the migration journey for secondary refugees, primary refugees, secondary non refugees and primary non refugees to Canada 148 Figure 5.2: Secondary migration and refugee status interact to increase the risk of very (22 31 weeks) and moderate (32 36 weeks) preterm birth 148 xv

16 LIST OF APPENDICES Chapter 1 Appendices 20 Chapter 2 Appendix 53 Chapter 3 Appendix 78 Chapter 4 Appendix 109 Chapter 5 Appendix 149 Chapter 6 Appendix 174 xvi

17 ABBREVIATIONS CIC Citizenship & Immigration Canada CIHI Canadian Institutes for Health Information CRH corticotrophin releasing hormone CLPR country of last permanent residence COB country of birth COR cumulative odds ratio DAD Discharge Abstract Database FGC female genital cutting GA gestational age GEE generalized estimating equations HIV/AIDS human immunodeficiency virus/acquired immune deficiency syndrome HIV human immunodeficiency virus ICC intra class correlation coefficient ICD 10 CA/CCI 10 th revision of the International Statistical Classification of Diseases and Related Health Problems, Canadian enhancement/canadian Classification of Health Interventions ICES Institute for Clinical & Evaluative Sciences IFH Interim Federal Health IME immigration medical exam IPV intimate partner violence IRPA Immigration & Refugee Protection Act, 2002 LBW low birth weight LMP last menstrual period NAOI neonatal adverse outcomes indicator NICU neonatal intensive care unit OHIP Ontario Health Insurance Plan OR odds ratio PTB preterm birth xvii

18 PTSD PHAC RPDB RR RR SEP SGA SGBV SMM SNM STROBE UNHCR post traumatic stress disorder Public Health Agency of Canada Registered Persons Database rate ratio risk ratio socio economic position small for gestational age sex and gender based violence severe maternal morbidity severe neonatal morbidity strengthening the reporting of observational studies in epidemiology United Nations High Commissioner for Refugees xviii

19 CHAPTER 1: INTRODUCTION 1.1 BACKGROUND Refugees around the world In 2012 the United Nations High Commissioner for Refugees (UNHCR) estimated that 15.4 million refugees, 28.8 million internally displaced persons and nearly 1 million asylum seekers were of concern to the UNHCR, with the combined figure being the second largest on record since These individuals were forcibly displaced as a result of persecution, conflict, generalized violence and human rights violations based on race, religion, nationality, political view or membership in a social group. 2 Refugees are those that have fled across an international border and are unable to return to their home country or are afraid to do so. Half of all refugees worldwide come from only five countries: Afghanistan, Somalia, Iraq, Syria and Sudan. Globally, 80 percent of the world s refugees are hosted by developing countries. 1 Globally, Pakistan, Iran, Germany and Kenya host the largest numbers of refugees. Maps representing the global distribution of refugees countries of origin as well as the countries of destination based on UNHCR 2012 statistics can be found in Appendix 1.1 (Figure A1.1.1). 3 UNHCR is mandated to seek durable solutions for refugees in exile which include voluntary repatriation to their home country, local integration into the country of asylum or resettlement to a third country. Despite this mandate, durable solutions are not readily available and at the end of 2009, 10.3 million refugees (2/3 of the global refugee population at the time) lived in protracted refugee situations in 30 countries around the world. 4 The UNHCR defines protracted refugee situations where refugee populations of 25,000 persons or more, originating from the same country have been in exile for five or more consecutive years. 5 A map representing protracted refugee situations around the world at the end of 2009 can be found in Appendix 1.1 (Figure A1.1.2) Refugees in Canada Immigrants are an integral thread in Canada s economic and social fabric. Every year more than 250,000 immigrants are admitted to Canada and become permanent residents. 6 While the majority enter to participate in the labour market (Economic class immigrants) or to reunite with family members (Family class immigrants), many arrive as refugee immigrants. Between 2009 and 2013, 62% of all immigrants arrived as economic immigrants, 25% as family class immigrants and 10% as refugee immigrants. 7 1

20 2 Between 2002 and 2011, 155,000 of the approximately 280,000 refugee immigrants to Canada (56%), arrived in Ontario, 6 making it the province with the most resettled refugees. Refugee immigrants arrive to Canada under two broad sub classes: 1) sponsored and 2) non sponsored. Sponsored refugees comprise both Government assisted and privately sponsored refugee immigrants, and are selected for re settlement by the UNHCR. Their re settlement is supported by the Government of Canada or a private organization, 8 and they are eligible for provincial healthcare coverage upon arrival. 9 Nonsponsored refugees are comprised of refugee claimants (also known as asylum seekers ), who request refugee status upon arrival to Canada. They are eligible for short term federal healthcare coverage while their application is being processed, and following approval of their claim coverage is transferred to the provincial health care insurance plan (Ontario Health Insurance Plan [OHIP] in Ontario). 9 Between 2002 and 2011, refugee immigrants to Canada primarily originated from Africa and the Middle East (38%) and Asia and the Pacific (33%); smaller proportions come from South and Central America (18%), Europe and the UK (8%) and the US (2%). 6 The origins of Canada s refugees are comparable to that for refugee immigrants to Ontario (see Appendix 1.2 Table A1.2.1). The top 10 refugee contributing countries to Ontario between 1985 and 2010 were, in ranked order were: Sri Lanka (13% of all refugees), Somalia (12%), Poland (9%), Afghanistan (6%), Ethiopia (5%), Vietnam (4%), El Salvador (4%), Iraq (4%), China (3%) and Iran (3%) Forced vs. voluntary migration The factors that precipitate migration serve as an important basis for separating refugee immigrants from non refugee immigrants. Refugee immigrants are often described as forced migrants, whereas, economic and family class immigrants are often referred to as voluntary migrants. 1,10,11 Social or political factors that lead to regional conflict and generalized violence often push refugee immigrants out of their native country, forcing them to seek refuge elsewhere. On the other hand, voluntary migrants are often pulled by receiving nations such as Canada, where their individual skills and/or resources are valued. 10 This simplified distinction 11 is a useful starting point for scientific inquiry, since refugee immigrants and non refugee immigrants likely differ in their baseline health and health determinants Determinants of refugee maternal and perinatal health Many refugee women are exposed to unique pre migration, transitional migration and postmigration factors that put them and their fetuses/newborns at particularly high risk compared to nonrefugee immigrants who also experience the effects of migration. 12 The following review will highlight

21 3 differences in determinants of maternal and perinatal health comparing refugee immigrants to both non refugee immigrants and non immigrants where evidence permits. Where evidence is not available, there is reason to believe that the distribution of the determinant being discussed would place refugee women and their infants at a health disadvantage when compared to the other groups. Relevant factors include: Pre migration health issues and Canadian Immigration Policies. Pre migration health issues may continue through a woman s reproductive years and thus affect maternal and perinatal health after immigration to Canada. Prior to migration, quality health care may be limited or non existent. 13 In addition, malnutrition, limited access to clean water and increased exposure to infectious diseases may arise during the migration process. 14 Two studies of government assisted refugee immigrants revealed a higher prevalence of several infectious diseases and intestinal parasites compared to the Canadian population. 15,16 These health issues have been shown to elevate the risk for preterm birth, low birth weight, hypertension and preeclampsia. 17,18 The health status of immigrants in Canada is also shaped, to some extent through Canadian immigration policies. Of relevance to maternal health is the 1991 policy that delisted HIV/AIDS as grounds for non admittance into Canada. The subsequent Immigration and Refugee Protection Act (2002) (IRPA) built upon the 1991 legislation. Firstly, IRPA instituted mandatory testing for HIV/AIDS (as part of the immigration medical exam) for all applicants for permanent and temporary residence in Canada Secondly, severe HIV disease was considered to place excessive demand on health and social services and was adequate grounds for rejecting an applicant. 22 Finally, IRPA created an exemption for the excessive demand clause for refugees, sponsored spouses and their dependent children. These applicants could not be rejected based solely on the severity of HIV disease (or other severe health conditions). A Canadian study examining HIV infection using immigration medical screening data found that the highest rates of HIV infection were among migrants from HIV endemic areas of the world and among those in the refugee and family classes. 21 Considering these differences, immigrant selection processes likely play an important role in shaping the pre migration health of both refugee immigrant and non refugee immigrant women in Canada and this may have an impact on maternal and perinatal health of these groups Pre departure social determinants of health also affect the existing and future health state of migrants. 23

22 4 Knowledge of Canadian official languages: Of all refugee mothers who immigrated to Ontario between 1985 and 2010, 45% had no knowledge of either spoken English or French (compared to 35% among non refugee immigrant mothers), with values varying by country of birth. Limited language facility may affect a refugee woman s ability to access important settlement resources, as well as healthcare and maternity services. 24 Educational attainment: Ontario refugee mothers immigrating between 1985 and 2010 were less likely to have more than 12 years of schooling compared to their non refugee counterparts (33% vs. 55%, respectively). This difference reflects the selective nature of migration among economic migrants to Canada. Maternal education is consistently associated with adverse maternal and perinatal outcomes in Canada. 25 For women immigrating after completion of their formal schooling, education level may not change substantially after migration and may, at least among recent immigrants, also limit their access to employment and indicate their literacy at the time of delivery. Country of birth. Previously published work consisting of a Delphi consensus process involving 38 experts from 22 countries, suggested that country of birth is the most relevant and robust indicator of an immigrant woman s perinatal health. 26 Research suggests that country of birth predicts birth outcomes, above and beyond the influence of post migration residential neighbourhood status. 27 This persisted up to 15 years after a woman s migration to Canada, the period during which more than 85% of births to female immigrants take place. 27 Country of birth is used in this dissertation to signify shared exposures that could influence the health of women residing in that country including relevant social conditions and access to health care. World region of birth may help identify broader groups of women who have experienced regional social and political instability from protracted, refugee generating conflicts. 2,11 The UNHCR identifies a major protracted refugee situation as that where > 25,000 refugee immigrants from the same country go into exile for > 5 years. 2 By this definition, in 2011, the vast majority of protracted refugee situations were in parts of Africa and Asia. 4 Country of last permanent residence. An additional characteristic related to country of birth, and of specific relevance to refugee immigrants, is prolonged political and social instability causing massive refugee movement into neighbouring countries. 2,11 Refugee migrants may live for years in refugee camps or in camp like conditions 1 within a transition country prior to their arrival in Canada. Immigrants who experience this two step migration process are referred to as secondary migrants. 28 In previous work, it has been identified that secondary migration is a determinant of health when contrasted between non refugee immigrants. 29 In general, secondary non refugee immigrants tend to be voluntary, highly educated and globally competitive and who choose to re migrate to attain upward global socioeconomic position. Therefore, they tend to be healthier than their non refugee

23 5 counterparts who undergo primary migration. 29 However, secondary refugee immigrants likely continue to experience adverse living conditions in their transition country, and thus, may be at higher risk for poor health compared to non refugee immigrants and primary refugees. Indirect evidence supporting this hypothesis is the prior observation that rates of major obstetric interventions 30 and child mortality 31 tend to be higher among refugee immigrants living within settlement camps than among their African host country populations Post migration determinants of health: Access and timing of prenatal care: Maternal and perinatal health depends upon the provision of information, supports and services through pregnancy, childbirth and the post delivery period. A number of factors may prevent refugee immigrant women from sufficiently accessing these services 32. One such factor may be access to health insurance. Non sponsored refugees (i.e., refugee claimants) are eligible for short term federal health care coverage (Interim Federal Health Program IFHP) while their claim is being processed however it has been noted that a limited number of physicians accept payment through the IFHP. 33 This is in contrast to sponsored refugees (i.e., government assisted and privately sponsored) who are eligible for provincial healthcare coverage when they arrive in Canada. 9 Non sponsored refugee claimants may experience deficiencies in antenatal care because of delayed access to health insurance 34 with the potential to negatively impact maternal and infant health. Socioeconomic position: There is important variation in formal supports available to refugee immigrants that affect socio economic position, particularly during the early years of re settlement. Specifically sponsored refugees (i.e., government assisted and privately sponsored refugees) receive resettlement assistance which includes start up funds and time limited income supports 35 while nonsponsored refugees (i.e., refugee claimants) are eligible for provincial social assistance but this is not assured and can be refused. 33 Government assisted refugee immigrants are more likely to encounter employment barriers and to have jobs with low pay and poor working conditions than nonimmigrants. 33,36 Beyond this, there is little data available on the post migration socio economic status of refugee immigrants but data for all newcomers (non refugee and refugee immigrants, together) indicate that 57.5% of newcomers who arrived in Toronto between 2001 and 2006 spent more than 30% of their income on housing (vs. 21.6% among Canadian born), 36 leaving less money for other necessities. A systematic review found that being socioeconomically disadvantaged was associated with an increased risk of adverse birth outcomes, robustly seen across different measures of SES and the countries represented. 37

24 6 Duration of residence: Duration of residence, often used as a proxy for the degree of acculturation experienced by an immigrant in her host country, is recognized as an important predictor of health outcomes among immigrants. 28 Immigrant populations have been found to be healthier than host country populations upon arrival, referred to as the healthy immigrant effect yet, they lose this advantage with duration of residence. 38,39 It has been suggested that the healthy immigrant effect may not apply to refugee immigrants. 40 A recent Swedish study 41 found that war refugees had twice the odds of very preterm birth (22 32 weeks gestation) in the first year after arrival compared to the second year Psychosocial risks: Sex and gender based violence (SGBV): SBGV encompasses several sexually based offenses from domestic violence to female genital mutilation. 42,43 Since incidents of SGBV often go unreported in refugee situations, the true magnitude of the problem is unknown. Attempts have been made to estimate the extent of abuses 44 which indicate the pervasive nature of this dismal situation in crisis regions around the world. SGBV in these situations can lead to a host of pregnancy related effects. 33 Female genital cutting (FGC): The World Health Organization (WHO) estimates that between 100 and 140 million girls and women worldwide have been subjected to FGC. A prevalence study conducted in Somalia indicated that 97.9% of year old women experienced FGC. There are three types of FGM that represent varying degrees of severity with respect to mutilation of the female genitalia. 45 All types have been documented in 28 countries in Africa and a few countries in Asia and the Middle East. In 2010, 72% of refugees immigrated to Canada from regions where FGC has been documented. In 2006, the WHO 46 conducted a study of 28,393 women in six African countries which found that FGC was associated with elevated risks of: c section, post partum haemorrhage, extended maternal stay, stillbirth and early neonatal death. No association was found with low birth weight. Intimate partner violence (IPV): A review of IPV in immigrant and refugee communities conducted in 2009 in the United States suggested that based on available data, IPV is probably less prevalent among immigrant and refugee population groups compared to other groups. 47 However, the review went on to indicate that the threat of deportation, language barriers, lack of familiarity with the social system and potential discrimination, prevents many immigrant and refugee victims from seeking help and preventing future IPV. 47 Two qualitative studies of Tamil Canadian immigrant women that have experienced IPV revealed a number of contributing factors including: violence experienced during border crossing, patriarchal social norms, changes in social networks and lack of social supports, as well

25 7 as shifting gender roles with respect to employment and income. 48,49 The potential consequences of IPV include preterm labour, LBW and neonatal deaths. 50 Social support is another key social determinant of health for newcomers that has been found to have a positive impact on immigrant health. 32,51,52 However systemic issues such as limited resources and lack of integration of policies and programs with narrow service provider mandates, limits the ability to meet newcomer needs 53. Consequently, immigrants depend on their own informal social network for social support. Informal social networks provide all forms of social support; from material aid, to opportunities for social participation and receipt of essential information about the host country. 54 The limited scope 55 and availability of formal social supports and the importance of informal social supports for new immigrants has significant implications for refugees without established communities in Canada 36,56 and who may be separated from family members during the migration process. Mental health: A review of psychiatric surveys in refugee populations revealed that 1/10 adult refugees living in western countries were diagnosed with post traumatic stress disorder (PTSD), 1/20 with major depression and 1/25 with generalized anxiety disorder. 57 Evidence suggests anxiety experienced during pregnancy is a key risk factor for PTB and that depression is implicated in the etiology of low birth weight 58. Depression and anxiety in early pregnancy are also associated with risk for subsequent preeclampsia (OR 3.1, 95% CI: ) Literature: Maternal & perinatal health of refugees There is limited systematic assessment of maternal and perinatal health among refugee immigrants in Canada or in other industrialized countries, despite refugees bearing a disproportionate number of risk factors (as outlined above). A comprehensive literature search revealed thirteen studies that assessed adverse maternal and perinatal health outcomes among refugee immigrant women, with no apparent pattern to the results. See Appendix 1.3 for a description of the search strategy (Table A1.3.1) and a summary of all studies (Table A1.3.2). The studies were conducted in Canada, Croatia, Greece, Hong Kong, Ireland, the Netherlands, Norway, Sweden and the US with refugees from a variety of countries including Cambodia, Vietnam, Somalia, Croatia, Yugoslavia, Afghanistan and Iraq. The outcomes examined include: professionally identified maternal health concerns, 60 unaddressed professionally identified maternal health concerns, 60 severe acute maternal morbidity, 61 human immunodeficiency virus (HIV), 61,62 Caesarean delivery, 62,63 gestational diabetes, 62 hypertension in pregnancy, 62 low birth weight (LBW), PTB, 41,64,67 small for gestational age (SGA), 68 macrosomia, 69 stillbirth, 68 perinatal morbidity 69 and perinatal death. 63,68 70 These studies indicated increased risks, no

26 8 differences in risk, as well as lower risks among refugee immigrants compared to a reference group. However, the majority of these studies were descriptive and of variable quality, many with small sample sizes that were underpowered given the frequency of the outcome. 61,62,65 67,69 71 Also many did not adequately adjust for important confounding factors, 61,62,64 66,69 71 such as maternal age or parity. A higher quality study published in 2012 examining professionally identified concerns (i.e., problems with feeding for infant or excessive bleeding for mother) revealed significant differences among refugee immigrants compared to Canadian born women, both at 1 week (odds ratio [OR] 2.27, 95% confidence interval [CI] ) and at 4 months (OR 1.91, 95% CI ) postpartum. 60 An additional two studies of higher methodological quality (i.e., with adequate sample sizes and power; adjustment for confounding) examining common maternal and perinatal outcomes included Somali 63 refugee women in Norway and African 68 (mostly Somali) refugee women in Sweden. These studies observed significantly increased risks of emergency Caesarean delivery (odds ratio [OR] 3.0, 95% confidence interval [CI] ), antepartum stillbirth (OR 2.5, 95% CI ), SGA (OR 1.9, 95% CI ) and perinatal mortality (OR 4.3, 95% CI ). The results of these two studies may be generalizable to refugee immigrants of the same origins in Canada since refugees are also eligible for universal health care in Norway and Sweden. On the other hand, there may be important differences in immigration selection policies among these countries that make refugees from the same origins in different host countries systematically different from each other. In either case, the findings of these two studies are not necessarily valid for other large refugee immigrant groups in Canada (i.e., western and south Asia, Latin America & Caribbean). 1.2 FILLING THE GAP It is evident from the summary of determinants of refugee maternal and infant health, that refugee women experience multiple and intersecting risk factors that are known to increase risk of adverse maternal and perinatal health outcomes in the general and immigrant populations. Despite this, little refugee specific research has been done with large study samples and with representation that reflects the diversity of Canada s immigrants. To fill this gap, I have used two large, linked, population based databases available at the Institute for Clinical and Evaluative Sciences (ICES) to examine the health of both refugee mothers and their newborns that immigrated to Ontario between 1985 and A brief summary of each database is provided here. For in depth information on methodology related to the databases and variables available in each, please see Chapter 2 and its appendix.

27 1.2.1 The Discharge Abstract Database (CIHI DAD) 9 The DAD originates from the Canadian Institute of Health Information (CIHI). It has been used for the purpose of conducting perinatal research through deterministic linkage of mother newborn pairs from 2002 onwards and through probabilistic linkage at ICES with 97% accuracy prior to 2002 (coverage: 1988/ /11, data used in this dissertation: 2001/ /11). This dataset contains data on parity, plurality, maternal age and all the information needed to define the study outcomes (based on up to 25 ICD 10 diagnoses, 20 procedural codes) The Citizenship and Immigration Canada Database (CIC) The CIC database is the official Canadian immigration registry. It is composed of all legal immigrants landing records, last updated on the date permanent residency was received in Canada (coverage: ). With the exception of spoken language, which is self reported for non principal applicants, the remaining socio demographic information is objectively ascertained using the legal documentation provided by each immigrant during the application process. This dataset has been linked to the DAD. The CIC Database has been used in studies covering diabetes mellitus, immunization, health care utilization and perinatal health. 27,29,39,72 75 This database contains information on refugee status, maternal country of birth, maternal country of last permanent residence, secondary migration and knowledge of official languages and education at arrival. 1.3 CONCEPTUAL FRAMEWORK I have adapted a pre existing conceptual framework to highlight three determinants of refugee maternal and perinatal health that I examined in the three manuscripts included in this dissertation (Figure 1.1 below). Bierman et al 76 produced a conceptual framework which was further adapted by Hyman et al. 77 The latter framework (included in its entirety in Appendix 1.4, Figure 1.4.1) emphasized macro, community and family/individual level factors operating both pre migration and post migration that influence the health status of an immigrant. Also within this framework is the potential for these factors to operate differently depending on some important individual level factors (i.e., age and ethnicity). My adapted diagram is abbreviated in terms of the items that fall under the various levels but with elaboration on places within the diagram where refugee health may diverge from non refugee immigrant health (indicated by bold font). I have included linear time in the direction from left to right. I have added a box describing transitional migration with the same levels and sub levels as the

28 10 resettlement box. For this dissertation, transitional migration will be operationalized as secondary migration. I have specified a resettlement macro level factor (The Immigration & Refugee Protection Act, 2002) and an individual level factor (refugee sponsorship status) that will be explored as potential determinants of refugee maternal and perinatal health. I have also added stress as an added construct associated specifically with the migration process, where stress is defined as demands that tax or exceed the adaptive capacity of an organism that results in psychological and biological changes. 78 While stress could not be explicitly examined in this thesis, it may be integral to understanding the findings of manuscript 3. Figure 1.1: Conceptual diagram highlighting the determinants of refugee maternal and perinatal health explored in this dissertation (in bold type). 1.4 MANUSCRIPTS: TITLES & OBJECTIVES Manuscript 1: Severe maternal morbidity among refugee immigrants in Ontario Objectives: To use a Canadian specific severe maternal morbidity indicator to assess the maternal health of refugee immigrant women in comparison to 1) non immigrant (majority Canadianborn) women and 2) non refugee immigrant women, in Ontario, Canada. We further examined whether differences varied according to: 1) SMM subtypes, and 2) periods of immigration (before and after 2002) coinciding with removal of an inadmissibility policy aimed at refugees with medical conditions which could place an excessive burden on the health care system. Manuscript 2: Severe neonatal morbidity among births to refugee Immigrants in Ontario. Objectives: 1) to assess differences in severe neonatal morbidity among infants born to refugee immigrant women in comparison to infants born to a) non immigrant (majority Canadian born) women and b) non refugee immigrant women, in Ontario, Canada; 2) to determine whether risk of severe neonatal morbidity varies by refugee sponsorship status (non sponsored vs. sponsored).

29 11 Manuscript 3: The migration journey of refugee women to Ontario, Canada and preterm birth of offspring. Objectives: 1) to determine if the relationship between refugee status and risk of PTB varied with the nature of migration journey (i.e., secondary vs. primary migration); 2) to determine whether this relationship varied with maternal region of birth; and 3) to determine if secondary migrants transition countries (countries of last permanent residence CLPR) contributes to PTB risk. 1.5 RESEARCH ETHICS APPROVALS This study received research ethics approval from: 1) The Institute for Clinical and Evaluative Sciences and Sunnybrook Health Sciences Centre on May 21, 2013 (Study number: ); 2) St. Michael s Hospital Research Ethics Office on October 29, 2013 (REB# c ); and 3) The University of Toronto Research Ethics Office on December 2, 2013 (# 29629). Development of the thesis proposal occurred under Dr. Urquia s related project entitled Immigration to Canada and other industrialized countries and maternal and infant morbidity which received research ethics approval from 1) The Institute for Clinical and Evaluative Sciences and Sunnybrook Health Sciences Centre on February 16, 2012 (Study number: ) and 2) St. Michael s Hospital Research Ethics Office on May 23, 2012 (#12 087). 1.6 DISSERTATION LAYOUT This dissertation consists of six chapters: Chapter 1: this introduction; Chapter 2: Methodology; Chapter 3 (Manuscript 1): Severe maternal morbidity among refugee immigrants in Ontario; Chapter 4 (Manuscript 2): Severe neonatal morbidity among births to refugee immigrants in Ontario; Chapter 5 (Manuscript 3): The migration journey of refugee women in Ontario, Canada and preterm birth of offspring; and Chapter 6: Conclusions. For Chapter 1, the references and appendices (the latter labelled Chapter 1 Appendix and containing the tables and figures mentioned in this chapter) are placed after the body of the chapter. In Chapter 2: Methodology, in depth methodological details and decisions relevant for all manuscripts is provided. For each manuscript chapter references and the main tables and figures (numbered in order of appearance) follow the body. Each manuscript chapter also has an appendix with supplementary files (specified as section A.1 containing additional information regarding methodology, supplementary tables and analyses) and sensitivity analyses (specified as section A.2 evaluating the impact of methodological decisions on study results). The concluding chapter outlines what I have learned from this dissertation, how it contributes to the field of refugee

30 12 maternal and perinatal health, the implications of the work and some recommendations for research, policy and practice based on the findings. I also describe some strengths and limitations of the work and I end by posing some additional research questions that build on my dissertation findings.

31 REFERENCES United Nations High Commissioner for Refugees. Displacement The new 21st century challenge, UNHCR Global Trends 2012 [Internet] Available from: 2. United Nations University. Protracted refugee situations Political, human rights and security implications. New York, New York: United Nations University Press; Hennig B. Global refugee trends [Internet]. Views World [cited 2014 Jan 10]. Available from: 4. US Department of State. Protracted refugee situations [Internet]. US Dep State Dipl Action [cited 2013 Aug 15]. Available from: 5. United Nations High Commissioner for Refugees. Chapter 5 Protracted refugee situations: the search for practical solutions [Internet] Available from: 6. Citizenship and Immigration Canada. Canada facts and figures Immigration overview permanent and temporary residents 2011 [Internet] Available from: stats/facts2011.pdf 7. Citizenship and Immigration Canada. Preliminary tables permanent and temporary residents, 2013 [Internet] [cited 2014 Jul 4]. Available from: preliminary/01.asp 8. Becklumb P. Background paper Canada s immigration program [Internet]. Library of Parliament; 2008 Sep. Available from: e.pdf 9. Citizenship and Immigration Canada. Health care Refugees [Internet] Dec. Available from: healthcare.asp 10. Castles S. Towards a sociology of forced migration and social transformation. Sociology. 2003;37(1):13 34.

32 Long K. Permanent crises?: unlocking the protracted displacement of refugees and internally displaced persons [cited 2013 Aug 15]; Available from: Gagnon AJ, Merry L, Robinson C. A systematic review of refugee women s reproductive health. Refuge Pottie K, Topp P, Kilbertus F. Case report: profound anemia. Chronic disease detection and global health disparities. Can Fam Physician. 2006;52(3): Oxman Martinez, J, Hanley, J. Health and social services for Canada s multicultural population: challenges for equity. Ottawa: Heritage Canada; Denburg A, Rashid M, Brophy J, et al. Initial health screening results for Karen refugees: a retrospective review. Can Commun Dis Rep. 2007;33(13): Pottie K, Janakiram P, Topp P, McCarthy A. Prevalence of selected preventable and treatable diseases among government assisted refugees. Can Fam Physician. 2007;53: Schieve LA, Handler A, Hershow R, Persky V, Davis F. Urinary tract infection during pregnancy: its association with maternal morbidity and perinatal outcome. Am J Public Health. 1994;84(3): Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. The Lancet. 2008;371(9606): Bisaillon LM. Human rights consequences of mandatory HIV screening policy of newcomers to Canada. Health Hum Rights Int J. 2010;12(2). 20. Citizenship & Immigration Canada. HIV screening [Internet] Nov. Available from: Zencovich M, Kennedy K, MacPherson DW, Gushulak BD. Immigration medical screening and HIV infection in Canada. Int J STD AIDS. 2006;17: Public Health Agency of Canada. Population specific HIV/AIDS status report. People form countries where HIV is endemic black people of African and Caribbean descent living in Canada. [Internet] Available from: aspc.gc.ca/aids sida/publication/pspd/africacaribbe/pdf/ps spreport eng.pdf

33 Gushulak BD, MacPherson DW. Health aspects of the pre departure phase of migration. PLoS Med. 2011;8(5):e McKeary M, Newbold B. Barriers to Care: The Challenges for Canadian Refugees and their Health Care Providers. J Refug Stud Nov 8;23(4): Luo Z C, Wilkins R, Kramer MS, Fetal and Infant Health Study Group of the Canadian Perinatal Surveillance System. Effect of neighbourhood income and maternal education on birth outcomes: a population based study. CMAJ. 2006;174(10): Gagnon AJ, Zimbeck M, Zeitlin J. Migration and perinatal health surveillance: An international Delphi survey. Eur J Obstet Gynecol Reprod Biol Mar 1;149(1): Urquia ML, Frank JW, Moineddin R, Glazier RH. Does time since immigration modify neighborhood deprivation gradients in preterm birth? A multilevel analysis. J Urban Health. 2011;88(5): Urquia ML, Gagnon AJ. Glossary: migration and health. J Epidemiol Community Health. 2011;65(5): Urquia ML, Frank JW, Glazier RH. From places to flows. International secondary migration and birth outcomes. Soc Sci Med. 2010;71(9): Orach CG, Brouwere V De. Postemergency health services for refugee and host populations in Uganda, The Lancet. 2004;364(9434): Hargreaves JR, Collinson MA, Kahn K, Clark SJ, Tollman SM. Childhood mortality among former Mozambican refugees and their hosts in rural South Africa. Int J Epidemiol. 2004;33(6): Grewal SK, Bhagat R, Balneaves LG. Perinatal beliefs and practices of immigrant Punjabi women living in Canada. J Obstet Gynecol Neonatal Nurs May 27;37(3): Gagnon A, Rousseau H, Welt M, et al. Appendix 20: Screening during pregnancy: evidence review for newly arriving immigrants and refugees. Can Med Assoc J Sep 6;183(12):E824 E Caulford P, Vali Y. Providing health care to medically uninsured immigrants and refugees. CMAJ Apr 25;174(9):

34 Elgersma S, Dolin B. Assistance for government sponsored refugees, chosen abroad [Internet]. Library of Parliament; 2008 Sep. Report No.: PRB 04 39E. Available from: e.pdf 36. Toronto Public Health, Access Alliance Multicultural Health and Community Services. The global city: newcomer health in Toronto [Internet] Nov. Available from: standards/map/f iles/pdf/global_city.pdf 37. Blumenshine P, Egerter S, Barclay CJ, Cubbin C, Braveman PA. Socioeconomic disparities in adverse birth outcomes. Am J Prev Med Sep;39(3): Ray JG, Vermeulen MJ, Schull MJ, Singh G, Shah R, Redelmeier DA. Results of the recent immigrant pregnancy and perinatal long term evaluation study (RIPPLES). CMAJ. 2007;176(10): Urquia M, Frank J, Moineddin R, Glazier R. Immigrants duration of residence and adverse birth outcomes: a population based study. BJOG Int J Obstet Gynaecol Apr;117(5): DesMeules M, Gold J, Kazanjian A, et al. Immigrant health assessment: Profiling health diversity in Canada. Can J Public Health. 95(3): Liu C, Urquia M, Cnattingius S, Hjern A. Migration and preterm birth in war refugees: a Swedish cohort study. Eur J Epidemiol Jan 14;29: World Health Organization. Female genital mutilation and other harmful practices [Internet] [cited 2014 May 5]. Available from: Holmes R, Bhuvanendra D. Preventing and respond to gender based violence in humanitarian crises [Internet] Jan. Report No.: 77. Available from: magazine/issue 13/evaluation of the ngo field cooperationprotocol/download/np77pdf 44. Stark L, Ager A. A systematic review of prevalence studies of gender based violence in complex emergencies. Trauma Violence Abuse Jul 1;12(3):

35 Perron L, Senikas V, Burnett M, Davis V. Female genital cutting. J Obstet Gynecol Can. 2013;35(11): Eke N, Nkanginieme K. Female genital mutilation and obstetric outcome. The Lancet Jun 3;367(9525): Family Violence Prevention Fund. Intimate partner violence in immigrant and refugee communities: challenges, promising practices and recommendations [Internet] Mar. Available from: 9.pdf 48. Guruge S, Khanlou N, Gastaldo D. Intimate male partner violence in the migration process: intersections of gender, race and class. J Adv Nurs Jan;66(1): Hyman I, Mason R, Guruge S, Berman H, Kanagaratnam P, Manuel L. Perceptions of factors contributing to intimate partner violence among Sri Lankan Tamil immigrant women in Canada. Health Care Women Int Sep;32(9): Sharps PW, Laughon K, Giangrande SK. Intimate partner violence and the childbearing year: Maternal and infant health consequences. Trauma Violence Abuse Apr 1;8(2): Bandyopadhyay M, Small R, Watson LF, Brown S. Life with a new baby: How do immigrant and Australian born women s experiences compare? Aust N Z J Public Health Aug 3;34(4): Chakrabarti R. Therapeutic networks of pregnancy care: Bengali immigrant women in New York City. Soc Sci Med. 2010;71(2): Simich L, Beiser M, Stewart M, Mwakarimba E. Providing social support for immigrants and refugees in Canada: challenges and directions. J Immigr Minor Health Oct;7(4): Hernández Plaza S, Alonso Morillejo E, Pozo Muñoz C. Social support interventions in migrant populations. Br J Soc Work. 2006;36(7): Lipman A, Longino CF. Formal and informal support: A conceptual clarification. J Appl Gerontol Jun 1;1(1):

36 Stewart M, Anderson J, Beiser M, et al. Multicultural meanings of social support among immigrants and refugees. Int Migr. 2008;46(3): Fazel M, Wheeler J, Danesh J. Prevalence of serious mental disorder in 7000 refugees resettled in western countries: a systematic review. The Lancet. 2005;365(9467): Dunkel Schetter C. Psychological science on pregnancy: Stress processes, biopsychosocial models, and emerging research issues. Annu Rev Psychol. 2011;62: Kurki T, Hiilesmaa V, Raitasalo R, Mattila H, Ylikorkala O. Depression and anxiety in early pregnancy and risk for preeclampsia. Obstet Gynecol. 2000;95(4): Gagnon AJ, Dougherty G, Wahoush O, et al. International migration to Canada: The post birth health of mothers and infants by immigration class. Soc Sci Med. 2012;76: Hanegem N Van, Miltenburg AS, Zwart JJ, Bloemenkamp KWM, Roosmalen J Van. Severe acute maternal morbidity in asylum seekers: a two year nationwide cohort study in the Netherlands. Acta Obstet Gynecol Scand Sep;90(9): Kandasamy T, Berniak R, Shah R, Yudin MH, Spitzer R. Obstetric Risks and Outcomes of Refugee Women at a Single Centre in Toronto. J Obstet Gynaecol Can. 2014;36(4): Vangen S, Stoltenberg C, Johansen REB, Sundby J, Stray Pedersen B. Perinatal complications among ethnic Somalis in Norway. Acta Obstet Gynecol Scand. 2002;81(4): Kuvacic I, Skrablin S, Hodzic D, Milkovic G. Possible influence of expatriation on perinatal outcome. Acta Obstet Gynecol Scand. 1996;75(4): Davis JM, Goldenring J, McChesney M, Medina A. Pregnancy outcomes of Indochinese refugees, Santa Clara County, California. Am J Public Health. 1982;72(7): Gann P, Nghiem L, Warner S. Pregnancy characteristics and outcomes of Cambodian refugees. Am J Public Health. 1989;79(9): Flynn PM, Foster EM, Brost BC. Indicators of acculturation related to Somali refugee women s birth outcomes in Minnesota. J Immigr Minor Health. 2011;13(2):

37 Essen B, Hanson BS, Ostergren PO, Lindquist PG, Gudmundsson S. Increased perinatal mortality among sub Saharan immigrants in a city population in Sweden. Acta Obstet Gynecol Scand. 2000;79(9): King PA, Duthie SJ, Li DFH, Ma HK. Obstetric outcome among Vietnamese refugees in Hong Kong: an age matched case controlled study. Int J Gynaecol Obstet Off Organ Int Fed Gynaecol Obstet. 1990;33: Lalchandani S, MacQuillan K, Sheil O. Obstetric profiles and pregnancy outcomes of immigrant women with refugee status. Ir Med J. 2001;94(3): Malamitsi Puchner A, Tzala L, Minaretzis D, Michalas S, Aravantinos D. Preterm delivery and low birthweight among refugees in Greece. Paediatr Perinat Epidemiol. 1994;8(4): Creatore MI, Moineddin R, Booth G, et al. Age and sex related prevalence of diabetes mellitus among immigrants to Ontario, Canada. Can Med Assoc J. 2010;182(8): Lofters A, Moineddin R, Hwang S, Glazier R. Predictors of low cervical cancer screening among immigrant women in Ontario, Canada. BMC Womens Health. 2011;11(20). 74. Urquia ML, Glazier RH, Blondel B, et al. International migration and adverse birth outcomes: role of ethnicity, region of origin and destination. J Epidemiol Community Health Aug 19;64(3): Urquia ML, Frank JW, Glazier RH, Moineddin R. Birth outcomes by neighbourhood income and recent immigration in Toronto. Health Rep. 2007;18(4): Agnew V. Racialized Migrant Women in Canada: Essays on Health, Violence and Equity. University of Toronto Press; Hyman I, McDermott S, DesMeules M. Development of a framework to examine the determinants of health among Canadian immigrants. Canadian Public Health Association Annual Conference, Halifax; Behrman RE, Butler AS. Preterm birth: causes, consequences, and prevention. National Academies Press; 2007.

38 APPENDIX 1.1: DESCRIPTIVE INFORMATION ON REFUGEES AROUND THE GLOBE Figure A1.1.1: Global distribution of refugees countries of origin and countries of destination in

39 Figure A1.1.2: Global protracted refugee situations as of

40 22 APPENDIX 1.2: DESCRIPTIVE INFORMATION ON REFUGEES IN CANADA Table A1.2.1: Distribution of Immigrants and Refugees, Ontario vs. Canada (both sexes) from Source Regions* 6 All Immigrants Refugees Ontario Canada excluding Ontario Ontario Canada excluding Ontario Source Region n % n % N % n % Africa & The Middle East 243, , , , Asia & Pacific 627, , , , South & Central America 116, , , , United States 44, , , ,504 1 Europe & The United Kingdom 158, , , ,036 5 Not Stated 409 <1 291 <1 257 <1 158 <1 Total 1,191,533 1,284, , ,160 *Refers to principal country of last permanent residence. For refugee claimants, source country refers to the principal country of alleged persecution. APPENDIX 1.3: DETAILS ON LITERATURE SEARCH AND SUMMARY OF STUDIES INCLUDED IN LITERATURE REVIEW Table A Search strategy summary: epidemiological literature examining refugee maternal and infant health 1. Medline Search Terms for exposure of interest exp Refugees Terms for outcomes of interest exp [Maternal Deprivation OR Maternal Mortality OR Maternal Child Health Centres OR Maternal Exposure OR Maternal Fetal Relations OR Maternal Behaviour OR Maternal Nutritional Physiological Phenomena OR Maternal Welfare OR Maternal Fetal Exchange OR Maternal Child Nursing OR Maternal Health Services OR Maternal Death OR HIV OR Stillbirth/ep [epidemiology] OR Congenital Abnormalities OR Premature Birth OR Infant, Small for Gestational age OR Fetal Growth Retardation/ep OR Intensive Care, Neonatal OR Intensive Care Units, Neonatal OR Fetal Death OR neonatal death.mp OR perinatal death.mp OR maternal morbidity.mp OR severe maternal morbidity.mp OR severe acute maternal morbidity. mp] 2. Hand Searching Reference Lists a) Relevant articles found in Medline from above b) Peer reviewed literature by Dr. Anita Gagnon c) Grey literature authored by Toronto based refugee serving health centres (i.e., Access Alliance, South Riverdale Health Centre)

41 23 Table A1.3.2: Summary of literature related to refugee maternal and perinatal health # Author, Year Country Design, Population, Data, Sample Size Outcome(s) Comparison, Analysis 1 Davis, USA Descriptive, all Many None; Indochinese refugee proportions of univariate birth records LBW, PTB, statistics (n=542), random complications review of 37 during obstetric charts, pregnancy/ in Santa labour/delivery Clara county, etc California 2 Flynn, USA Descriptive, All hospital based births to Somali women between in mid west (n=581) 3 Essen, Sweden Cohort study, a single hospital s births from between , linked to population register (All foreign born n=15,639; all African born n=11,140) 4 Gagnon, Canada Cohort study recruited from 11 hospitals in cities receiving the largest number of refugees, (n=1127) 5 Gann, USA Descriptive, Recruitment of all Cambodian women in Lowell, Mass. From prenatal care clinic (n=554) Gestational age and birth weight Perinatal mortality Unaddressed professionallyidentified concerns postpartum Low birth weight, preterm birth Internal comparison by time period; bivariate statistics Foreign born (mostly refugees) vs. Swedish born; logistic regression Canadianborn population; logistic regression Clinic comparison; bivariate General Results impressions are presentation of greater risk, lack prenatal care, more LBW and pregnancy complications Proportion of preterm birth increased with longer duration or residence (4.3% to 9.9%) All foreign born, AOR=1.5. ( ); all African origin AOR=4.0 ( ) One week postpartum OR=2.27 ( ); four months postpartum OR=2.18 ( ) LBW 6.4% vs. 10.3%; PTB 9.1% vs. 6.4% Key Challenges descriptive study; small sample size; no comparison group; unable to adjust for confounding descriptive study; small sample size; did not account for previous preterm birth in earlier time period Not representative of Canada s large Asian refugee immigrant population Outcome related to quality of care; uncommon, not easily reproducible outcome Descriptive study; small sample size; unable to adjust for confounding 6 Kandasamy, Canada Retrospective cohort, many Immigrants, Prior c section Small sample size; did not adjust

42 24 # Author, Year Country Design, Population, Data, Sample Size all refugee (n=274) and non refugee (n=8,529) women delivering in hospital Jan 1, 2008 to Dec 31, King, Hong Kong Descriptive, all Vietnamese refugees delivering between (n=308) and age matched Hong Kong patients (n=308) 8 Kuvacic, Croatia Descriptive, n=9150 deliveries to free (undisplaced) and displaced/refugee Croatians, n=9389 neonates, unstated data source 9 Lalchandani, Ireland Descriptive, consecutive hospital births to refugee women between 1999 and 2000 (n=7608 with 271 refugees), clinic data 10 Liu, Sweden Register based study, all singleton live births between 1992 and 2008 during the first five years of residence. Refugees from Yugoslavia, Iraq, Somalia and Outcome(s) Many Many Many Preterm birth, weeks (very PTB) and weeks (moderately PTB) Comparison, Analysis permanent residents, CDN citizens (native & foreign born) Native born age matched group; bivariate Free Croatians; bivariate Hospital based; bivariate Internal comparison births occurring in 2 nd year of duration of residence; multinomial General Results 21.5% vs. 12.8% (p=0.007); HIV positive 0.4 % vs. 3.6% (p=0.006); late prenatal care 1.8% vs. 10.2% (p<0.001); poor social support 1.8% vs. 3.6% LBW 8.4% vs. 4.2%; small for dates 17.5% vs 4.5% Perinatal mortality without major malformation 18.9% vs 9.2% Low epidural administration, oxytocin acceleration, instrumental delivery and episiotomy rates Increased risk of preterm in the 1 st year of duration of residence compared to 2 nd year OR=1.39 ( ) Key Challenges for confounding Descriptive study; small sample size; unable to adjust for confounding Unknown methods of data collection and sample selection; small sample sizes; unable to adjust for confounding Descriptive study; unknown methods of data collection; small sample size; unable to adjust for confounding No non refugee immigrant or native born population comparison

43 25 # Author, Year Country Design, Population, Data, Sample Size Afghanistan (n=20,723) 11 Malamitsi Greece Hospital based, Puchner, Refugees (n=638 from Eastern Europe, Soviet Union, Middle East and Africa) and parity and delivery date matched indigenous Greek (n=1231) giving birth between Jan 1990 and June Vangen, Norway Cross sectional population based registry study, all births to Somali women (n=1733) and Norwegian women (n=702,192) giving birth between Sweden Population based 13 Van Hanegem, cohort study, all births to asylum seekers in the Netherlands (n=1310), nonwestern immigrants (n=61,022), general Dutch population (n=369,711) Outcome(s) Preterm birth and low birthweight Perinatal complications Severe acute maternal morbidity (SAMM); HIV Comparison, Analysis logistic regression Indigenous Greek; conditional logistic regression Norwegian born women; logistic regression All pregnant women in the Netherlands; calculation of relative risks General Results PTB and LBW: all groups nonsignificant or significantly lower than Greek population. Fetal distress (OR=2.6, 95% CI: ); emergency c section (OR=3.0, ); Agpar below 7 (OR=3.1, ); prelabor fetal death (OR=2.5, ) SAMM vs. Non Western RR=3.6 ( ); vs. General Dutch population RR=4.5 ( ) Among SAMM cases HIV vs. Non Western RR=8.9 ( ); vs. Dutch RR=40.3 (13 130) Key Challenges Unknown methods of data collection and data quality; small sample size; unable to adjust for confounding Not representative of all refugee women in Ontario Small sample size for asylum seekers; no adjustment for confounders;

44 APPENDIX 1.4: DETERMINANTS OF IMMIGRANT HEALTH CONCEPTUAL FRAMEWORK (HYMAN et al, 2007) Figure A1.4.1: Determinants of Immigrant Health Conceptual Framework (Hyman et al, 2007) 26

45 CHAPTER 2: METHODOLOGY The format of this methods chapter follows the Strengthening the reporting of observational studies in epidemiology (STROBE) guidelines 1 for reporting of observational studies. In this chapter, emphasis is placed on additional methodological details and rationale for decisions that were either described briefly or not included in the methodology sections of the manuscripts included in this dissertation. 2.1 STUDY DESIGN, SETTING AND PARTICIPANTS A table with a methodological overview of the three manuscripts included in this dissertation can be found in the appendix to this chapter (see Table A2.1). The details summarized in this table include: study setting, study design, data sources and dates, inclusion and exclusion criteria, outcome, exposure as well as confounding and mediating variables Study Design and inclusion criteria All three studies take place in the province of Ontario, Canada. In terms of study design, manuscripts one and two (Chapters 3 and 4) are population based database studies with retrospective assessment of exposure. These two studies are described in this manner since the units of analysis (either deliveries or births) are chosen based on the outcome and not the exposure. For manuscript one, all hospital based deliveries were included that occurred between April 1, 2002 and March 31, 2011 whereas for manuscript two, all singleton live births were included that occurred over the same time period. All deliveries and singleton live births were identified using the Discharge Abstract Database (DAD) which contains diagnostic and procedural data on all hospital deliveries in the province of Ontario. These deliveries and births were then retrospectively linked to the Citizenship & Immigration Canada (CIC) database to determine exposure status that is refugee immigrant and compared to non refugee immigrants and non immigrants. For manuscript 3 (Chapter 5), a retrospective population based cohort was assembled by identifying all refugee immigrants (exposure of interest) and non refugee immigrants in the CIC database that arrived in Ontario between April 1, 2002 and March 31, 2011 and also had a singleton live birth over the same period of time, determined through linkage to the DAD. 27

46 Exclusions. Two exclusion criteria were applied to the study populations described above. Firstly, from the refugee immigrant and non refugee immigrant populations, all girls who were <15 years of age at the time of arrival in Canada were excluded from analyses. This exclusion is based on literature suggesting that children experience migration differently from adults and that educational attainment in the host country and subsequent integration into the host country varies with the age at arrival 2 with likely impacts on health outcomes (see Chapter 1, section ). Improvements in language and education may be particularly relevant for immigrant girls who arrive at school age since they will be enrolled in school upon arrival. As a by product of participating in formal education in Ontario, Canadian language ability is also likely to improve if it was at a deficit prior to arrival. These differences in education and integration have important implications on the validity of the measures of education and language ability (measured at arrival) available in the CIC database, in that there is a higher likelihood that these measures will be misclassified with increasing duration of residence for the immigrants who arrive as children compare to those that arrive as adults. This is not to say that women who immigrate as adults do not improve their education and language ability but given the complex relationship between factors that influence post migration education (i.e., child development theory, pre migration education, acceptance of credentials and skills in the host country 2,3 ), for simplicity sake it is assumed: 1) immigrants who arrive as girls will be more likely to integrate into Canadian society due to participation in formal education in Canada; 2) immigrants who arrive as women are less likely to participate in education in Canada and 3) because of points 1 and 2, education and language ability measured at arrival are less likely to suffer from misclassification among immigrant women who arrive as adults. To reduce chances of misclassification of these socio demographic measures and to examine a group of women with more uniform chances of integration, both refugee immigrant and non refugee immigrant women < 15 years at arrival are excluded from all manuscripts included in this dissertation. The second exclusion criterion occurs as a result of the first, which is to exclude non immigrant women in manuscripts one and two (Chapter 3 and 4) who were less than 15 years of age at the time of delivery or birth. This was done to ensure that the maternal ages of all groups being compared were similar. Considering the importance of exposures over the life course, an additional important reason for the exclusion cut point at younger than 15 years of age is that maternal and perinatal outcomes may be influenced by exposures that occurred prior to reaching reproductive age. Since this dissertation focuses on differences between refugee immigrants, non refugee immigrants and non immigrants, the

47 29 conditions women are exposed to prior to reaching reproductive age may be an important contribution to any differences seen in maternal and perinatal outcomes. Excluding immigrant women who were under 15 years of age at arrival eliminates the mixture of effects resulting from this excluded group of women being exposed to both foreign and Canadian social and cultural environments leading up to reproductive age. Including only those women that were 15 year of age or older ensures that different environments for development (outside of Canada for refugee immigrants and non refugee immigrants and Canada for non immigrants) are also being compared. Sensitivity analysis was conducted to examine the impact associated with age related exclusions for each manuscript. Adjusted risk ratios were estimated from study populations which excluded women who were < 15 years at arrival (as reported in the manuscripts) to one where this population was included ( BOTH ). The results of these analyses are summarized in the appendix of each manuscript (Tables A3.2.1, A4.2.1, A5.2.1). Overall, for manuscript two and three (Chapters 4 and 5), the exclusions made no difference to risk ratios (RR) and cumulative odds ratios (COR) reported. In manuscript one (Chapter 3), RRs including both age groups, were about 10% lower than those reported in the manuscript when compared to non immigrants and roughly the same when compared to non refugees. However, the conclusions remained the same. 2.2 DATA SOURCES, VARIABLES, MEASUREMENT AND POTENTIAL BIASES This section of the methods describes in detail the data sources used in the three manuscripts as well as the salient methodology (deterministic and probabilistic linkage) associated with the data sources, study variables and their measurement as well as information and selection biases that may result. Details of all study variables and other relevant information is summarized in the appendix to this chapter (Tables A2.2a, A2.2b and A2.2c) Citizenship & Immigration Canada database & the Ontario Registered Persons Database. The Citizenship and Immigration Canada Database (CIC) is the official Canadian immigration registry, composed of all legal immigrants landing (arrival) records, in which the landing date indicates the date of receiving permanent residency in Canada. All immigrant women who intended to live in Ontario upon arrival to Canada between 1985 and 2010 were included in the dataset prior to linkage to the health care registry (Registered Persons Database RPDB, described in the next paragraph). In

48 30 addition to landing information, the CIC database also contains information on refugee status (refugee immigrant and non refugee immigrant), country of birth, and country of last permanent residence, as well as marital status, education level and knowledge of Canadian official languages upon arrival to Canada. The objective ascertainment of refugee status in administrative population level data is unique to this database. Its use is an important contribution to refugee health research as primary data collection studies in this area suffer from small sample sizes and high attrition. The CIC database has been used previously in several epidemiological studies on topics such as diabetes mellitus, immunization, health care utilization and perinatal health Exposure status who is included? The CIC dataset was deterministically and probabilistically linked to the aforementioned RPDB, which is Ontario s health care registry. The RPDB began keeping track of all Ontario residents eligible for provincial health care coverage (Ontario Health Insurance Plan or OHIP) in 1991 but includes residents who were eligible for OHIP prior to initiation of the registry. Women who were eligible for health care at any time between the fiscal years 1991/92 to 2010/11 were included in the linkage process with the CIC database. Linkage of the CIC database to the RPDB allowed us: 1) to determine which OHIP eligible residents immigrated to Ontario between 1985 and 2010 and as a result, which OHIP eligible residents would henceforth be described refugee immigrants or non refugee immigrants (positive linkage to the CIC) as well as non immigrants (no linkage to the CIC), and 2) to identify the unique encrypted health card number for immigrants in the CIC database, further facilitating linkage to the DAD. The step wise linkage (both deterministic and probabilistic) was based on first names and surnames, sex, date of birth and time of arrival and registration to the provincial health insurance plan. This rigorous linkage methodology resulted in 85.8% (67% through deterministic and 19% through probabilistic methods) of both males and females in the CIC database being successfully matched to Ontario residents with valid health cards, demonstrating the high accuracy of the linkage. 9 It was also found that there were no major differences in socio demographic characteristics between matched and non matched individuals (selection bias). 7,10 Individuals in the CIC database that were unmatched to the RPDB may have: 1) re migrated to other provinces and therefore were not present in the RPDB to begin with, 2) returned to their countries of origin after arrival (likely very small proportion), or 3) may be misclassified and included in the nonimmigrant group. Relevant to point 1), a recent report on the interprovincial mobility of immigrants to Canada (June 2010) found that 8% of immigrants that intended to migrate to Ontario (between 1991 and 2006) were living in another province as of If we remove this 8% from the 14% unmatched,

49 this leaves a potential 6% of CIC immigrants (migrating between 1985 and 2010) in the non immigrant pool. 31 Operationalizing exposure groups. Database details and operationalizing exposure status is summarized in Table A2.2a. All manuscripts use a dichotomous categorization of refugee status comparing refugee immigrants to non refugee immigrants. For manuscripts one and two, an ordinal categorization of refugee status is also used to compare refugee immigrants to non immigrants and non refugee immigrants to non immigrants. The sample sizes included for each manuscript are summarized in Table 2.1 below. Table 2.1: Sample sizes by exposure status included in each manuscript Manuscript (chapter) Unit of analysis Refugee immigrants (n) Non refugee immigrants (n) Nonimmigrants (n) 1 (3) Deliveries 30, , ,975 2 (4) Singleton live births 29, , ,617 3 (5) Singleton live births 12, ,895 Not included Socio demographic information for immigrants in the Citizenship & Immigration Database. The CIC database also contains valuable socio demographic information for immigrants that arrived between 1985 and This includes education level and knowledge of Canadian official languages upon arrival to Canada as well as country of birth (COB) and country of last permanent residence (CLPR). With the exception of language, which is self reported for non principal applicants, the remaining information is ascertained based on legal documentation provided by the immigrants during their application process. Education and language at arrival. Both level of education and knowledge of official languages were considered confounders of the relationship between refugee status and the health outcome of interest in all manuscripts. This is because both of these variables are determinants of being a successful non refugee immigrant, in that economic immigrants are generally highly educated and have a working knowledge of one or both official languages. This is less of a requirement for refugee immigrants. Level of educational attainment at arrival was entered into all models as an ordinal categorical variable with the highest category of education (Bachelor s, Master s or Doctorate degree) used as the reference level. The other categories for education were categorized as: 0 9 years, 10 12

50 32 years, 13+ years and trade certificate/non university diploma. Spoken language ability was entered into all models as a dichotomous variable with no knowledge of either Canadian official language (English, French) compared to that have knowledge of one or both languages. As explained in the exclusions section of this chapter (section 1.1.1), education and language are measured upon arrival in the CIC database (specifically assessed during the application process). These variables are likely to become increasingly misclassified with increasing duration of residence, particularly for those immigrants that came to Canada as children. To address this likely misclassification, women who were <15 years old at arrival were excluded from the analysis. The effect of these exclusions (minimal) are described in the sensitivity analysis sections (Tables A3.2.1, A4.2.1, A5.2.1) of each paper s appendix. Maternal country of birth (COB) is considered to be the most important indicator variable for perinatal health surveillance of immigrant women. 12 In order to take into account the differences that may be present between groups of women from the approximately 160 countries represented in the dataset and the subsequent non independence of outcomes between women from the same country of birth, this variable was conceptualized as a level in a hierarchical data structure. For all manuscripts, deliveries or births were nested within maternal COBs. Maternal COBs were also categorized according to the United Nations geographical regions or sub regions. 13 This regional categorization was used to further investigate effect modification, through stratification, in all manuscripts. Please see Appendix Table 2c for additional information on these variables and their categorizations. Country of last permanent residence (CLPR). In manuscript three, the CLPR on its own is conceptualized as a hierarchical variable in the same way that COB is used as a hierarchical variable in all manuscripts. Potential similarities among mothers that transitioned through the same CLPR were investigated in this manner for secondary migrants only. Secondary migration. In addition to the above variables, another variable was derived for manuscript three which involved combining information from the COB and CLPR. In the CIC database, CLPR is reported as any country that the applicant lived in for 6 months in the 5 years prior to applying for Canadian permanent residency. Individuals that indicated a CLPR that was different than their COB were described as secondary migrants in this manuscript. Individuals that indicated a CLPR that was the same as their COB were described as primary migrants. Personal communication with a representative of Citizenship & Immigration Canada 14 confirmed that refugee applicants who reside in a

51 33 refugee camp for the requisite amount of time prior to arriving in Canada would indicate the country of the camp as their CLPR. While this may not be the case for all secondary refugee immigrants, secondary migration for refugees may serve as an indicator for refugee camp, refugee camp like or other precarious living circumstances. Due to restrictions on the data released by the CIC regarding migration prior to immigration to Canada, secondary migration for both refugee immigrants and non refugee immigrants may be a misnomer as some secondary migrants may in fact be serial migrants who lived in more than one country in the five years preceding arrival in Canada and chose the most recent country to report to the CIC. Alternatively, there may be some primary migrants who lived in another country further than five years prior to immigrating to Canada and may have been secondary migrants in their distant past. Regardless of the inability to capture the complete picture of all migration events prior to arrival in Canada, secondary migration in this work does capture those who experienced at least one migration event that lasted for 6 months in the previous five years. Secondary migration was investigated as an effect modifier in manuscript 3 and operationalized as a dichotomous variable. Please see the Methodology Appendix (Table A2.2c) for details Selection biases associated with the CIC database The non immigrant group. All manuscripts included in this dissertation include comparisons made between refugee immigrants and non refugee immigrants and manuscript two compares nonsponsored refugee immigrants to sponsored refugee immigrants. For these analyses, there is very little, if any, misclassification of the groups compared, considering similar linkage rates among immigrant classes. 15 However, manuscripts one and two also include comparisons of refugee immigrants to nonimmigrants and this non immigrant group likely contains women who are actually immigrants. These immigrants classified as non immigrants fall into three groups: 1) those who immigrated prior to 1985 and are therefore not captured in the CIC database, 2) those CIC immigrants arriving between 1985 and 2010 who were not successfully linked to the RPDB and 3) those immigrants who initially migrated to another province and then re migrated to Ontario. With respect to 1), considering all manuscripts focus on deliveries or births occurring between 2002 and 2010 (and arrival 15 years of age), the probability of an immigrant arriving before 1985 and giving birth in this period is extremely small. This assertion is supported by examining Table 3.1 in manuscript one and Table 4.1 in manuscript two. Here we see that a small number of births occurred to women that arrived between 1985 and These numbers would likely be smaller for all those that migrated prior to 1985, but even if the numbers were the same (~9000) this would contribute 1% of all births in the non immigrant group (9000 misclassified births to immigrant women among ~862,000 births to non immigrant women). However since we can only make

52 34 estimates of bias using linkage values using the full CIC dataset ( ), I use historical immigration estimates to Ontario from Statistics Canada to liberally estimate ~39,000 immigrants arriving before 1985 gave birth between 1988 and In terms of 2) this group was described briefly in the previous paragraph. Given the information about interprovincial migration and the knowledge that about 26% of all CIC females gave birth in Ontario at least once between 1988 and 2010, I estimate that approximately 1.2% (~18,000) of mothers in the non immigrant group are actually unmatched CIC immigrants (see Methodology Appendix Table A2.3a and A2.3b, for estimates). Using the same report to develop an estimate for the number of interprovincial immigrants (re migrating to Ontario) (see Table A2.3c), I estimate that approximately another 2.5% (~35,000) of mothers in the non immigrant group are immigrant women who initially migrated to another part of Canada and then re migrated to Ontario. Combining estimates for all three groups, this means that approximately 6% of the nonimmigrant group are actually immigrants (see Table A2.3d). This proportion is likely to be smaller for manuscripts one and two considering that these analyses were restricted to deliveries/births occurring between while the above estimates are based on mothers giving birth between However, even if we double this estimate, assuming that 10% of the non immigrant group are actually immigrants, the bias to risk ratios comparing refugee immigrants to non immigrants reported in these manuscripts is virtually non existent (see Table A2.4 and A2.5). According to my estimation the reported values, SMM RR=1.378 (manuscript 1) and SNM=0.923 (manuscript 2), are biased very slightly towards the null compared to the true values, SMM RR=1.382 and SNM RR= Given the many assumptions regarding misclassification of the non immigrant group that I made (i.e., applying proportions that exist for groups of mothers to groups of births to those mothers, applying a correction factor for numbers published in the CIC interprovincial migration of immigrants report, using a constant rate of in migration to Ontario of immigrants from other provinces), it is possible that the extent of misclassification is larger. Even if this was the case, reported estimates in manuscripts one and two are biased towards the null and true RRs taking into account non immigrant group misclassification would exacerbate the effects seen (greater rate of SMM among refugee immigrants compared to non immigrants and lower risk of SNM among refugee immigrants compared to non immigrants). Ontario health insurance coverage. An important caveat of using the CIC database is that this database only includes immigrants who have obtained permanent residency status and were therefore eligible for health care coverage. On a related note, the analytical population in this dissertation includes a small proportion of refugee immigrant and non refugee immigrant women who had a

53 35 delivery or birth in Ontario before they received permanent residency status and were not yet eligible for Ontario Health Insurance. These events are picked up in the DAD since these individuals eventually became eligible for OHIP. For refugee immigrants, 98% of those giving birth before permanent residency were refugee claimants (rather than being government assisted, privately sponsored or dependent refugees) and should have been eligible for the Interim Federal Health (IFH) Program funded by the Government of Canada during the time of these deliveries and births, although this cannot be confirmed with the data available. In the appendix of each manuscript sensitivity analyses (Tables A3.2.2, A4.2.2, A5.2.2) are described which examine potential differences in risk between those that had OHIP coverage at the time of delivery/birth and those that did not and potential biasing of RRs. Overall there were few significant differences in risk between those with and without OHIP eligibility at the time of delivery or birth. In addition, there were no differences in RRs excluding those without OHIP eligibility at the time of delivery or birth compared to RRs including those without OHIP eligibility at the time of birth (as reported in the manuscripts). Cohort effects (the healthy immigrant effect). Given that the CIC database includes immigrant arrivals occurring over a 25 year time span ( ), there is the potential for cohort effects to influence the findings. Specifically of relevance are Canadian immigration policies that change entry criteria for refugee immigrants and non refugee immigrants. Two such policies are explored in secondary analyses for manuscript one. Firstly, starting in 1991, individuals with positive HIV/AIDS status were no longer considered a danger to public health and safety because of their positive status. 16 This meant that potential immigrants with HIV/AIDS could no longer be rejected based solely on those grounds. 16 The Immigration & Refugee Protection Act, 2002 (IRPA) built upon this 1991 legislation by instituting mandatory HIV/AIDS testing as a part of the compulsory Immigration Medical Exam (IME) that all permanent and temporary residence applicants are required to undergo. 17,18 Further, if the IME found that HIV disease was of a severe form, this could place excessive demand on health and social services and was considered adequate grounds for rejecting an applicant. 19 However, IRPA also placed greater emphasis on the need to protect refugee immigrants and thus refugee immigrants were exempt from rejection based on severe HIV disease, as well as other diseases that were considered to put excessive demand on health and social services. Since immigration policies, such as those described above, can create health disparities between refugee immigrants and non refugee immigrants at arrival times coinciding with immigration policies, it is important to examine potential cohort effects. This will add to our understanding of the healthy immigrant effect, whether it applies to refugee immigrants 20 and whether there are changes over time. In manuscript one, results comparing the rate of SMM

54 among refugee immigrants to non refugee immigrants and to non immigrants were stratified by period of immigration ( , , ) Discharge Abstract Database (DAD) The DAD originates from the Canadian Institute for Health Information (CIHI) and contains both mother and infant hospital admission records (delivered by any health care provider i.e., obstetrician, midwife, family physician), including information on diagnoses and procedures. The DAD was used to derive a perinatal specific dataset created by linking maternal and infant hospitalization records, resulting in maternal infant delivery pairs. This linkage was conducted in two stages: for births taking place between the fiscal years 1988/89 and 2001/02 and for those taking place between fiscal years 2002/03 and 2009/ The earlier dataset was developed by linkage between a maternal dataset and a newborn dataset based on matching: 1) institution (hospital) number, 2) newborn birth date falling between the mother s admission and discharge dates and 3) postal codes. The yearly average percent matched for this period was 92% for maternal records and 85% for newborn records. Births and deliveries taking place between 2002/03 and 2009/10 were linked based on: 1) institution (hospital) number, 2) newborn birth date falling between the mother s admission and discharge dates and 3) chart numbers (preferred) or postal code. The yearly average percent matched between 2002/03 and 2008/09 was 97% for maternal records and 94% for newborn records. 10 The data from this latter period was used in this dissertation to take advantage of the 10 th revision of the International Statistical Classification of Diseases and Related Health Problems, Canadian enhancement, and Canadian Classification of Health Interventions (ICD 10 CA/CCI) which has been determined to be more comprehensive and more specific than the previous coding system (ICD 9CM). 21 Using this period also allowed us to take advantage of routinely collected gestational age reported in completed weeks and self reported parity (continuous) which were not previously recorded. All health outcomes for the three manuscripts included in this dissertation were derived from the DAD Measuring maternal health Manuscript 1. Maternal mortality has traditionally been used to monitor the health of mothers. Maternal mortality is indicative of the status of women, their access to health care and the capacity of the health care system to respond to their needs. 22 However in industrialized countries, the numbers of maternal deaths is very low. 23 In Canada between 1997 and 2000 there were 64 direct and indirect maternal deaths with a maternal mortality ratio of 6.1 per 100,000 births. 22 Given such low numbers and rates, particularly with further stratification (e.g., by

55 37 province or cause of death) one can see that maternal mortality is limited in its ability to serve as a basis to improve maternal health. 24 As an alternative to maternal mortality there has been increasing emphasis on examining nearmiss or severe maternal morbidity. A World Health Organization working group defined maternal nearmiss morbidity as a woman who nearly died but survived a complication that occurred during pregnancy, childbirth or within 42 days of termination of pregnancy. 25 In practice, there is no international consensus on what constitutes a severe maternal morbidity but typically includes disease (i.e., severe post partum hemorrhage), intervention (i.e., blood transfusion) or organ system (i.e., abruptio placentae) specific entities. 26 Many countries have developed their own severe maternal morbidity measures based on single or mixed criteria that have shown risks between 0.79% in Europe and 1.38% in North America. 24 Measuring severe maternal morbidity overcome the problem of low counts and rates that are typically associated with maternal mortality, making it more suitable for evaluating the general health of mothers. The Canadian Perinatal Health Surveillance system at the Public Health Agency of Canada (PHAC) developed a composite (mixed) severe maternal morbidity (SMM) indicator that uses over 40 disease, intervention and organ dysfunctions (based on ICD 10 CA/CCI codes) reported in the labour or delivery record of the DAD. 27 This measure was further validated by identification of temporal trends and regional disparities in SMM as well as specific maternal illnesses using DAD data covering the period The overall risk of SMM in Canada (excluding Quebec) was found to be 1.38%. 27 Please see the Appendix of manuscript one (Table A3.1.1) for a list of diagnosis and intervention descriptions and codes included in the SMM indicator. Operationalizing SMM and SMM subtypes. This SMM indicator, used in manuscript one of this dissertation, was present when one or more of the ICD 10 CA/CCI diagnosis or procedure codes was indicated on the delivery record. The SMM indicator and the subtypes that are included as a part of the indicator were all operationalized as dichotomous variables (present or absent) (See Table A2.2b) Measuring neonatal health Manuscript 2. There are a number of health outcomes that are typically used to evaluate the health of newborns. These outcomes may include low birth weight, preterm birth, small for gestational age and perinatal mortality. However these outcomes do have limitations. All of these outcomes are indicators for numerous co morbid health conditions (e.g., preterm birth may be indicative of feeding problems) with a range of severities, without having the ability to be more specific about which related health conditions can be acted upon for prevention

56 38 purposes. In this way a composite outcome examining severe neonatal morbidities is beneficial since it includes severe conditions and it can be disaggregated into its components. No gold standard exists for the measurement of severe neonatal morbidity and the measures that do exist have limitations. The use of neonatal intensive care unit (NICU) admission is often used 29 but this measure has limited generalizability, since it captures only a select group of the sickest neonates requiring a very high level of care that is not available in all hospital settings. 30 Other validated measures depend on clinical and laboratory data that are not captured in population based hospitalization records and therefore not readily estimated. 30,31 To overcome many of the issues described above, a neonatal morbidity indicator that enumerates multiple severe neonatal outcomes simultaneously based on hospitalizations data is both practical and efficient while still being informative for prevention efforts. In 2012, Lain et al, 30 published findings based on development of such an indicator, referred to as the Neonatal Adverse Outcomes Indicator (NAOI). The authors found that their composite indicator was predictive of infant mortality and hospital readmission in New South Wales, Australia. This outcome was based on birth admissions in hospitalization and midwifery data and coded using the ICD 10 coding system for diseases and interventions for Australian data. Considering the relative similarities of the Australian and Canadian health care systems (universal, single payer systems) and the coding system, an adapted version of this indicator was possible. I collaborated with an expert clinician (Dr. Joel Ray St. Michael s Hospital, Toronto) and a general surgeon (Dr. David Urbach Toronto General Hospital, Toronto) to adapt the indicator for use with Canadian hospitalizations data. This adapted version is referred to as the Severe Neonatal Morbidity (SNM). This outcome was used to evaluate the health of infants born to refugee immigrant women in manuscript two. A list of the diagnoses and interventions included in the SNM indicator is included in the appendix of manuscript two (Table A4.1.1). In terms of criterion (predictive) validity, Lain et al reported that infants with NAOI were 10 times more likely to die in the first year of life than infants without NAOI. Modifications to NAOI for SNM. In the process of adapting diagnoses and procedures of the NAOI indicator some important modifications were made. Firstly, Lain et al included both singletons and multiples in the evaluation of the NAOI indicator. Given the much higher likelihood of severe morbidities associated with multiple births, it was decided that for the purposes of reporting that multiples would be excluded from the Canadian indicator. In addition, neonatal mortality was included in the formulation of the NAOI. Since one of the purposes of such an indicator is to evaluate the ability of the indicator to predict perinatal and infant death, it was deemed including neonatal mortality in the

57 39 SNM would be inappropriate because of co linearity. Lastly, adult codes for sepsis and pneumonia (those not starting with P, or the adult codes) were not used in the formulation of the SNM indicator as it came to my attention that these are not commonly recorded for newborns in Canada. Operationalizing SNM and SNM subtypes. The SNM indicator was used in manuscript two of this dissertation. The dichotomous SNM indicator was operationalized as any singleton live birth with one or more of the ICD 10 CA/CCI diagnosis or procedure codes recorded at the birth admission (See Table 2.2b). All births with a gestational age <22 weeks or birth weights < 500 grams were excluded. Supplementary and sensitivity analyses. Concurrent validity was assessed by conducting a side byside comparison of the SNM and its subtypes with the NAOI and its subtypes as published in Lain et al (2012). This assessment can be found in the appendix of manuscript 2 (Table A4.1.2). For the purposes of the comparisons in this table, multiple births were included but neonatal deaths remained excluded. Personal communication with Dr. Samantha Lain (lead author of the Australian study) indicated that risk of ventilator support should be 2.77% rather than the 1.99% reported in their manuscript, 32 therefore the overall NAOI reported in this table does not consider the higher rate of this subtype. Keeping in mind these differences, the risk of SNM among all singleton and multiple hospital births in Ontario was 5.43% (excluding adult codes for sepsis and pneumonia) between 2002 and 2010 while the rate of NAOI was only slightly lower at 4.59% among singleton and multiple hospital and midwifery births in New South Wales, Australia between 2001 and 2006 (15% difference). Comparison of individual subtypes (with the caveat of possible under ascertainment in both datasets) revealed reasonable similarity, with the majority of values for the NAOI subtypes within 30% of the values for the SNM subtypes. Overall this comparison suggests reasonable concurrent validity. There were larger differences in: resuscitation with intubation or CPR, ventilator support, sepsis and central line. The differences seen in interventions likely result from different intervention codes being used in Canada and Australia but also could reflect different medical practices. The difference in sepsis could be due to reporting errors because of the need for rapid treatment before confirmation with lab testing is available. Dr. Jonathan Hellmann (Sick Kids Hospital, Toronto) reviewed the frequencies for SNM and the SNM subtypes and did not suggest that any were out of the ordinary. Criterion validity, more specifically predictive validity, was also assessed. The SNM (including multiples to facilitate comparison to the NAOI, neonatal mortality still excluded) performed substantially better than the NAOI indicator, with infants with SNM being 98 times more likely to die in

58 40 the first 28 days and 41 times more likely to die in the first year of life than those infants without SNM. See the appendix of manuscript two for these results (Table A4.1.3). Since there is no gold standard for measuring severe neonatal morbidity construct validity of the SNM indicator could not assessed. The other measures that would be relevant for severe neonatal morbidity are incorporated into SNM measure as SNM subtypes (i.e., very preterm, birth weight <1500, sepsis) therefore examining agreement with these alternative measures of severity would be misleading Preterm birth Manuscript 3. Preterm birth (PTB) is traditionally defined as any birth occurring at less than 37 weeks gestational age. This outcome was determined using the gestational age indicated on the DAD hospitalization record. Gestational age has been recorded in the DAD in completed weeks since 2002, and is based on the best clinical estimate in a woman s medical chart. Best clinical estimates of gestational age in the DAD are based on either the first day of the mother s last menstrual period (LMP) or ultrasonography. Due to the often inaccurate assumptions associated with estimating LMP including the length of a woman s menstrual cycle (28 days) and when ovulation occurs in the cycle (14 th day), ultrasonography is considered more precise 33 with gestational age estimates accurate within 3 5 days if ultrasound is done in the first trimester and accurate within 1 week if done in the second trimester. 26 The use of prenatal ultrasonography is the dominant method used for pregnancy dating in Ontario 4 with annual rates of ultrasonography utilization steadily rising between 1996 and The adjusted relative risks and 95% confidence intervals comparing prenatal ultrasound utilization rates in 2006 to 1996 were RR=1.98 ( ) for the first trimester, and RR=1.38 ( ) for the second trimester. The rate of first and second trimester prenatal ultrasound use rose from 1.4 per pregnant woman in Ontario in 1996 to 2.2 per pregnant woman in Operationalizing PTB. Due to the fact that infant health outcomes and survival differ across the range of preterm gestational ages and diverse causal mechanisms are implicated, investigation of PTB sub groups based on gestational age has been recommended. 26 In recognition of this, preterm birth was measured as an ordinal three category variable in manuscript three of this dissertation, specifically very preterm (22 31 weeks), moderately preterm (32 36 weeks) and full term ( 37 weeks). All births with birth weights < 500 grams were excluded.

59 Other variables available in the DAD that were used in all manuscripts were: Maternal age at delivery or birth was calculated based on maternal date of birth and available on the admission record. Maternal age was a confounder of the relationship between refugee status and each of the three health outcomes examined. This variable was entered as a 6 level categorical variable in all analyses: 15 19, 20 24, (reference category), 30 34, 35 39, 40 years. Parity was self reported by the mother and refers to how many live children were born prior to the index delivery or birth. Parity is an important confounder of the relationship between refugee status and each health outcome examined in all three manuscripts. Parity is recorded as a continuous variable in the DAD but given the skewed distribution (most women had none or one previous live birth), the variable was categorized into a nominal four level variable with categories: none, 1, 2 and 3 previous live births. The reference category was 1 previous live birth for all manuscripts. These reference categories were chosen based on the category that had the lowest risk of the outcome. Duration of residence was measured for all refugee immigrants and non refugee immigrants and was calculating by subtracting the date of permanent residency recorded in the CIC database from the date of delivery or the date of birth. This variable was conceptualized as a confounder of the relationship between refugee status and the outcomes and included as a continuous variable in all analyses when comparing refugee immigrants to non refugee immigrants Validity of perinatal health outcomes using hospitalizations data. In 2009, Joseph & Fahey conducted a study to validate the use of CIHI hospitalization data for perinatal research. In this study the DAD was compared to a Nova Scotian perinatal database which was known to have a high degree of accuracy. The authors reported that major procedures (e.g., caesarean delivery) were coded accurately. Information on more minor diagnoses (e.g., perineal lacerations) and more challenging diagnoses (e.g., induction) were also reasonably accurate. Disagreements found between the perinatal database and the DAD were mostly for diagnostic entities which could be defined in a number of ways. However, the authors found that accuracy could be increased by using both neonatal and adult codes as well as restricting to more severe forms. Overall, this study supported the use of the DAD for perinatal research in Canada. More generally, a recent systematic review of the quality of data in perinatal population based databases, which included Canadian studies, found that there were few false positives and high positive predictive values particularly for more common conditions. 35

60 42 However, under ascertainment of specific conditions and procedures is a potential concern for the use of hospitalizations data for perinatal research. Potential under ascertainment is addressed in this dissertation in three ways. Firstly, outcomes examined in manuscripts one and two are severe indicators of maternal and infant morbidity respectively. It has been suggested that in addition to being more accurately reported, 33 severe conditions are less likely to suffer from under reporting. 35 Secondly, the measures of both severe maternal and neonatal morbidity are composite indicators. Operationally, this means that definitions of the severe maternal morbidity and severe neonatal morbidity indicators include, respectively, approximately 40 and 20 individual diagnostic entities or procedures. This is important since a mother or infant with severe illness is likely to have co morbidities and by using numerous diagnostic and procedure codes the chances of identifying all individuals with illness is increased. Lastly, these indicators include both diagnoses and procedures. Considering that procedures are more accurately reported than diagnoses, 33 and routinely have high positive predictive values and few false positives, 35 including procedural codes also increases the chances of identifying mothers and babies with severe illness and reduces under ascertainment Selection bias associated with use of the DAD. Since health care in Canada operates under a universal health care system and almost all births in the province of Ontario take place in hospital with an obstetrician or a midwife in attendance, there is very little selection bias associated with using data on hospital births available in the DAD. In Canada it is estimated that less than two percent of total births do not take place in hospital and the largest subset of these are low risk pregnancies attended to by midwives in the home. 36 It has been suggested that immigrant women in Canada have a stronger preference for hospital birth than Canadian women. This difference in preference may cause outcomes in the non immigrant comparator in our studies to have slightly higher risks than if the low risk home births occurred in hospital and were included in the DAD. This may cause attenuated RRs when comparing outcomes of refugee immigrants to non immigrants (manuscripts one and two), however considering the size of the non immigrant population giving birth in hospital, this attenuation will be minimal.

61 2.3 STATISTICAL METHODS AND DATA ANALYSIS Descriptive Analysis A similar approach to data analysis was taken for all three manuscripts included in this dissertation. Analytical work began with univariate and bivariate descriptions. Univariate statistics were conducted for all variables included in these analyses (exposure, outcomes and all covariates) to examine the distributions, outliers and the extent of missingness. This included examining frequencies for categorical variables and means, standard errors and minimum and maximum values for continuous variables. The very few unlikely values that were found were reset to missing (i.e., maternal age at 103 years). Less than 0.5% of refugee immigrant and non refugee immigrant deliveries or births were excluded because of missing data, whereas less than 1% were excluded among non immigrant deliveries or births. To examine associations, bivariate analyses were conducted between all pairs of variables included in each manuscript. This included Pearson 2 tests for pairs of categorical variables, t tests for continuous variables with dichotomous variables and ANOVAs for continuous variables with categorical variables (p<0.05). If continuous variables were not normally distributed, nonparametric equivalents of the t test and ANOVA (Wilcoxon test and nonparametric ANOVA, respectively) were conducted Multivariate Analysis Choice of regression model Log binomial regression. For manuscripts one and two, log binomial regression was chosen to estimate rate ratios for SMM and SNM risk ratios, which compare refugee immigrants to nonimmigrants and refugee immigrants to non refugee immigrants. This regression model was chosen since cases were being modeled to estimate relative rates or the relative risk of the outcome. Alternatively, logistic regression could also be used since odds ratios approximate risk ratios with rare outcomes (<10%) which is the case with both SMM and SNM. However since neither of these studies have case control designs, where logistic regression is the most suitable due to the necessity to estimate odds rather than risk, it was decided log binomial regression was the most appropriate regression method to use. One of the main criticisms of the log binomial model is convergence problems. Therefore model convergence was closely monitored for all regression models in manuscripts one and two. For those few models where convergence was not achieved, a modified Poisson model with robust error variances was used instead. When the prevalence of the outcome is

62 44 low and the sample size is large, as is the case with manuscripts one and two, probabilities from the Poisson distribution can be used to approximate the probabilities from the binomial distribution. 37 Coefficients for all variables included in the main models examined in manuscripts one and two can be found in the Appendix of each manuscript (Tables A3.1.2 to A3.1.5 & A4.1.4 to A4.1.6). Cumulative odds regression. For manuscript three, generalized linear models were used estimate the cumulative odds ratios (COR) for preterm birth comparing refugee immigrants to non refugee immigrants. The cumulative probability of preterm birth was estimated in this manuscript since preterm birth was operationalized as an ordinal three category variable: weeks GA (very PTB), weeks GA (moderate PTB) and 37 weeks GA (full term). Ordinal models are considered more efficient than dichotomous models since all information can be fully utilized which increases statistical power. Dichotomizing preterm birth, as is typically done, results in a loss of information because of the collapsing and ignoring of relevant categories. 38,39 There are a number of ways to interpret a COR generated from these models: 1) the cumulative probability of the outcome, starting at the extreme end is significantly smaller/greater among the exposed compared to the unexposed; 2) the exposed group is more likely to have a higher probability of the adverse categories of the outcome compared to the unexposed group; or 3) compared to the unexposed group, the odds of having the outcome rather than not having he outcome is X% higher for the exposed group. An important assumption that must be evaluated when using a cumulative odds model is the proportionality of the odds ratios between ordered aggregations of the outcome. In order to assess this, two binary logistic regression models were estimated, one estimating an odds ratio for very PTB (22 31 wk GA vs. the other categories) and another estimating an odds ratio for PTB (<37 week vs. full term).. If these ORs were reasonably similar, the cumulative odds models were deemed appropriate for use with the data. In addition to the proportionality test, the Score test was also estimated in a cumulative odds model with the null hypothesis indicating that the odds were proportional. A nonsignificant Score test (p 0.05) suggested that the null hypothesis could not be rejected and that the proportionality assumption was met. 39 Both tests of the proportionality assumption were conducted for all analyses in manuscript three.

63 45 Coefficients for all variables included in the main models examined in manuscript three, and not included in the main tables, can be found in the Appendix of this manuscript (Tables A5.1.1, A5.1.2) Random effect (multi level) models. In all three manuscripts, RRs or CORs were estimated for refugee immigrants compared to non refugee immigrants. For these regression models, maternal COB was evaluated as a potentially important contributor to total variation in health outcomes. Maternal COB was included as a random effect (or 2 nd level individual deliveries or births nested within maternal COB) in regression models. To determine whether deliveries and births for mothers from the same COB were significantly correlated, the statistical significance of a randomintercept for country of birth in an intercept only model (no explanatory variables) was determined. If the covariance parameter estimate for the random intercept divided by its standard error was greater than 1.96, incorporating maternal COB as a random effect into the regression models was deemed justified. Further, if the random intercept was significant the intra class correlation coefficient (ICC) was estimated for each model. The ICC can be interpreted as either: 1) the proportion of the total variance in the outcome that exists at the group level, or 2) the expected correlation between two randomly drawn units that are in the same group. 40 The ICC for dichotomous or ordinal outcomes is calculated using the following formula, ICC = /( + 2 /3) where is 3.14 and is the covariance parameter estimate for the group level intercept. 40,41 In manuscript three, country of last permanent residence (CLPR) was also explored as a potentially important contributor to the cumulative probability of preterm birth using cumulative odds regression with a random effect restricted to secondary migrants (without COB as a 2 nd level). The same process was followed as that described above to determine the importance of the CLPR. Alternatives to random effects models. Generalized Estimating Equations (or GEEs) can also be used to take into account potential non independence of outcomes of mothers from the same COB or CLPR. However random effect models offer much more flexibility than GEEs, including the ability to estimate parameters such as the ICC and handle more than two levels of clustering. Given this flexibility and the greater efficiency of maximum likelihood estimation over GEE methods, multi level models were used in all manuscripts. 40 Deliveries or births within mothers. An additional methodological issue related to hierarchies is that in the analytic dataset used for all manuscripts approximately 30% of all deliveries and births have

64 46 one or more sibling present in the database. This means that there may be correlation among deliveries and births to the same mother that is unaccounted for. Buck Louis et al (2006) 42 published a discussion of the methodological challenges of analyzing repeated pregnancy outcomes. The authors compare different approaches including analyzing the first pregnancy only or using more complex models that address the correlation directly. It is suggested that if the research goal is to detect subtle effects (arguably like the effect of refugee status), models that build in correlation are necessary. To examine the impact of clustering of deliveries and births within mothers instead of clustering of deliveries and births within maternal countries of birth (as was done in all manuscripts), sensitivity analyses were conducted for each manuscript. These results are outlined in the Appendix of each manuscript (Tables A3.2.3, A4.2.3, A5.2.3). Unadjusted risk ratios were compared for models with no hierarchy, the maternal country of birth hierarchy (as reported) and the maternal hierarchy. Models including both the maternal country of birth and the mother were too complex given the large number of mothers in the dataset, so were not included as a comparison in these sensitivity analyses. For manuscripts one and two, there were no differences in the RRs when taking into account clustering within mothers compared to clustering within maternal countries of birth. For manuscript three, PTB ORs (not cumulative ORs for sensitivity analysis) for primary migrants were lower and non significant for the model taking into account clustering within mothers when compared to the model taking into account clustering within maternal countries of birth. For secondary migrants there was no difference between the two models. Given very similar findings between clustering approaches, deliveries and births clustered within maternal countries of birth was retained for all manuscripts Adjusting for confounders. Conceptualization of covariates as confounders is described in the Data Sources, variables, measurement and potential biases section of this chapter (above) as well as summarized for each manuscript in Table A2.2c in the Methodology Appendix. All confounders were included in regression models and no formal examination of confounding was conducted since all variables are suspected to be important determinants of perinatal health. Supplementary analyses. In manuscript three, I also examined whether the significant modification of the relationship between refugee status and preterm birth by secondary migration could be explained by maternal co morbidities. These maternal co morbidities were reported in the mother s DAD birth admission record and were categorized into systemic morbidities (i.e., hypertension, diabetes) and localized morbidities (i.e., hemorrhage, cervical incompetence). A list of these maternal co morbidities and their ICD 10 CA codes can be found in the appendix of manuscript

65 47 three (Table A5.1.3). The risk of each maternal co morbidity was estimated for those primary and secondary refugee and non refugee immigrants who delivered preterm. For each co morbidity risk ratios and 95% confidence intervals comparing refugee immigrants to non refugee immigrants were also estimated. These results can be found in the appendix of manuscript three (Table A.5.1.4) of manuscript three Subgroup analysis & interactions. Stratified risk ratios (refugee immigrants vs. nonrefugee immigrants) were estimated to explore the potential confounding or modifying effects of maternal region of birth in manuscripts one and two. These results can be found in the appendices of these manuscripts (Table A3.1.7 and Table A4.1.7, respectively). In the main analysis of manuscript three, RRs for Asian and Sub Saharan African immigrants were also estimated to explore potential confounding or modifying effects by maternal region of birth (Table 5.3). In manuscript one, this was also explored for period of immigration by stratification of RRs, the results of which can be found in this manuscript s appendix (Table A3.1.6). The main objective of manuscript three was to determine whether the relationship between refugee status and preterm birth was modified by secondary compared to primary migration. Since, in manuscript three, the relative odds (specifically, the cumulative odds for an ordinal outcome) was the effect measure of choice rather than additive risk, the presence of a more than multiplicative interaction between refugee status (refugee immigrant vs. non refugee immigrant) and secondary migration was tested by adding an interaction term between these two variables in the regression model. If the coefficient on the interaction term was statistically significant (p<0.05), significant multiplicative interaction was deemed to be present and the interaction term was retained for all models. For interest sake, assessment of a more than additive interaction was also assessed using the crude homogeneity of effects method summarized by Szklo & Nieto (2007). 43 By examining crude preterm birth risk for each of the four groups making up the interaction between refugee status and secondary migration, it was found that the observed joint additive effect of refugee and secondary migration status was greater than the expected joint additive effect, suggesting that a more than additive interaction was also present.(results not shown) In all models for which refugee immigrants are compared to non refugee immigrants, the presence of an additional two variable interaction between education level and duration of residence was explored. This interaction was tested since these variables are particularly relevant for immigrant

66 48 health outcomes and it was hypothesized that a synergistic effect may result when levels of the two variables are present simultaneously. For the sake of keeping the models as parsimonious as possible, these interaction terms were retained only if the interaction term was statistically significant (p<0.05) in a full model and there was a substantial change in main effect estimates and interpretation. Tables summarizing the significance of the interaction terms and its influence on the main effects are included in the appendix of each manuscript (Tables A3.1.8, A4.1.8, A5.1.6) Multiple comparisons In manuscripts one and two analyses were conducted which estimate a series of RRs for each subtype of the composite indicators as well as for RRs stratified by maternal region of birth. Since multiple tests were conducted simultaneously, statistically speaking there is a possibility that some significant findings may be false positives. In such situations, an adjustment for multiple comparisons is often suggested to compensate for findings that could be due to chance. The downside of this correction is that while false positives may be reduced, false negatives may be created. Rothman (1990) 44 suggests that adjustment for multiple comparisons trivializes the interpretive problem of what scientists should do with unanticipated findings particularly when the finding is one of many and are not the main focus of the study. Instead of missing potentially important and interesting findings (by creating false negatives through adjustment for multiple comparisons), Rothman suggests all that needs to be done is to further investigate unexpected findings. Consequently, this exploratory approach was taken for manuscripts one and two. It is important to recognize that by examining subtypes and regions of maternal birth, a number of independent questions are being asked. If there is an unexpected finding, it seems less important that these findings might be due to chance and more important to examine whether the findings are meaningful and deserve closer scrutiny. For these reasons, adjustments for multiple comparisons were not conducted herein Sensitivity analyses A number of sensitivity analyses have been described in detail in other parts of this methods chapter. These were conducted to explore the impact of methodological decisions on the results reported in the manuscripts. These analyses were common to all three manuscripts. Analyses included examining the impact of: i) age at arrival exclusions, ii) immigrant births that occurred before OHIP eligibility (before achieving permanent residency) and iii) deliveries and births clustered within mothers. The results of these sensitivity analyses can be found in the appendix of each chapter (Sections A3.2, A4.2, A5.2).

67 REFERENCES Elm EV, Altman DG, Egger M, et al. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ Oct 20;335(7624): Beck A, Corak M, Tienda M. Age at immigration and the adult attainments of child migrants to the United States. Ann Am Acad Pol Soc Sci Sep 1;643(1): Adamuti Trache M, Sweet R. Adult immigrants participation in Canadian education and training. Can J Study Adult Educ. 2010;22(2): Urquia M, Frank J, Moineddin R, Glazier R. Immigrants duration of residence and adverse birth outcomes: a population based study. BJOG Int J Obstet Gynaecol Apr;117(5): Creatore MI, Moineddin R, Booth G, et al. Age and sex related prevalence of diabetes mellitus among immigrants to Ontario, Canada. Can Med Assoc J. 2010;182(8): Lofters A, Moineddin R, Hwang S, Glazier R. Predictors of low cervical cancer screening among immigrant women in Ontario, Canada. BMC Womens Health. 2011;11(20). 7. Urquia ML, Frank JW, Glazier RH. From places to flows. International secondary migration and birth outcomes. Soc Sci Med. 2010;71(9): Urquia ML, Frank JW, Moineddin R, Glazier RH. Does time since immigration modify neighborhood deprivation gradients in preterm birth? A multilevel analysis. J Urban Health. 2011;88(5): Institute for Clinical and Evaluative Sciences. Source of matched data in the CIC linkage [Internet] [cited 2014 Jun 12]. Available from: Institute for Clinical and Evaluative Sciences Intranet 10. Fung K. Methods for creating/updating MOMBABY file [Internet] [cited 2012 Aug 31]. Available from: Institute for Clinical & Evaluative Sciences Intranet. 11. Okonny Myers I. The interprovincial mobility of immigrants in Canada [Internet]. Citizenship and Immigration Canada; 2010 Jun. Available from: mobility.pdf

68 Gagnon AJ, Zimbeck M, Zeitlin J. Migration and perinatal health surveillance: An international Delphi survey. Eur J Obstet Gynecol Reprod Biol Mar 1;149(1): United Nations Statistics Division. Composition of macro geographical (continental) regions, geographical sub regions, and selected economic and other groups [Internet] [cited 2014 Apr 3]. Available from: Tara Gilkinson, CIC representative. Personal communication Re: country of last permanent residence Jul. 15. Cernat G, Wall C, Iron K. Initial Validation of Landed Immigrant Data System (LIDS) with the Registered Person s Database (RPDB) at ICES Klein A. HIV/AIDS and immigration final report [Internet]. Canadian HIV AIDS Legal Network; Available from: Bisaillon LM. Human rights consequences of mandatory HIV screening policy of newcomers to Canada. Health Hum Rights Int J. 2010;12(2). 18. Zencovich M, Kennedy K, MacPherson DW, Gushulak BD. Immigration medical screening and HIV infection in Canada. Int J STD AIDS. 2006;17: Public Health Agency of Canada. Population specific HIV/AIDS status report. People form countries where HIV is endemic black people of African and Caribbean descent living in Canada. [Internet] Available from: aspc.gc.ca/aids sida/publication/pspd/africacaribbe/pdf/ps spreport eng.pdf 20. DesMeules M, Gold J, Kazanjian A, et al. Immigrant health assessment: Profiling health diversity in Canada. Can J Public Health. 95(3): Canadian Institute for Health Information. Benefits of ICD [cited 2014 Jul 1]; Available from: extportal/internet/en/document/standards+and+data+submission/standards/classification+and+cod ing/codingclass_icd10bene 22. Health Canada, Canadian Perinatal Surveillance System. Special report on maternal mortality and severe morbidity in Canada [Internet] Available from: E.pdf

69 23. World Health Organization. Trends in maternal mortality: 1990 to 2008 [Internet] Available from: Tunçalp ö, Hindin M, Souza J, Chou D, Say L. The prevalence of maternal near miss: a systematic review: Prevalence of near miss: systematic review. BJOG Int J Obstet Gynaecol May;119(6): Khan KS, Wojdyla D, Say L, Gülmezoglu AM, Look PF Van. WHO analysis of causes of maternal death: a systematic review. The Lancet Apr;367(9516): Buck Louis GM, Platt RW. Reproductive and perinatal epidemiology. New York, New York: Oxford University Press; Joseph KS, Liu S, Rouleau J, et al. Severe maternal morbidity in Canada, 2003 to 2007: surveillance using routine hospitalization data and ICD 10CA codes. J Obstet Gynaecol Can. 2010;32(9): Liu S, Joseph KS, Bartholomew S, et al. Temporal trends and regional variations in severe maternal morbidity in Canada, 2003 to J Obstet Gynaecol Can. 2010;32(9): Claydon JE, Mitton C, Sankaran K, Lee SK. Ethnic differences in risk factors for neonatal mortality and morbidity in the neonatal intensive care unit. J Perinatol. 2007;27(7): Lain SJ, Algert CS, Nassar N, Bowen JR, Roberts CL. Incidence of severe adverse neonatal outcomes: use of a composite indicator in a population cohort. Matern Child Health J Apr;16(3): Strobino D, Johns Hopkins Bloomberg School of Public Health. Measures of neonatal morbidity [Internet] [cited 2014 Apr 5]. Available from: Lain SJ. Personal communication Re: NAOI indicator Jun. 33. Joseph KS, Fahey J. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute for Health Information. Chronic Dis Can. 2009;29(3): You J, Alter D, Stukel T, et al. Proliferation of prenatal ultrasonography. CMAJ. 2010;182:

70 Lain SJ, Hadfield RM, Raynes Greenow CH, et al. Quality of data in perinatal population health databases. Med Care. 2012;50(4):e7 e Danielle Bachove. Are home births safe? McLeans [Internet] Aug 26 [cited 2014 Apr 3]; Available from: try this at home/ 37. Petersen MR, Deddens JA. A comparison of two methods for estimating prevalence ratios. BMC Med Res Methodol. 2008;8(9). 38. Ananth CV, Kleinbaum DG. Regression models for ordinal responses: a review of methods and applications. Int J Epidemiol. 1997;26(6): Agresti A. Categorical data analysis. 3rd ed. Wiley; Hox JJ. Multilevel analysis techniques and applications. 2nd ed. Great Britain: Routledge; Snijders TA, Bosker R. Multilevel analysis: an introduction to basic and advanced multilevel modeling. 2nd ed. SAGE Publications Ltd; Louis GB, Dukic V, Heagerty PJ, et al. Analysis of repeated pregnancy outcomes. Stat Methods Med Res Apr 1;15(2): Szklo M, Nieto J. Epidemiology: Beyond the Basics. Jones & Bartlett Learning; Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology. 1990;1(1):43 46.

71 APPENDIX Table A2.1: Methodological overview summary of manuscripts included in this dissertation Manuscript (Chapter) 1(3) 2(4) 3(5) Study setting & design Ontario Populationbased database study, retrospective exposure assessment Ontario Populationbased database study, retrospective exposure assessment Ontario Retrospective populationbased cohort Data sources and dates DAD: April 2002 March 2011 CIC: January 1985 December 2010 DAD: April 2002 March 2011 CIC: January 1985 December 2010 CIC: April 2002 December 2010 DAD: April 2002 March 2011 Inclusion Criteria All deliveries (multiple births counted as 1) April 2002 March 2011, born to immigrants arriving and all nonimmigrants All singleton live births April 2002 March 2011, born to immigrants arriving and all nonimmigrants All singleton live births April 2002 and March 2011, born to immigrants arriving and all nonimmigrants Exclusion Criteria 1. Maternal age at arrival: <15 years 2. Maternal age at birth: <15 years 1. Maternal age at arrival: <15 years 2. Maternal age at birth: <15 years 1. Maternal age at arrival: <15 years DAD: Discharge Abstract Database; CIC: Citizenship & Immigration Canada Outcome Severe maternal morbidity Severe neonatal morbidity Preterm birth Comparison groups (n) 1. Refugee (n=30 381) vs. Non immigrant (n= ) 2. Refugee vs. Nonrefugee (n= ) 1. Refugee (n=29 765) vs. Non immigrant (n= ) 2. Refugee vs. Nonrefugee (n= ) 3. Non sponsored refugee (n=15 045) vs. sponsored refugee (n=10 519) 1. Primary refugee (n=11 671) vs. primary nonrefugee (n= ) 2. Secondary refugee (n=1 300) vs. secondary nonrefugee (n=9 773) Confounding (C) & mediating (M) variables Both comparisons: C maternal age, parity Only comparison 2: C education, language, duration of residence. All comparisons: C maternal age, parity. Comparison 2 & 3: C education, language, duration of residence. Both comparisons: C maternal age, parity, education, language, duration of residence 53

72 54 Table A2.2: List of exposures, outcomes and covariates Abbreviations: DAD = Discharge Abstract Database of the Canadian Institute for Health Information; CIC=Citizenship and Immigration Canada Database; RPDB=Registered Persons Database A2.2a. Exposures Exposure Refugee status Refugee sponsorship status* Sponsored refugees: supported by the Govt of Canada with travel to Canada, social support benefits, have access to health care upon arrival Non sponsored: arrive on their own, access to Interim Federal Health care, potential access to provincial funding for supports Source Database CIC Variable: FCLAS; RPDB CIC Variable: FCLAS Definition Refugee immigrant, nonrefugee immigrant or nonimmigrant Codes 31=Governmentassisted refugees 32=Privately sponsored refugees 33=Refugees landed in Canada 34=Refugee dependents All other FCLAS codes 31=Governmentassisted refugees 32=Privately sponsored refugees 33=Refugees landed in Canada Operationalization for description and analyses Two comparisons: 1. Refugee vs. Non Refugee 1=Refugee (codes 31 34) 0=Non Refugee (all other fclas codes) 2. Refugee vs. Non Immigrant 0=Refugee (codes 31 34) 1=Non Refugee (all other fclas codes) 2=Non immigrant (unlinked) 0 = sponsored refugee (REFERENCE) 0 = sponsored refugee (REFERENCE) 1 = non sponsored refugee

73 55 A2.2b. Outcomes Outcome Paper1: Severe maternal morbidity (SMM) indicator Paper 2: Severe neonatal morbidity (SNM) indicator Paper 3: Preterm birth Source Database Definition As defined in Joseph et al, DAD Variables: Diagnosis and procedure fields As defined in Lain et al, 2012, 30 DAD Variables: modified for Canadian context Diagnosis and procedure fields Gestational age (GA) at birth Very PTB: completed weeks Moderate PTB: completed weeks Full term: 37 completed weeks DAD Variable: B_GESTWKS_DEL Codes Numerous see Appendix of Chapter 3 Numerous see Appendix of Chapter 2 ICD 10: gestational age in completed weeks Operationalization for description and analyses 1 = presence of SMM 0 = absence of SMM (REFERENCE) 1 = presence of SNM 0 = absence of SNM (REFERENCE) 2 = weeks GA 1 = weeks GA 0 = 37 weeks GA (REFERENCE) A2.2c. Covariates (variables: hierarchical, descriptive, confounder stratify or adjusted, effect modifier, cohort effect, mediator explored) Source Database Operationalization for Variable type Covariate Definition Coding description and analyses Maternal encrypted health card number Unique encrypted health card number for each mother DAD Variable: M_IKN Unique codes 2 nd level in multilevel analyses (deliveries/births nested within mothers), sensitivity analysis all manuscripts Hierarchical Country of birth * Principal country of birth CIC Variable: FCOB CIC based coding Transition country* Country of last permanent residence, if country country of birth CIC Variables: FCOB, FCLPR CIC based coding 2 nd level in multilevel analyses (deliveries/births nested within maternal countries of birth), all manuscripts 2 nd level in multilevel analyses (deliveries/births nested within maternal countries of birth), manuscript three Hierarchical Hierarchical

74 56 World region of birth and transition* Duration of residence* Period of immigration Secondary Migration status* Level of Education* Based on country of birth, defined using United Nations regional classification system ethods/m49/m49regin.htm ((Date of birth or date of delivery) Date of landing)/ days Landing date categories based on relevant immigration policy changes: 1991 removal of HIV as grounds to reject application, 2002 Immigration and Refugee Protection Act enacted Woman whose country of last permanent residence (transition country) country of birth Eight levels of education are shown, based on the number of years of schooling or the certificate, diploma or degree obtained CIC Variable: FCOB CIC Varable: LANDING_DATE; DAD Variable: M_ADMDATE, B_BDATE CIC Variable: LANDING_DATE CIC Variables: FCOB, FCLPR CIC Variable: FEDUC CIC based coding Continuous in years Continuous in years, extracted from landing_date CIC based coding 1=Sub Saharan Africa 2=South Asia 3=Latin America & Caribbean 4=Mid East, N. Africa, Central Asia 5=Eastern Europe 6=Southeast Asia, Oceania Islands 7=Southern Europe 8=Industrialized Continuous 5 yr categories 1= = = =secondary migrant 0=non secondary migrant (REFERENCE) 1= 0 9 years 1=0 9 years 2=10 12 years 2=10 12 years 3=13 or more 3=13+years years 4=trade 4=trade certificate, nonuniversity diploma certificate 5=nonuniversity diploma 6=Bachelor s 5=bachelors degree, masters Confounder stratify, Effect modifier stratify Descriptive Descriptive, Cohort effect, Effect modifier stratify Effect modifier Confounderadjusted Confounderadjusted

75 Spoken language ability* Ability to communicate in one of Canada s official languages CIC Variable: CAN_LANG Maternal age Age at delivery or birth DAD Variable: M_AGE 7=Master s 8=Doctorate 1=English 2=French degree, doctorate (REFERENCE) 3=Both (above) 4=Neither 1=neither 3=Separated, divorced, widowed Continuous 57 0=knowledge of one or both official languages (REFERENCE) 3=Separated, divorced, widowed Categorical: 1=15 19 years 2=20 24 years 3=25 29 years (REFERENCE) 4=30 34 years 5=35 39 years 6=40+ years Confounderadjusted Confounderadjusted Parity Number of previous children DAD: PREVBIRTH Continuous 0=none Confounderadjusted 1=1 previous (REFERENCE) 2=2 previous 3= 3 previous Multi fetal pregnancy Delivery/record associated with multiple birth DAD Variable: B_MULTIBIRTH T=true F=false Descriptive, Identify exclusion Categorized into localized (i.e., hemorrhage) and systemic (i.e., hypertension) co morbidities DAD Variables: Diagnosis and procedure fields Numerous see supplementar y tables of manuscript 3 1 = presence of co morbidity 0 = absence of co morbidity (REFERENCE) Maternal comorbidities Mediatorexplored * In analyses that compare refugee and non refugee immigrants to non immigrants (manuscripts one and two), these covariates will not be included as they are not available for non immigrant women

76 58 Table A2.3a. Proportion of CIC females having at least one birth between 1988 and 2010 Notes Calculations All CIC database ( ) women ~1,459,367 Linked number of Females in CIC ( ) 1,275,057 Observed number of women giving birth at 329, 867 least once between 1988 and 2010 % of women immigrating between giving birth at least once between =329,687/1,275,057 =26% Table A2.3b. Estimated proportion of NON IMMIGRANT group that are unmatched CIC female immigrants that remained in Ontario and having at least one birth between Proportion of all CIC immigrants that now live =1,459,367 females * 0.08 in another province (CIC document states 8% of =116,749 immigrant women intended for Ontario now immigrants intending to live in Ontario between 1991 live in another province and 2006 lived in another province as of ) Unmatched female CIC records =184,310 females Estimated number of unmatched CIC females =184, ,749 remaining in Ontario =67,650 Estimated number of unmatched CIC females remaining in Ontario and giving birth between Number of non immigrants giving birth between % of NON IMMIGRANT group giving birth between that are actually unmatched CIC immigrants =67,650 immigrants * 0.26 (table a, above) =17,566 1,412,899 =17,566/1,412,899 =1.2%

77 Table A2.3c. Estimated proportion of NON IMMIGRANT group that are interprovincial immigrants having at least one birth between 1988 and 2010 Notes Calculations 1 Proportion of immigrants immigrating to Canada between NOT linked to tax file in 2006 correction factor for numbers of immigrants reported in this CIC report =3.7 million 2.1 million/3.7 million = 43% 1.43 Yearly interprovincial migration of female immigrants to Ontario between =39,313*1.43 =56,217 =56,217/9 years =6246 females per year Estimated number of interprovincial immigrants to Ontario between =6246*15 years =93,695 2 Yearly interprovincial migration of female immigrants to Ontario between =17,578*1.43 =25,136 =25,136/7 years =3,591 females per year Estimated number of interprovincial immigrants to Ontario between =3591*11 years =39,500 3 Estimated total interprovincial female immigrants between =93,695+39,500 =133,195 4 Estimated number of interprovincial immigrant females giving birth between =133,195 immigrants * 0.26 =34,631 % of NON IMMIGRANT group giving birth between that are interprovincial immigrants =34,631/1,412,899 =2.5% Table A2.3d. Estimated proportion of immigrant mothers (arriving before 1985, unlinked CIC immigrants and interprovincial immigrants) among NON IMMIGRANT mothers giving birth between 1988 and 2010 Notes Calculations Non immigrants giving birth between 1988 =1,412, % of NON IMMIGRANT group giving birth between that are unlinked CIC immigrants or interprovincial immigrants =(39,000+23,469+34,631)/ 1,412,899 =6% 59

78 60 Table A2.4 Assuming 10% of non immigrant group are truly immigrants. Potential misclassification of counts (N, deliveries), rate and rate ratios (RR) reported for refugee immigrant vs. non immigrant model reported in manuscript 1 for Severe Maternal Morbidity (SMM) Refugee Non Refugee All Immigrants Non Immigrants RR (refugees vs. non immigrants) Reported Values N (deliveries) 30, , , , SMM cases 519 2,856 3,375 10,878 SMM risk Potentially misclassified counts N (deliveries) 10,050 b 77,820 87,871 a 87,871 a N/A SMM cases 171 c 944 1,115 1,115 True N deliveries and SMM cases (reported + misclassified), rate ( true SMM cases/ true N deliveries), rate ratio ( true rate refugees/ true rate non immigrants) N (deliveries) 40, , , , SMM cases 690 3,800 4,490 9,763 SMM risk a 10% of non immigrant group potentially misclassified as immigrants b 11% of immigrant deliveries are to refugee immigrants c 15% of SMM cases to immigrants are to refugee immigrants Table A2.5 Assuming 10% of non immigrant group are truly immigrants. Potential misclassification of counts (N, births), risk and risk ratios (RR) reported for refugee immigrant vs. non immigrant model reported in manuscript 2 for Severe Neonatal Morbidity (SNM) Refugee Non Refugee All Immigrants Non Immigrants RR (refugees vs. non immigrants) Reported Values N (births) 29, , , , SNM cases 1,380 9,638 11,018 43,220 SNM risk Potentially misclassified counts N (births) 9,827 b 76,235 86,062 a 86,062 a SMM cases 456 c 3,182 3,638 3,638 True N deliveries and SNM cases (reported + misclassified), risk ( true SNM cases/ true N deliveries), risk ratio ( true risk refugees/ true risk non immigrants) N (births) 39, , , , SNM cases 1,836 12,820 14,656 39,582 SNM risk a 10% of non immigrant group potentially misclassified as immigrants b 11% of immigrants births are to refugee immigrants c 13% of SNM cases to immigrants are to refugee immigrants

79 CHAPTER 3 (MANUSCRIPT 1): SEVERE MATERNAL MORBIDITY AMONG REFUGEE IMMIGRANTS IN ONTARIO 61

80 62 ABSTRACT Background: Little is known about severe maternal morbidity and its determinants among refugee childbearing women. This study examines severe maternal morbidity among refugees, compared to non refugee immigrant and non immigrant women and assesses the impact of changes in immigration admission policies. Methods: Ontario hospital deliveries ( ) linked to immigration records were used to calculate rate ratios (RR) with 95% confidence intervals (95% CI) for a severe maternal morbidity (SMM) composite indicator comparing refugees to non refugee immigrants and non immigrants using logbinomial regression. Analyses were stratified by SMM subtypes and period of immigration (before and after enactment of the Immigration & Refugee Protection Act, 2002) to assess the impact of elimination of a policy which stated refugees were inadmissible if they placed excessive demand on Canadian health and social services. Results: SMM was significantly elevated among refugees (n=30 581) compared to non refugee immigrants (n= ) (adjusted RR: 1.22, 95% CI: ) and non immigrants (n= ) (ARR: 1.34, 95% CI: ). Of all SMM subtypes, the greatest disparity was for HIV: RR=7.94 (95% CI: ) vs. non refugee immigrants, and RR=17.37 (95% CI: ) vs. non immigrants. SMM RRs were highest among refugees who arrived after the policy change came into effect. When deliveries with HIV were excluded, the association between refugee status and SMM disappeared. Interpretation: Refugee women had higher rates of SMM than non refugee immigrants and nonimmigrants but these differences were explained by higher rates of HIV among refugees. Removal of inadmissibility criteria for refugees was associated with increasing disparities in SMM among refugees.

81 INTRODUCTION Globally, in 2012, approximately 15 million persons feared persecution or violence due to their race, religion, nationality or political views and were forced to flee their home countries 1 and in doing so became refugees of concern to the United Nations High Commissioner for Refugees. Given the reasons and nature of their flight, refugee women may experience different health risks compared to their immigrant counterparts, who instead migrate for economic opportunities or as family members. Accordingly, the maternal health profile of refugees may be poorer than that of immigrants due to exposure to pre (e.g., sex and gender based violence, poor quality health care) and post migration risk factors (e.g., low socio economic position and access to care). 2 Few studies have examined disparities in maternal health comparing refugees to non refugee immigrants, who have migrated to high income countries, or to host country populations. The differential health status of refugees and non refugee immigrants may also be shaped by immigration policies. In Canada, the Immigration and Refugee Protection Act (IRPA) of 2002 stated that new immigration applicants could be rejected if it was determined that they would exert excessive demand on Canadian health and social services. 3 However, IRPA also shifted refugee immigration to emphasize humanitarian protection over pre existing medical conditions and thus refugees were exempt from this inadmissibility criteria. 3 The influence of this policy on the maternal health of refugee women in Ontario will be examined in this study. Severe maternal morbidity (SMM) or near miss morbidity has become an increasingly important indicator to assess the maternal health of populations 4 6 and it has been used in studies with immigrant populations. 7 9 The objective of this study was to use a Canadian specific SMM surveillance indicator 10 to assess the maternal health of refugees in comparison to 1) non refugee immigrants and 2) nonimmigrants, in Ontario, Canada. We also examined whether differences varied according to: 1) SMM subtypes, and 2) periods of immigration (before and after 2002) coinciding with removal of an inadmissibility policy aimed at refugees with medical conditions which could place an excessive burden on the health care system. 3.2 METHODS Study design and inclusion/exclusion criteria This population based database study included all Ontario hospital admissions for childbirth that occurred between April 1, 2002 and March 31, Deliveries to refugee and non refugee immigrants

82 64 were identified retrospectively through linkage of hospital deliveries to the Citizenship & Immigration Canada (CIC) database ( ). Our definition of refugee immigrants was consistent with the United Nations definition: persons who cannot return to their home country due to a well founded fear of persecution based on race, religion, political opinion, nationality or membership in a particular social group. This definition was agreed to by Canada in 1969 when it acceded to the 1951 United Nations Convention Relating to the Status of Refugees and the 1967 Protocol Relating to the Status of Refugees. 11 Non refugee immigrants, on the other hand, had non refugee migration histories. Deliveries not linked to the CIC database were attributed to non immigrants of whom the vast majority were Canadian born. Refugees and non refugee immigrants who were <15 years old at the time of becoming permanent residents were excluded. Accordingly, non immigrants who were <15 at the time of delivery were also excluded. All women included in this study were eligible for provincial health care insurance. The unit of analysis was the delivery admission where multiple births were counted as a single delivery Data Sources We linked two administrative databases held at the Institute for Clinical and Evaluative Sciences (ICES) in Toronto, Ontario The Discharge Abstract Database (DAD). Records for women admitted to an Ontario hospital for childbirth were obtained from the DAD originating from the Canadian Institute of Health Information (CIHI). Diagnosis and procedure codes (using the 10the revision of the International Statistical Classification of Diseases and Related Health Problems, Canadian enhancement, and Canadian Classification of Health Interventions ICD 10 CA/CCI) identified women who did and did not have a severe maternal morbidity. 10 This dataset also contains information on maternal age at the time of delivery, self reported parity and birth plurality The Citizenship and Immigration Canada Database (CIC) is the official Canadian registry containing all legal immigrants immigration records. All immigrants arriving in Ontario between 1985 and 2010 were included. The CIC database contains information on refugee status (refugee and non refugee immigrant), the date of becoming a permanent resident, country of birth, as well as education level and knowledge of Canadian official languages upon arrival to Canada. The CIC linked database has been used in several epidemiological studies

83 Outcome of Interest The health outcome of interest, severe maternal morbidity (SMM), was measured by a composite indicator 5,10 developed by the Canadian Perinatal Health Surveillance System. The SMM indicator was operationalized as any woman having one or more of the ICD 10 CA/CCI diagnosis or procedure codes relating to specific diseases (i.e., HIV), interventions (i.e., blood transfusion) or organ dysfunctions (i.e., hepatic failure) recorded during a hospital admission for labour or delivery (see Table A3.1.1 for full list of codes). 10 This SMM indicator has been demonstrated to be a useful indicator to identify SMM temporal trends and disparities in Canada. 5 A study examining the validity of perinatal data available in the DAD suggested that both diagnoses and procedures are accurately coded and supported use of this database for perinatal research Exposure and Covariates Main Exposure Refugee Status The exposure group of interest were refugee immigrants who were compared to both nonimmigrants and non refugee immigrants Covariates Covariates included: Maternal age recorded at the time of delivery (categories: 15 19, 20 24, 25 29, 30 34, years), Parity (categories: 1, 2 and 3 previous births) and Birth plurality (categories: singleton, multiple birth). Some covariates were only available for refugees and non refugee immigrants since this information was collected by Citizenship and Immigration Canada. Maternal birth place included country of birth, world sub region and world region. The latter two covariates categorized according to the United Nations Geographical classification system. 18 Education at the time of arrival was described as 0 9 years, years, 13+ years (exclusive of other categories), trade certificate and non university diploma and bachelor/masters/doctorate. Knowledge of official languages at arrival was categorized as knowledge of one or both official Canadian languages (English, French) or knowledge of neither. Duration of residence in Canada defined as the time (years) elapsed between the date of becoming a permanent resident and the date of delivery. Period of immigration was categorized into three time periods: , ,

84 Analytic Methods The chance of any severe maternal morbidity (SMM) was reported as a rate (SMM cases per 1000 deliveries). To assess the disparities in SMM between refugees and a) non refugee immigrants and b) non immigrants, rates were further modeled using log binomial regression models to calculate unadjusted rate ratios (RR) and 95% confidence intervals (95% CI). 19 This process was repeated for each sub type of SMM. We also calculated adjusted RRs (ARR) and 95% CIs for any SMM, SMM excluding deliveries with significant SMM subtypes (other than HIV) and SMM excluding deliveries with HIV. To compare refugees to non immigrants (Model 1), RRs were adjusted for maternal age and parity (chosen a priori). Few confounders were available for adjustment in Model 1 since many covariates were unique to the CIC database. To compare refugees to non refugee immigrants (Model 2), we included a random effect for country of birth to take into account the potential similarity of SMM among immigrants from the same country of birth. From an intercept only model (no explanatory variables), the significance of the random intercept co efficient for each model was noted and if significant, the intra class correlation coefficient (ICC) was calculated 20 to quantify the correlation in SMM rates among mothers from the same country of birth. Confounders (identified a priori) adjusted for in model 2 included maternal age, parity, education level, language ability and duration of residence (as fixed effects). To examine the impact of changes in immigration policy on SMM, we decided a priori to stratify by those who became permanent residents before and after removal of the excessive demand policy for refugee admission (enacted by the Immigration & Refugee Protection Act [2002]). A posteriori, we also stratified by those who became permanent residents before and after 1991 since at this time having HIV/AIDS was removed as grounds for rejecting immigration applicants. 21 The Cochran Armitage test was used to examine linear trends in unadjusted SMM rates across periods of immigration for refugees and non refugee immigrants. In addition, models estimating ARRs were stratified according to periods of immigration. In secondary analysis, models were further stratified by region and country of maternal birth. 3.3 RESULTS The study sample consisted of , and eligible deliveries to refugee immigrant, non refugee immigrant and non immigrant women, respectively. The characteristics of these groups are summarized in Table 3.1. The proportion of deliveries to refugee mothers with parity 3 was

85 67 almost three times (17.2%) that of non refugee immigrants (5.9%) and non immigrants (5.8%). Only 9.1% of refugee immigrants completed a Bachelors degree or higher at the time of delivery compared to 31.7% among non refugee immigrants. In addition, the proportion of refugee immigrant women born in Sub Saharan Africa (29.5%) was 6 times higher than that of non refugee immigrants (5.3%). The rate of any SMM was significantly elevated among refugee immigrants compared to a) nonrefugee immigrants (RR=1.40, 95%: ) and b) non immigrants (RR=1.35, 95% CI: ) (Table 3.2). Of all SMM subtypes, HIV had the highest prevalence ratio: RR=7.94 (95% CI: ) vs. non refugee immigrants, and RR=17.37 (95% CI: ) vs. non immigrants. After adjustment (Table 3.3), the rate of any SMM remained significantly higher among refugee immigrants in comparison to non immigrants (ARR=1.34, 95% CI: ) and in comparison to nonrefugee immigrants (ARR=1.22, 95% CI: ). For model 2 (vs. non refugee immigrants), the ICC indicated that there was 5.66% correlation in the SMM rate between mothers from the same country of birth. When we excluded deliveries with HIV, point estimates were reduced towards the null and became non significant: ARR=1.07 (95% CI: ) (refugee immigrants vs. non immigrants) and ARR=1.11 (95% CI: ) (refuge immigrants vs. non refugee immigrants) (ICC=1.81%). See Tables A3.1.2 to A3.1.5 in this chapter s appendix for coefficients and standard errors for all variables included in these models. To investigate whether refugee immigrant women experienced additional health risks as a result of a higher prevalence of HIV disease, SMM sub types among those with HIV disease were examined. Of 115 and 112 deliveries with HIV disease to refugee immigrant and non refugee immigrant women respectively, 5 deliveries in each group exhibited one or more other SMM sub types. Among the 184 deliveries to non immigrants with HIV disease, eight had one or more other SMM sub types. As a proxy for severity of HIV disease, we also examined mode of delivery since caesarean delivery is recommended for more severe disease. 22 There was no significant difference in caesarean delivery between exposure groups (p=0.5375). The Cochran Armitage test revealed a significant positive trend in SMM rates with recency of the immigration period for refugees (p<0.01) but not for non refugee immigrants (p=0.25). Refugee immigrants who became permanent residents after IRPA came into effect (2002) had a 45% greater rate of any SMM compared to non refugee immigrants over the same period (ARR=1.45, 95% CI: ) (appendix section A3.1 Table A3.1.6). This elevated rate was greater than among those who became permanent residents between 1991 (after which immigration applicants could not be rejected solely because of HIV/AIDS) and 2001 (ARR=1.11, 95% CI: ) and between 1985 and 1990 (ARR=1.01,

86 68 95% CI: ). When deliveries with HIV were excluded, the associations were reduced towards the null and no longer significantly elevated in the post IRPA period. The rate of any SMM among African refugee immigrants was significantly higher compared to their non refugee immigrant counterparts (appendix section A3.1 Table A3.1.7). 3.4 INTERPRETATION This study found that the rates of SMM and its subtype, HIV, were significantly higher among refugee immigrant women compared to both non refugee immigrants and non immigrants. The rate of SMM among refugee immigrants relative to non refugee immigrants also varied by period of immigration with refugee immigrants who became permanent residents post IPRA (2002) having the greatest excess rate of SMM. However, after exclusion of deliveries with HIV, the rate of SMM among refugee immigrants was no longer significantly elevated, overall or in the post IRPA period, suggesting excess prevalence of HIV was likely the primary cause of elevated SMM rates. Additionally, HIV disease among refugee immigrants was no more a severe disease when compared with other groups. Three recently published studies have examined severe maternal morbidity among immigrants in Sweden 7 and the Netherlands 9 and among asylum seekers (women without permanent residency) in the Netherlands. 8 In Sweden, immigrants from low income countries had a 2.5 times greater risk for developing a near miss condition as compared to Swedish born women. 7 The Dutch study identified Sub Saharan African immigrants to be at a 3.5 times greater risk compared to western women. 9 Neither study reported an association with HIV. The Dutch study focusing on asylum seekers found a 4.5 times greater risk of severe acute maternal morbidity (SAMM) compared to the general Dutch population and a 3.6 times greater risk compared to Dutch non western immigrants. 8 This study also noted that among those with SAMM, asylum seekers had a 9 times greater prevalence of HIV compared to non western immigrants and a 20 times greater prevalence of HIV disease compared to Dutch born women. A small Canadian hospital based study examining obstetric outcomes (not SMM) also found a greater proportion of refugee claimants (without permanent residency) with HIV (3.6%) compared to control subjects (0.4%). 23

87 Strengths & Limitations Our study has limitations. Firstly, home births (<2% of births in Ontario 24 ) could not be captured. A possible preference for hospital births among refugees and non refugee immigrants may cause rate ratios comparing to non immigrants to be attenuated, however considering the size of the non immigrant group any attenuation was minimal. Secondly, the non immigrant group contains a small proportion of immigrants which may bias reported rate ratios slightly towards the null. Lastly, the SMM indicator includes both incident (i.e., placental abruption) and prevalent (i.e., HIV) SMM subtypes which makes it difficult to interpret the SMM ratio as either a risk ratio or a prevalence ratio. Under ascertainment of specific disease, organ dysfunction and intervention codes in administrative data is a potential concern for use of the SMM indicator. However the SMM indicator limits potential under ascertainment in three ways. Firstly, SMM is a severity indicator. It has been suggested that in addition to being more accurately reported, 17 severe conditions are less likely to suffer from under reporting. 25 Secondly, SMM is a composite indicator consisting of 40 diagnostic entities or procedures. A mother with severe illness is likely to have co morbidities and by using numerous diagnostic and procedure codes the chances of identifying all mothers with severe illness is increased. Lastly, the SMM indicator includes procedures which are known to be more accurately reported than diagnoses 17, with high positive predictive values and few false positives. 25 As such, including procedural codes reduces under ascertainment through accurate reporting. In terms of strengths, to our knowledge this is the first study to examine the rate of SMM in a large refugee immigrant population with diverse representation consistent with Canada s refugee and non refugee immigrants. We had sufficient statistical power to address the research questions. 3.5 CONCLUSION This study found that refugee women in Ontario had higher rates of SMM and HIV compared to other groups. Removal of refugee inadmissibility criteria associated with the Immigration & Refugee Protection Act (IRPA) (2002) resulted in greater disparities in SMM between refugees and non refugee immigrants arriving after 2002 than those arriving in previous years, implying that the excess in SMM among refugee women can likely be attributed to this policy. The increase in SMM among refugees with recency of immigration suggests that IRPA was effective in shifting refugee immigration towards humanitarian protection. Our study also found that refugee mothers with HIV were no more a burden on the health care system, in terms of excess severe co morbidities and interventions, than others with

88 70 HIV. Additional research is needed to examine less severe morbidities among refugee mothers in Canada. Reductions in health care coverage available to refugee claimants in Canada were announced in June Since HIV is considered a Public Health and Safety concern 27 treatment remains covered under the Interim Federal Health Program. 27 However loss of some health care benefits 26 and drug coverage 28 for conditions not deemed a public health and safety concern (i.e., hypertension, diabetes), may have important implications for the maternal health of refugee claimants and their Canadian children. ACKNOWLEDGEMENTS The authors thank Dr. Darrell Tan for providing guidance on the implications of study findings.

89 REFERENCES United Nations High Commissioner for Refugees. Displacement The new 21st century challenge, UNHCR Global Trends 2012 [Internet] Available from: 2. Gagnon AJ, Merry L, Robinson C. A systematic review of refugee women s reproductive health. Refuge. 2002; 3. Citizenship and Immigration Canada. Evaluation of government assisted refugees (GAR) and resettlement assistance program [Internet] Available from: rap/summary.asp 4. Tunçalp ö, Hindin M, Souza J, Chou D, Say L. The prevalence of maternal near miss: a systematic review: Prevalence of near miss: systematic review. BJOG Int J Obstet Gynaecol May;119(6): Liu S, Joseph KS, Bartholomew S, et al. Temporal trends and regional variations in severe maternal morbidity in Canada, 2003 to J Obstet Gynaecol Can. 2010;32(9): Zwart J, Richters J, Öry F, Vries J de, Bloemenkamp K, Roosmalen J van. Severe maternal morbidity during pregnancy, delivery and puerperium in the Netherlands: a nationwide population based study of pregnancies. BJOG Int J Obstet Gynaecol Jun;115(7): Wahlberg å, Rööst M, Haglund B, Högberg U, Essén B. Increased risk of severe maternal morbidity (near miss) among immigrant women in Sweden: a population register based study. BJOG Int J Obstet Gynaecol Dec;120(13): Hanegem N Van, Miltenburg AS, Zwart JJ, Bloemenkamp KWM, Roosmalen J Van. Severe acute maternal morbidity in asylum seekers: a two year nationwide cohort study in the Netherlands. Acta Obstet Gynecol Scand Sep;90(9): Zwart JJ, Jonkers MD, Richters A, et al. Ethnic disparity in severe acute maternal morbidity: a nationwide cohort study in the Netherlands. Eur J Public Health Apr 1;21(2):

90 Joseph KS, Liu S, Rouleau J, et al. Severe maternal morbidity in Canada, 2003 to 2007: surveillance using routine hospitalization data and ICD 10CA codes. J Obstet Gynaecol Can. 2010;32(9): Canadian Council for Refugees. Brief history of Canada s responses to refugees [Internet] [cited 2014 Oct 14]. Available from: history canadas responsesrefugees 12. Urquia ML, Frank JW, Moineddin R, Glazier RH. Does time since immigration modify neighborhood deprivation gradients in preterm birth? A multilevel analysis. J Urban Health. 2011;88(5): Urquia M, Frank J, Moineddin R, Glazier R. Immigrants duration of residence and adverse birth outcomes: a population based study. BJOG Int J Obstet Gynaecol Apr;117(5): Urquia ML, Frank JW, Glazier RH. From places to flows. International secondary migration and birth outcomes. Soc Sci Med. 2010;71(9): Creatore MI, Moineddin R, Booth G, et al. Age and sex related prevalence of diabetes mellitus among immigrants to Ontario, Canada. Can Med Assoc J. 2010;182(8): Lofters A, Moineddin R, Hwang S, Glazier R. Predictors of low cervical cancer screening among immigrant women in Ontario, Canada. BMC Womens Health. 2011;11(20). 17. Joseph KS, Fahey J. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute for Health Information. Chronic Dis Can. 2009;29(3): United Nations Statistics Division. Composition of macro geographical (continental) regions, geographical sub regions, and selected economic and other groups [Internet] [cited 2014 Apr 3]. Available from: Knol MJ, Cessie S Le, Algra A, Vandenbroucke JP, Groenwold RH. Overestimation of risk ratios by odds ratios in trials and cohort studies: alternatives to logistic regression. Can Med Assoc J. 2012;184(8): Snijders TA, Bosker R. Multilevel analysis: an introduction to basic and advanced multilevel modeling. 2nd ed. SAGE Publications Ltd; 2012.

91 21. Klein A. HIV/AIDS and immigration final report [Internet]. Canadian HIV AIDS Legal Network; Available from: Hacker NF, Gambone JC, Hobel CJ. Hacker & Moore s Essentials of Obstetrics and Gynecology. Saunders Elsevier; Kandasamy T, Berniak R, Shah R, Yudin MH, Spitzer R. Obstetric Risks and Outcomes of Refugee Women at a Single Centre in Toronto. J Obstet Gynaecol Can. 2014;36(4): College of Midwives of Ontario. The facts about home birth in Ontario [Internet]. Available from: OURCES_2012.pdf 25. Lain SJ, Hadfield RM, Raynes Greenow CH, et al. Quality of data in perinatal population health databases. Med Care. 2012;50(4):e7 e Barnes S. The real cost of cutting the Interim Federal Health Program [Internet]. Toronto, Ontario: Wellesley Institute; 2013 Oct. Available from: Health Impacts of IFHP.pdf 27. Citizenship & Immigration Canada. Information sheet for Interim Federal Health Program Beneficiaries [Internet] [cited 2014 Oct 21]. Available from: info sheet.asp 28. Canadian Pharmacists Association. Cuts to refugee benefits: CPhA leads the fight [Internet] [cited 2014 Oct 21]. Available from: events/cphablog/cuts to refugee benefits cpha leads the fight/

92 TABLES Table 3.1: Characteristics of deliveries ( ) to refugee immigrant, non refugee immigrant mothers and non immigrants (n, proportion of total column, unless otherwise indicated) Deliveries to Refugee Immigrants (N=30,581) Deliveries to Non Refugee Immigrants (N=236,565) Deliveries to Non Immigrants (N=886,975) Characteristic N % N % N % Severe maternal morbidity subtypes (count) None 30,059 (98.3) 233,683 (98.8) 875,786 (98.7) (1.1) 1,793 (0.8) 6,858 (0.8) (0.4) 747 (0.3) 3,316 (0.4) 3 40 (0.1) 180 (0.1) 614 (0.1) 4 or more 25 (0.1) 162 (0.1) 401 (0.1) Maternal age at delivery (years) (0.9) 1,085 (0.5) 40,993 (4.6) ,930 (9.6) 22,729 (9.6) 127,594 (14.4) ,862 (25.7) 69,132 (29.2) 245,239 (27.7) ,491 (34.3) 84,127 (35.6) 298,500 (33.7) ,034 (23.0) 48,382 (20.5) 146,354 (16.5) 40+ 1,981 (6.5) 11,110 (4.7) 28,385 (3.2) Parity (previous births) none 9,638 (31.6) 100,713 (42.6) 408,189 (46.1) 1 9,930 (32.5) 90,063 (38.1) 311,950 (35.2) 2 5,719 (18.7) 31,682 (13.4) 114,163 (12.9) 3 or more 5,258 (17.2) 13,870 (5.9) 51,414 (5.8) Plurality Singleton 30,275 (99) 234,199 (99) 869,235 (98) Higher order 461 (1) 3,305 (1) 16,217 (2) Language ability at arrival English 17,880 (58.5) 139,973 (59.2) French 777 (2.5) 2,493 (1.1) Not available Both English & French 639 (2.1) 6,933 (2.9) Neither 11,285 (36.9) 87,161 (36.9) Education level at arrival 0 9 years 8,979 (29.4) 32,293 (13.7) years 11,525 (37.7) 57,661 (24.4) 13+ years a 3,372 (11.0) 25,935 (10.9) Not available Trade certificate, non university diploma 3,909 (12.8) 32,907 (13.9) Bachelors, Masters, Doctorate 2,796 (9.1) 87,769 (37.1) Region of maternal birth Sub Saharan Africa 9,005 (29.5) 12,573 (5.3) South Asia 8,984 (29.4) 85,294 (36.1) Latin America & Caribbean 3,113 (10.2) 26,563 (11.2) Western & Central Asia 2,762 (9.0) 14,667 (6.2) East Asia 1,674 (5.5) 33,787 (14.3) Not available Eastern Europe 1,978 (6.5) 16,365 (6.9) Southern Europe 1,923 (6.3) 6,097 (2.6) 74

93 Southeast Asia, Oceania Islands 847 (2.8) 26,031 (11.0) North Africa 170 (0.6) 3,697 (1.6) Industrialized b 120 (0.4) 11,409 (4.9) Mean age at arrival (standard deviation) 25 (6) 26 (5) Not applicable Arrival year grouping ,736 (5.7) 7,613 (3.2) c 14,901 (48.7) 109,880 (46.5) Not applicable d 13,994 (45.6) 119,072 (50.3) Duration of residence at birth 4 years e 14,621 (47.8) 130,734 (55.3) 5 9 years 8,407 (27.5) 67,006 (28.3) years 5,506 (18.0) 28,961 (12.2) Not applicable years 1,889 (6.2) 8,935 (3.8) 20+ years 156 (0.5) 924 (0.4) a 13+ age category is exclusive of other categories, did not complete a subsequent qualification b Industrialized Countries: North America, Western and Northern Europe, Australia & New Zealand c in 1991 HIV/AIDS was delisted as a condition posing threat to Canadian public health and safety 21 d In 2002, testing for HIV/AIDS became a mandatory part of the Immigration Medical Exam (IME) ( 15 years or with known risk factor). The IME is used to assess inadmissibility based either on health conditions deemed a danger to public safety or that will put excessive demand on Canadian health and social services. Also, in this year a shift in immigration policy change made it so refugees could not be deemed inadmissible based on potential excessive demand of health and social services 3 e 4 category includes a small proportion of women who delivered prior to receiving permanent residency and prior to eligibility for the Ontario Health Insurance Plan (OHIP) 75

94 Table 3.2: Severe Maternal Morbidity (SMM) and select SMM subtypes, number of cases, rates (/1000 deliveries) and unadjusted rate ratios (RR) with 95% confidence intervals (95% CI) comparing refugees to a) non refugees b) non immigrants for deliveries from n cases, (rate, /1000 deliveries) Unadjusted RR (95% CI) Refugee Non Refugee Non Immigrants a) Refugees vs. b) Refugees vs. SMM or SMM Subtype a Immigrant Immigrants Non Refugee Non immigrants (n=30,581) (n=236,565) (n=886,975) Immigrants Any SMM 522 (17.1) 2,882 (12.2) 11,189 (12.6) 1.40 ( ) 1.35 ( ) HIV 115 (3.8) 112 (0.5) 192 (0.2) 7.94 ( ) ( ) Cerebral edema or coma b b b b 14 (0.0) b significant b significant Hepatic failure b b 11 (0.1) 25 (0.0) b b b significant Disseminated intravascular coagulation 6 (0.2) 26 (0.1) 42 (0.1) 1.79 ( ) 4.14 ( ) Rupture of uterus before onset of labor 33 (1.1) 173 (0.7) 841 (1.0) 1.48 ( ) 1.14 ( ) Embolization or ligation of pelvic vessels + post partum hemorrhage 30 (1.0) 165 (0.7) 452 (0.5) 1.41 ( ) 1.93 ( ) Placenta previa with hemorrhage + blood transfusion 20 (0.7) 111 (0.5) 366 (0.4) 1.39 ( ) 1.58 ( ) Cardiac arrest or failure, myocardial infarction and pulmonary edema 41 (1.3) 250 (1.1) 813 (0.9) 1.27 ( ) 1.46 ( ) Blood transfusion (whole blood: plate plasma and red blood cell) 227 (7.4) 1,441 (6.1) 5,694 (6.4) 1.22 ( ) 1.16 ( ) Assisted ventilation through tracheostomy or endotracheal tube 20 (0.7) 134 (0.6) 356 (0.4) 1.15 ( ) 1.63 ( ) Eclampsia 19 (0.6) 185 (0.8) 924 (1.0) 0.79 ( ) 0.60 ( ) Bolded values indicate statistical significance a SMM subtypes listed are those with statistically significant RRs comparing refugee immigrants to non refugee immigrants or to non immigrants b suppressed due to case counts 5, noted if statistically significant 76

95 Table 3.3: Any Severe Maternal Morbidity (SMM), SMM excluding deliveries with HIV disease and SMM excluding deliveries with significant subtypes (other than HIV disease) number of cases, rate (/1000 deliveries) and adjusted rate ratio (95% CI) comparing refugee immigrants to non immigrants (model 1) and refugee immigrants to non refugee immigrants (model 2) for deliveries from Model 1 Model 2 Refugee Immigrant Non Refugee Immigrants Non Immigrants Refugee Immigrant Non Refugee Immigrants Any SMM Deliveries (N) 30, , ,709 30, ,540 n cases (rate/1000 deliveries) 519 (17.1) 2,856 (12.1) 10,878 (12.4) 519 (17.1) 2,875 (12.2) Rate Ratios (95% CI) 1.34 ( ) 0.97 ( ) 1.00 (ref) 1.22 ( ) 1.00 (ref) ICC (%) N/A 5.66 a SMM, excluding deliveries with significant SMM subtypes, other than HIV b,c Deliveries (N) 30, , ,386 30, ,956 n cases (rate/1000 deliveries) 234 (7.8) 957 (4.1) 3,555(4.1) 259 (8.6) 1,272 (5.4) Rate Ratios (95% CI) 1.85 ( ) 0.99 ( ) 1.00 (ref) 1.31 ( ) 1.00 (ref) ICC (%) N/A a SMM, excluding deliveries with HIV disease Deliveries (N) 30, , , , ,428 n cases (rate/1000 deliveries) 404 (13.3) 2,744 (11.7) 10,694 (12.2) 404 (13.3) 2,744 (11.7) Rate Ratios (95% CI) 1.07 ( ) 0.95 ( ) 1.00 (ref) 1.11 ( ) 1.00 (ref) ICC (%) N/A 1.81 a Bolded values indicate statistical significance Model 1: Rate ratios adjusted for maternal age at delivery and parity Model 2: Rate ratios account for maternal country of birth (random effect), and adjust for maternal age at delivery, parity, education level, language ability and duration of residence (fixed effects) a random intercept co efficient for country of birth significant b See Table 3.2 For model 1, excludes SMM subtypes with significantly different rates (except HIV disease) among refugee immigrant mothers compared to non immigrants (cerebral edema or coma, hepatic failure, disseminated intravascular coagulation, emobolization or ligation of pelvic vessels and post partum hemorrhage, placenta previa with hemorrhage and blood transfusion, cardiac arrest or failure, myocardial infarction of pulmonary edema, blood transfusion whole blood: plate plasma, and red cell, assisted ventilation through tracheostomy or endotracheal tube, eclampsia) c See Table 3.2 For model 2, excludes SMM subtypes with significantly higher rates (except HIV disease) among refugee immigrant mothers compared to non refugee immigrant mothers (cerebral edema or coma, rupture of uterus before onset of labor, blood transfusion) ICC=intra class correlation coefficient; N/A=not applicable 77

96 SECTION A3.1: SUPPLEMENTARY FILES Table A3.1.1: Severe Maternal Morbidity (SMM) Indicator ICD 10 CA diagnosis codes and CCI codes Joseph et al, 2010 Variable ICD 10 CA Diagnosis Code Canadian Classification of Health Intervention Codes Acute abdomen K35, K37, K65, N73.3, N73.5 Acute renal failure 090.4, N17, N19, N99.0 Acute psychosis F53.1, F23 Adult respiratory syndrome J80 Assisted ventilation through 1.GZ.31.CA ND endotracheal tube Assisted ventilation through 1.GZ.31.CR ND tracheostomy Blood transfusion Special CIHI variable Cardiac arrest, cardiac failure, myocardial infarction, or pulmonary edema Cardiomyopathy in the puerperium Cerebral edema or coma G93.6, R40,2 Cerebral infarction Cerebral venous thrombosis in the peurperium Cerebral venous thrombosis in pregnancy Cerebrovascular diseases: I60 I64 subarachnoid and intracranial hemorrhage, stroke Death death, obstetric, after 42 days but <1 year after delivery 096 eclampsia death, obstetric, cause 095 unspecified 097 death from sequelae of direct R96 R99 obstetric causes sudden death, death from unspecified cause Dialysis Disseminated intravascular D65 coagulation Eclampsia 015 Evacuation of incisional hematoma Hemorrhage antepartum hemorrhage with , 074.2, 075.4, I21 I22, I46, I50, J , I42, I PZ.21^^ 5.PC.73.JS

97 Variable ICD 10 CA Diagnosis Code Canadian Classification of Health Intervention Codes coagulation defect intrapartum hemorrhage with coagulation defect embolization or ligation of pelvic vessels or suturing of uterus + post partum hemorrhage placenta previa with hemorrhage + blood transfusion intrapartum hemorrhage + blood transfusion postpartum hemorrhage + blood transfusion postpartum hemorrhage + hysterectomy Hepatic failure HIV disease Hysterectomy caesarean hysterectomy blood transfusion blood transfusion blood transfusion hysterectomy K71 K72 B20 B24, Z21 total hysterectomy (open approach) subtotal hysterectomy Obstetric embolism 088 Obstetric shock 075.1, R57, T80.5, T88.6 Placental abruption with coagulation defect Pre existing hypertensive heart disease Pre existing hypertensive heart and renal disease Puerperal sepsis 085 Pulmonary, cardiac, and central nervous system complications of anaesthesia during pregnancy, labour and delivery, and the peurperium 029.0, 029.1, 029.2, 089.0, 089.1, 089.2, 074.0, 074.1, 074.3, Septicemia during labour Repair of bladder, urethra or intestine Rupture of uterus before onset of labour Rupture of uterus during labour Sickle cell anemia with crisis D57.0 Status asthmaticus J45.01, J45.11, J45.81,J45.91 Status epilepticus G41 + (1.RM.13^^ or 1.KT.51 OR 5.PC.91.LA) 5.MD.60.RC, 5.MD.60.RD, 5.MD.60.KE, 5.MD.60.CB 1.RM.89.LA 1.RM.87.LA GX 5.PC.80.JR, 1.NK.80^^, 1.NM.80^^ 79

98 80 Table A3.1.2: SMM Model 1 (refugee immigrants vs. non immigrants) coefficients (n=1,144,674) Coefficient Estimate (SE) Rate Ratio (95% CI) Main Effects Refugees vs. Non Immigrants (0.045) 1.34 ( ) Non Refugees vs. Non Immigrants (0.021) 0.97 ( ) Confounders Maternal age (years) (0.045) 1.10 ( ) (0.029) 0.97 ( ) (ref) Reference (0.023) 1.09 ( ) (0.025) 1.34 ( ) (0.040) 1.72 ( ) Parity (0.020) 1.46 ( ) 1 (ref) Reference (0.029) 1.01 ( ) (0.034) 1.40 ( )

99 81 Table A3.1.3: SMM without HIV Model 1 (refugee immigrants vs. non immigrants) coefficients (n=1,144,263) Coefficient Estimate (SE) Rate Ratio (95% CI) Main Effects Refugees vs. Non Immigrants (0.051) 1.07 ( ) Non Refugees vs. Non Immigrants (0.021) 0.95 ( ) Confounders Maternal age (years) (0.046) 1.10 ( ) (0.030) 0.97 ( ) (ref) Reference (0.022) 1.09 ( ) (0.026) 1.34 ( ) (0.040) 1.72 ( ) Parity (0.020) 1.46 ( ) 1 (ref) Reference (0.030) 1.01 ( ) (0.035) 1.40 ( )

100 82 Table A3.1.4: SMM Model 2 (refugee immigrants vs. non refugee immigrants) coefficients (n=265,960) Coefficient Estimate (SE) Rate Ratio (95% CI) Main Effects Refugees vs. Non Refugees (0.056) 1.22 ( ) Confounders Maternal age (years) (0.253) 0.76 ( ) (0.070) 0.89 ( ) (ref) Reference (0.047) 1.22 ( ) (0.054) 1.57 ( ) (0.076) 2.06 ( ) Parity (0.041) 1.49 ( ) 1 (ref) Reference (0.059) 0.93 ( ) (0.070) 1.10 ( ) Education 0 9 years (0.064) 1.25 ( ) years (0.054) 1.13 ( ) 13+ years (0.063) 1.10 ( ) trade certificate, non university diploma (0.057) 1.04 ( ) Bachelors, Masters, Doctorate Reference 1.00 Language Ability neither English or French (0.044) 0.98 ( ) one or both of English and French Reference 1.00 Duration of residence (years) (0.005) 0.99 ( )

101 83 Table A3.1.5: SMM without HIV Model 2 (refugee immigrants vs. non refugee immigrants) coefficients (n=265,733) Coefficient Estimate (SE) Rate Ratio (95% CI) Main Effects Refugees vs. Non Refugees (0.060) 1.11 ( ) Confounders Maternal age (years) (0.280) 0.70 ( ) (0.072) 0.91 ( ) (ref) Reference (0.049) 1.20 ( ) (0.056) 1.58 ( ) (0.079) 2.09 ( ) Parity (0.042) 1.54 ( ) 1 (ref) Reference (0.061) 0.95 ( ) (0.074) 1.15 ( ) Education 0 9 years (0.065) 1.20 ( ) years (0.055) 1.08 ( ) 13+ years (0.065) 1.07 ( ) trade certificate, non university diploma (0.059) 0.96 ( ) Bachelors, Masters, Doctorate Reference 1.00 Language Ability neither English or French (0.044) 0.98 ( ) one or both of English and French Reference 1.00 Duration of residence (years) (0.005) 0.99 ( )

102 84 Table A3.1.6: Stratification by arrival cohort any SMM and SMM excluding mothers with HIV, number of cases, rate (/1000 deliveries) and adjusted rate ratios (95% CI) comparing refugee immigrants to non immigrants for deliveries from SMM SMM excluding deliveries with HIV disease Refugee Non Refugee Non Refugee Non Refugee Non Immigrant Immigrants Immigrants Immigrant Immigrants Immigrants Total Deliveries (N) 30, , 545 a 87,709 30, ,433 c 878, Deliveries (N) 1,728 7, ,709 1,727 7, ,525 n cases (rate/1000 deliveries) 21 (12.2) 97 (12.8) 10,878 (12.4) 20 (11.6) 95 (12.7) 10,694 (12.2) Model 1 Rate Ratios (95% CI) 0.86 ( ) 0.93 ( ) 1.00 (ref) 0.83 ( ) 0.93 ( ) 1.00 (ref) Model 2 Rate Ratios (95% CI) 1.01 ( ) 1.00 (ref) 0.95 ( ) 1.00 (ref) 1991 e 2001 Deliveries (N) 14, ,405 b 878,709 14, ,328 d 878,525 n cases (rate/1000 deliveries) 235 (15.9) 1,350 (12.3) 10,878 (12.4) 210 (14.2) 1,273 (11.6) 10,694 (12.2) Model 1 Rate Ratios (95% CI) 1.23 ( ) 0.99 ( ) 1.00 (ref) 1.12 ( ) 0.95 ( ) 1.00 (ref) Model 2 Rate Ratios (95% CI) 1.11 ( ) 1.00 (ref) 1.14 ( ) 1.00 (ref) 2002 f 2010 Deliveries (N) 13, , ,709 13, , ,525 n cases (rate/1000 deliveries) 263 (19.0) 1,409 (11.9) 10,878 (12.4) 174 (12.6) 1,375 (11.6) 10,694 (12.2) Model 1 Rate Ratios (95% CI) 1.54 ( ) 0.96 ( ) 1.00 (ref) 1.04 ( ) 0.95 ( ) 1.00 (ref) Model 2 Rate Ratios (95% CI) 1.45 ( ) 1.00 (ref) 1.12 ( ) 1.00 (ref) a Model 2 N=235,540; b Model 2 N=109,400; c Model 2 N=235,428; d Model 2 N=109,323 e In 1991 HIV/AIDS was delisted as a condition posing a threat to Canadian public health and safety 19 f In 2002, testing for HIV/AIDS became a mandatory part of the Immigration Medical Exam (IME) ( 15 years or with known risk factor). The IME is used to assess inadmissibility based either on health conditions deemed a danger to public safety or that will put excessive demand on Canadian health and social services. Also, in this year a shift in immigration policy change made it so refugees could not be deemed inadmissible based on potential excessive demand of health and social services. 3

103 85 Table A3.1.7: Stratification by world region, sub regions and major refugee contributing countries of birth SMM and SMM excluding deliveries with HIV disease, rate (/1000 deliveries) and rate ratios (95% CI) comparing refugee vs. non refugee immigrant for deliveries from SMM (n cases=3,404) SMM excluding deliveries with HIV disease (n=3,177) World region, sub regions and selected maternal countries of birth (n) Total deliveries (N) Refugee (rate /1000) Non Refugee (rate Unadjusted a Rate Ratio (95% CI) Adjusted b Rate Ratio (95% CI) Refugee (rate /1000) Non Refugee (rate Unadjusted a Rate Ratio (95% CI) Adjusted b Rate Ratio (95% CI) /1000) /1000) Africa 25, ( ) 1.35 ( ) ( ) 1.04 ( ) Sub Saharan Africa 21, ( ) 1.23 ( ) ( ) 0.98 ( ) Somalia 4, ( ) 0.89 ( ) ( ) 1.01 ( ) Ethiopia 3, ( ) 0.74 ( ) ( ) 1.01 ( ) Asia 173, ( ) 1.03 ( ) ( ) 1.03 ( ) Middle East/ N. Africa/Central Asia 21, ( ) 1.29 ( ) ( ) 1.24 ( ) Iraq 4, ( ) 0.97 ( ) ( ) 0.97 ( ) South Asia 93, ( ) 1.13 ( ) ( ) 1.13 ( ) Sri Lanka 15, ( ) 0.87 ( ) ( ) 0.86 ( ) Afghanistan 5, ( ) 0.97 ( ) ( ) 0.97 ( ) Pakistan 23, ( ) 1.59 ( ) ( ) 1.59 ( ) East Asia 35, ( ) 0.84 ( ) ( ) 0.85 ( ) China 28, ( ) 0.90 ( ) ( ) 0.90 ( ) S.E. Asia/Oceania Is. 26, ( ) 0.70 ( ) ( ) 0.68 ( ) Latin America & Caribbean 29, ( ) 1.16 ( ) ( ) 1.15 ( ) Colombia 2, ( ) 1.89 ( ) ( ) 1.89 ( ) Europe 32, ( ) 1.12 ( ) ( ) 1.04 ( ) Southern Europe 7, ( ) 1.11 ( ) ( ) 1.11 ( ) Former Yugoslavia 2, ( ) 1.02 ( ) ( ) 1.02 ( ) a Unadjusted: did not account for country of birth as a random effect due to non convergence b Adjusted: did not account for country of birth as a random effect due to non convergence, adjusted for maternal age at delivery, parity, education at arrival, Canadian language ability and duration of residence.

104 86 Table A3.1.8: Coefficient estimates (SE) comparing risk of SMM among refugee immigrants to nonrefugee immigrants, fully adjusted models with specified interaction No Interaction Duration No Interaction ( ) Education ( ) a p value on interaction term statistically significant (p<0.05) Full model: fcob (random effect), refugee status, maternal age, parity, education category, language ability and duration of residence

105 87 SECTION A3.2: SENSITIVITY ANALYSES Table A3.2.1: Comparing rates of SMM among refugee immigrants (vs. non immigrants and vs. non refugee immigrants) stratified by age at arrival categories (<15, 15 years old and BOTH age groups) SMM % (n) Adjusted Main Effect 15 yrs at arrival Adjusted Main Effect including BOTH <15 years at arrival 15 years at arrival BOTH Coefficient (SE) RR (95% CI) Coefficient (SE) RR (95% CI) Refugees vs (73) a 1.71 (519) 1.58 (592) (0.045) 1.34 ( ) (0.042) 1.26 ( ) Non Immigrants Ref 1.00 Ref 1.00 Refugees vs (73) a 1.71 (519) 1.58 (592) (0.056) 1.22 ( ) (0.0524) 1.18 ( ) Non Refugees 1.14 (325) 1.21 (2856) 1.20 (3181) Ref 1.00 Ref 1.00 Adjusted RR for vs. non immigrants includes maternal age and parity Adjusted RR for vs. non refugees includes random effect for country of birth and maternal age, parity, education, language and duration of residence (fixed effects) a Chi square p value <0.05 for difference in SMM proportion between women <15 years old at arrival and only women 15 at arrival (tests rows)

106 88 Table A3.2.2: Comparing rates of SMM for refugee immigrants compared to non refugee immigrants, with and without Ontario Health Insurance Plan (OHIP) eligibility at the time of delivery Refugee Non Refugee SMM RR Total in Paper (n) 30, ,565 SMM % (n) 1.71 (522) 1.22 (2,882) 1.40 SMM no HIV % (n) 1.33 (404) 1.17 (2744) 1.14 Without OHIP eligibility at birth Total (n, % of total in paper) 1,887 (6.2) 4,371 (1.8) N/A Government Assisted (n) 9 N/A N/A Landed/Claimants (n) 1,852 N/A N/A Dependents (n) 26 N/A N/A SMM with OHIP at birth ONLY % (n) 1.67 (478) 1.22 (2,831) 1.37 SMM without OHIP at birth ONLY % (n) 2.33 (44) a 1.17 (51) 1.99 SMM no HIV with OHIP at birth ONLY % (n) 1.32 (378) 1.17 (2,722) 1.13 SMM no HIV without OHIP at birth ONLY % (n) 1.55 (29) 1.10 (48) 1.41 a significant difference between those with and without OHIP at delivery or birth NOTE: 15 refugee immigrants and 5 non refugee immigrants gave birth before OHIP eligibility Table A3.2.3: Comparing nesting structures: random effect for country of birth or for mother. Coefficient estimates and unadjusted rate ratios with 95% confidence intervals comparing rates of SMM among refugee immigrants to non refugee immigrants Unadjusted Main Effect Results Levels included Coefficient (SE) RR (95% CI) Model 2 None a (0.0473) 1.41 ( ) Refugees vs. Deliveries within countries of birth b (0.0486) 1.41 ( ) Non Refugees Deliveries within mothers c (0.0547) 1.45 ( ) a model estimated using PROC GENMOD (LOG BINOMIAL) b models estimated using PROC NLMIXED with log distribution (using starting parameters from None model) c models estimated using PROC NLMIXED with Poisson distribution (using starting parameters from None model) (this distribution used since prevalence is low and sample size is large and since log distribution does not converge)

107 CHAPTER 4 (MANUSCRIPT 2): SEVERE NEONATAL MORBIDITY AMONG BIRTHS TO REFUGEE IMMIGRANTS IN ONTARIO 89

108 90 ABSTRACT Background: Little is known about the severity of neonatal morbidity of infants born to refugee women in Canada and other industrialized refugee receiving countries. The objectives of this study were: 1) to assess severe neonatal morbidity of infants born to refugee immigrant women and; 2) to determine whether severe neonatal morbidity differs between infants born to non sponsored refugee immigrant women and sponsored refugee immigrant women. Methods: Immigration records ( ) linked to Ontario hospital discharge data ( ) were used to calculate risk ratios (RR) with 95% confidence intervals (95% CI) for any severe neonatal morbidity (SNM) and SNM subtypes comparing refugee immigrants to non refugee immigrants and non immigrants using log binomial regression. We also compared the risk of SNM of non sponsored refugees to that of sponsored refugees. Results: Risk of any SNM was significantly higher among newborns of refugee immigrant women (n=29,364) in comparison to non refugee immigrant women (n=230,085) (adjusted RR=1.10, 95% CI: ) but significantly lower in comparison to non immigrants (N=853,643) (adjusted RR=0.94, 95% CI: ). Risk of SNM was not significantly elevated among non sponsored refugees compared to sponsored refugees. Conclusion: Results comparing SNM among infants born to refugee immigrant women to that of nonimmigrant women and non refugee immigrant women suggest that the healthy migrant effect applies to both immigrant groups but it is stronger for non refugee immigrants. Assumed difficulty in accessing formal support among non sponsored refugee immigrants compared to sponsored refugee immigrants did not appear to impact the incidence of severe neonatal morbidity among refugees.

109 INTRODUCTION In 2013, the United Nations High Commissioner for Refugees (UNHCR) recorded the highest level of human displacement since statistics have been kept, estimating 16.7 million refugees in numerous conflict regions around the globe. 1 Refugees are those that have fled their country of origin because of persecution, conflict or generalized violence and are unable or afraid to return home. 2 This is in contrast to non refugee immigrants who migrate primarily for economic purposes or for family reunification. There are a number of reasons to suspect that infants born to refugee immigrant women may be at increased risk for severe neonatal health outcomes. Firstly, a recent systematic review estimated that 1 in 5 refugee or displaced women in complex humanitarian settings experienced sexual violence and that this estimate is likely underestimated considering the stigma associated with reporting sexual violence. 3 Violence has been shown to lead to high rates of trauma related disorders including posttraumatic stress disorder (PTSD). 4 Major depression, often co morbid with PTSD, has been linked to increased admission rates to a neonatal care unit. 5 In addition, a population based study conducted in British Colombia, Canada showed that prenatal exposure to anti depressants was associated with increased incidence of low birth weight and rates of respiratory distress. 6 Secondly, high rates of acute malnutrition are present in many protracted refugee situations around the world including those in Kenya, Ethiopia, Sudan and Sierra Leone. 7 Malnutrition reduces maternal immunity and is associated with increased risk for infection which is known to impact neonatal morbidity. 8 In addition, from a postmigration perspective, low educational attainment among refugee immigrants may lead to reduced health seeking behaviours, particularly for prenatal care, which may have important implications for neonatal morbidity and mortality. 8 In Canada, as well as in other refugee receiving industrialized countries, little is known about the severity of neonatal health outcomes experienced by infants born to refugee immigrant women. Another important factor that may determine neonatal outcomes among refugee immigrant women is the levels of settlement assistance provided, as set by the sponsorship status of the refugee. Refugee immigrants arrive to Canada under two broad sub classes: 1) sponsored and 2) non sponsored. Sponsored refugees comprise both Government assisted and privately sponsored refugee immigrants, who are selected for re settlement by the UNHCR. Their re settlement is supported by the Government of Canada or a private organization. 9 Upon arrival, they are eligible for provincial healthcare coverage 10 and resettlement assistance which includes support navigating their new communities, basic life skills training, one time start up funds and time limited income support, among other benefits. 11 Nonsponsored refugees comprise refugee claimants (or asylum seekers ) who request refugee status upon

110 92 arrival to Canada. They are eligible for short term (interim) federal health (IFH) coverage while their application is being processed however it has been noted that a limited number of clinicians accept IFH and this may pose a barrier to care. 12 In addition, claimants may be eligible for provincial social assistance to help pay for basic necessities (e.g., rent, food) although this assistance is not assured and can be refused. 13 With approval of their claim and attainment of permanent residency status, refugee claimants are transferred to a provincial health insurance plan [Ontario Health Insurance Plan or OHIP in Ontario]). 10 I hypothesize that infants born to non sponsored refugee immigrant women will have higher risk of severe neonatal health outcomes compared to sponsored refugee immigrant women since the former may experience more difficulty in accessing formal supports upon arrival. There have been a number of studies that have examined the perinatal health of newborns of refugee immigrant women These studies covered a range of perinatal outcomes including low birth weight, preterm birth, small for gestational age and perinatal mortality but many studies are of uncertain quality and indicate no obvious pattern of risk among them. Of those studies with higher quality, 16,17 the outcomes examined are not specific and generally unable to capture the degree and severity of complications experienced by neonates. 23,24 That being said, no gold standard exists for the measurement of severe neonatal morbidity and the measures that do exist have limitations. The use of neonatal intensive care unit (NICU) admission is often used 25 but this measure has limited generalizability, since it captures only a select group of the sickest neonates requiring a very high level of care that is not available in all hospital settings. 24 Other validated measures depend on clinical and laboratory data that are not captured in population based hospitalization records and therefore not readily estimated. 23,24 Individual severe neonatal morbidities captured in hospitalizations data are often estimated 26 (i.e., sepsis) but pose a methodological challenge for research since outcomes are often too rare to detect important differences between groups. To overcome many of the measurement issues associated with severe neonatal morbidity described above, a neonatal morbidity indicator that enumerates multiple severe neonatal outcomes simultaneously based on hospitalizations data is both practical and efficient and still meaningful. In 2012, Lain et al, published findings based on development of such an indicator, referred to as the Neonatal Adverse Outcomes Indicator (NAOI). They found that their composite indicator had reasonable predictive validity with 10 times greater risk of infant mortality and hospital readmission among those with NAOI than without. This indicator was adapted for use with Canadian hospitalizations data for this study. The objectives of this research were first to assess differences in severe neonatal morbidity among infants born to refugee immigrant women in comparison to infants born to 1) non immigrant women

111 93 (majority Canadian born women) and to 2) non refugee immigrant women. Our second objective was to compare severe neonatal morbidity of infants born to non sponsored refugee immigrant women to that of sponsored refugee immigrant women. 4.2 METHODS Study design and inclusion/exclusion criteria This population based database study included all Ontario hospital based singleton live births that occurred between April 1, 2002 and March 31, Births to refugee women and non refugee women were identified retrospectively through linkage of hospital deliveries to the Citizenship & Immigration Canada (CIC) database ( ). Our definition of refugee immigrants was consistent with the United Nations definition. 2 Non refugee immigrants, on the other hand, had non refugee migration histories. Those births not linked to the CIC database were attributed to non immigrant women. The non immigrant group consists of a majority of Canadian born women (see Chapter 2 for detailed discussion of the non immigrant group). All women included in this study were eligible for provincial health care insurance. Implicit in this study is that exposures prior to reaching reproductive age likely play an important role in any differences that may exist. To ensure that these exposures occurred prior to immigrating to Canada for refugee and non refugee immigrants in this study, those who were <15 years old at the time of arrival were excluded. Accordingly, non immigrants who were <15 at the time of birth were also excluded. The unit of analysis was the singleton live birth, excluding births with a gestational age < 22 weeks or a birth weight < 500 grams Data Sources We linked two administrative databases held at the Institute for Clinical and Evaluative Sciences (ICES) in Toronto, Ontario. See below for a detailed description of each database The Discharge Abstract Database (DAD). Records for any singleton live birth occurring in an Ontario hospital were obtained from the DAD originating from the Canadian Institute of Health Information (CIHI). Diagnosis and procedure codes (defined by, the 10 th revision of the International Statistical Classification of Diseases and Related Health Problems, Canadian enhancement, and Canadian Classification of Health Interventions ICD 10 CA/CCI) were used to identify newborns who did and did not have a severe neonatal morbidity (adapted from Lain et al, please see below

112 for further details). This dataset also contains information on maternal age at the time of birth and selfreported parity The Citizenship and Immigration Canada Database (CIC) is the official Canadian immigration registry, composed of all legal immigrants arrival records. All immigrant women who arrived in Ontario between 1985 and 2010 were included. The CIC database contains information on refugee status (refugee immigrant and non refugee immigrant), the date of becoming a permanent resident, country of birth, as well as education level and knowledge of Canadian official languages upon arrival to Canada. The CIC linked database has been used in several epidemiological studies Outcome of Interest Severe neonatal morbidity was measured using a composite indicator based on birth admissions data referred to as the neonatal adverse outcomes indicator (NAOI) which was developed by perinatal researchers in Australia. 24 The diagnostic and procedure codes included generally related to the different systems of the body (i.e., respiratory, central nervous, gastrointestinal, circulatory and others). The researchers found that infants identified by the NAOI were ten times more likely to die and twice as likely to be readmitted to hospital in the first year of life compared to infants not identified by the NAOI. The authors concluded that the NAOI would be a cost effective way to monitor temporal trends, assess obstetric and neonatal interventions and the quality of perinatal care. To our knowledge this indicator has not yet been used in other published peer reviewed studies. The lead author (SW), in collaboration with an expert clinician and a general surgeon, mapped the Australian diagnosis and procedure codes over to the ICD 10 CA/CCI system (see appendix section A4.1 Table A4.1.1 for a list of codes). The major modifications to the NAOI for use in Canada were to exclude neonatal mortality and adult codes for sepsis and pneumonia (those not beginning with P ) and to exclude multiple births from analyses. Further discussion on these issues can be found in Chapter 2 (section ). The NAOI adapted for use with Canadian hospitalizations data will be hence forth be referred to as the severe neonatal morbidity (SNM) indicator. In terms of validity of the SNM indicator, a discussion of the concurrent validity of NAOI compared to SNM and their subtypes can be found in Chapter 2 (section ) and the demonstration of this in the appendix to this chapter (section A4.1 Table A4.1.2). The side by side comparison shown in Table A4.1.2 suggests reasonable similarity in the frequencies of the NAOI and SNM overall and their subtypes. The overall SNM proportion (5.4%) was within 15% of the overall NAOI proportion (4.6%). There appeared to be a larger discrepancy between the NAOI and SNM in terms of interventions (i.e.,

113 95 resuscitation with intubation) as well as sepsis. The discrepancy in interventions is likely due to the use of different coding systems for interventions between Australia and Canada. The difference in sepsis may represent a true difference since diagnosis codes are identical in the Australian and Canadian ICD 10 systems. Alternatively, there may be a propensity in Canada to treat urgent circumstances in the neonate before an infection is lab confirmed. The frequencies in this table were reviewed by a neonatologist who did not suggest that any were out of the ordinary. Adequate predictive validity of the SNM was suggested by investigations examining the association between SNM (including multiples to facilitate comparison to NAOI, but neonatal mortality still excluded) and mortality which revealed that infants with SNM had a 98 and 41 times greater risk of neonatal and infant death, respectively, compared to infants without any SNM (see appendix section A4.1, Table A4.1.3). Construct validity could not assessed since no gold standard exists for the measurement of severe neonatal morbidity. The other measures that would be relevant for severe neonatal morbidity are incorporated into the SNM measure as SNM subtypes (i.e., very preterm, birthweight <1500, sepsis) therefore examining agreement with these alternative measures of severity would be misleading. A study examining the validity of perinatal data available in the DAD suggested that both diagnoses and procedures are accurately coded and supported use of this database for perinatal research. 32 The dichotomous SNM indicator was operationalized as any singleton live birth with one or more of the ICD 10 CA/CCI diagnosis or procedure codes for SNM recorded at the birth admission Exposure and Covariates Main Exposure Refugee Status For the primary objective, the exposure group of interest were refugee immigrants who were compared to non refugee immigrants and non immigrants. For the second objective non sponsored refugees (refugee claimants) were compared to sponsored refugees (government assisted and privately sponsored refugees) Confounders Covariates included: Maternal age recorded at the time of delivery (categories: 15 19, 20 24, 25 29, 30 34, years) and Parity (categories: 1, 2 and 3 previous births). Some covariates were only available for refugees and non refugee immigrants since this information was collected by Citizenship and Immigration Canada. Maternal birth place included

114 96 country of birth, world sub region and world region. The latter two covariates categorized according to the United Nations Geographical classification system. 33 Education at the time of arrival was described as 0 9 years, years, 13+ years (exclusive of other categories), trade certificate and non university diploma and bachelor/masters/doctorate. Knowledge of official languages at arrival was categorized as knowledge of one or both official Canadian languages (English, French) or knowledge of neither. Duration of residence in Canada defined as the time (years) elapsed between the date of becoming a permanent resident and the date of delivery Analytic Methods Cumulative incidence of SNM was reported as a proportion of all singleton live births. Logbinomial regression models were used to calculate unadjusted risk ratios (RR) and 95% confidence intervals (95% CI) to compare the incidence of SNM among refugee immigrants to a) non refugee immigrants and to b) non immigrants. The above process was repeated for each sub type of the SNM indicator. Logistic regression could also be used to approximate RRs since odds ratios (OR) can approximate RRs when the outcome of interest is not common (<10%), as is the case in this study. However, since incident data in this study are being modeled as risks rather than odds (as would be appropriate in a case control study where modeling incident data is not possible), it was deemed more appropriate to use log binomial regression to estimate risk ratios directly. 34 A possible disadvantage of using log binomial regression is model convergence problems. 34 As such, if non convergence resulted, the model was modified from log binomial to robust Poisson. In this case, the Poisson distribution can be used to approximate the probabilities from the binomial distribution since the prevalence of SNM is low and the sample size is large. 35 We also used log binomial regression models to calculate adjusted RRs (ARR) and 95% CIs for any SNM. Refugee immigrants were compared to non immigrants (Model 1) and to non refugee immigrants (Model 2) and non sponsored refugee immigrants were compared to sponsored refugee immigrants (Model 3). In Model 1, RRs were adjusted for maternal age and parity (identified a priori). Few confounders were available for adjustment in Model 1 since many socio demographic are only captured for refugee immigrants and non refugee immigrants (in the CIC database). In Models 2 and 3, a random effect for country of birth was included to take into account potential similarity of SNM among infants born to women from the same country of birth. From an intercept only model (no explanatory variables), the significance of the random intercept co efficient for each model was noted and if significant, the intra class correlation co efficient (ICC) was calculated 36 to quantify the

115 97 correlation in SNM risk between infants born to mothers from the same country of birth. Confounders were identified a priori and models were adjusted for maternal age, parity, education level, language ability and duration of residence (as fixed effects). In supplementary analyses, RRs comparing infants born to refugee immigrant women to infants born to non refugee immigrant women were stratified by maternal world region and world sub region of birth. 4.3 RESULTS The study sample consisted of , and eligible births to refugee immigrant, non refugee immigrant and non immigrant (majority Canadian born) women, respectively. The characteristics of each of these groups are summarized in Table 4.1. The crude incidence of SNM was 4.64%, 4.17% and 5.02% among newborns born to refugee immigrant, non refugee immigrant and non immigrant women respectively (Table 4.2). Risk of any SNM among births to refugee immigrants was significant elevated compared to births to non refugee immigrants (RR=1.12, 95% CI: ) and significantly lower when compared to non immigrant women (RR=0.92, 95% CI: ) (Table 4.2). Nine SNM subtypes were significantly higher among refugees compared to non refugees with many related to the respiratory system (ventilatory support, respiratory distress syndrome, primary atelectasis respiratory failure and bronchio pulmonary dysplasia) (Table 4.2). The SNM subtype with the greatest disparity was for hypoxic ischemic encephalopathy (RR=2.46, 95% CI: ). In comparison to non immigrants, there were five subtypes that were significantly elevated among births to refugee immigrant women and another three that were significantly lower. Multivariable models (Table 4.3) revealed that infants born to refugee immigrant women had significantly lower risk of SNM when compared to infants born to non immigrants (ARR=0.94, 95% CI: ) but significantly higher risk when compared to infants born to non refugee immigrants (ARR=1.10, 95% CI: ). Infants born to non sponsored refugee women did not have significantly different risk of SNM when compared to infants born to sponsored refugee women (ARR=1.06, 95% CI: ). The ICCs for Model 2 was statistically significant and indicated that there was only a 1% correlation in risk of SNM among infants born to mothers from the same country of birth. See Tables A4.1.4, A4.1.5 and A4.1.6 in this chapter s appendix for coefficients and standard errors for all variables included in these models.

116 98 Supplementary analysis examining results stratified by maternal region and sub region of birth did not reveal significant differences between refugee immigrants and non refugee immigrants, after adjustment for confounders (see appendix section A4.1 Table A4.1.7). 4.4 DISCUSSION The findings of this study are three fold. Firstly, this study found that infants born to refugee immigrant women had a small but significantly increased risk of any SNM compared to non refugee immigrant women, with no significant variation by maternal region of birth. In comparison to infants born to non immigrant women, infants born to refugee immigrant women had a small and significantly lower risk of SNM. These findings may suggest that a very mild healthy immigrant effect (the phenomenon whereby foreign born individuals are healthier than native born residents, 37 perhaps because a certain level of health is needed to undergo migration) applies to infants born to both refugee immigrant women and non refugee immigrant women. However any healthy immigrant effect present is stronger for non refugee immigrants than for refugee immigrants. This may be due to factors known to have a high prevalence among refugees (i.e., malnutrition) as outlined in the introduction. This finding may also be related to immigration policies such as those outlined in the Immigration & Refugee Protection Act (2002) which actively exclude non refugee immigrants that could place excessive demand on health and social services (thereby those accepted are healthy) while the same restrictions are relaxed for refugee immigrants (thereby health criteria is less likely to be used as the basis for rejection). 38 Secondly, risks of five SNM subtypes related to the respiratory system were significantly greater among infants born to refugee immigrants compared to those born to non refugee immigrants. This warrants further investigation since severe lung injury may have important developmental implications for the infant. Thirdly, if a difference in formal support exists between non sponsored and sponsored refugees it does not appear to influence severe neonatal morbidity, since no significant difference in SNM was found between these two groups. We were unable to corroborate these findings in other refugee immigrant populations since no peer reviewed studies could be identified that examined one or more of the SNM subtypes that made up the SNM indicator. However four studies were identified among immigrant, ethnic or foreign born populations. The subtypes examined in these studies included ventilatory support (or deviations of it i.e., intubation), necrotizing enterocolitis, birth trauma, bronchio pulmonary dysplasia and intraventricular hemorrhage. NICU admission was also examined. A study from Greece found that rates of emergency intubation and NICU admission among immigrant women originating primarily from

117 99 Albania and the former Soviet Union countries were significantly lower when compared to native women. 39 A study of infants born to Somali immigrant women in the United States found a significantly higher rate of assisted ventilation when compared to both US born black women and US born white women. There were no significant differences in the rate of admission to the NICU or birth trauma. 40 One Canadian and one Australian study focused on outcomes of infants born to immigrant women that were admitted to the NICU 25,41 and therefore have limited comparability to the current study. The Australian study found that infants born to immigrant women were at increased risk of mechanical ventilation but lower risk of necrotizing enterocolitis compared to women born in Australia. 41 The Canadian study found that infants born to women of African or South Asian ethnicity (not necessarily immigrants) had 7% higher rates of any major neonatal morbidity (comprising: bronchopulmonary dysplasia, intraventricular hemorrhage, necrotising enterocolitis, patent ductus arteriosus or retinopathy or permaturity) compared to infants born to Caucasian women although the statistical significance of these differences were not indicated. 25 The non refugee immigrant population in the current study is most comparable to the immigrant populations in the non NICU based studies. However, given the lack of diversity represented in the non NICU based studies (Albania, Soviet Union and Somalia), it is difficult to draw comparisons with the present study Strengths & Limitations Using an outcome based on specific diagnostic and procedure codes in administrative data does have potential limitations, the most important of which is under ascertainment. 24,42 Differential underascertainment for refugee and non refugee immigrants seems unlikely considering the characteristics of immigrants that commonly influence under reporting of health related information (i.e., language ability, education) are less likely to come into play in administrative datasets since the physician and facility are responsible for collecting and managing this clinical data. However, potential underreporting by specific facilities or physicians that may serve large refugee immigrant and non refugee immigrant populations could be a factor and may serve to attenuate RRs when the comparison group is non immigrants. Potential under ascertainment should not be problematic for comparisons between refugee immigrants and non refugee immigrants because potential under reporting should be equal for both groups. Construct validity of the SNM indicator could not be conducted since no gold standard exists for measuring severe neonatal morbidity. The overall risk of SNM in the singleton live birth population (4.8%) indicates that the outcome is quite common and further suggests that some of the SNM subtypes may not necessarily represent severe neonatal conditions.

118 100 A major strength of our work is the use of large population based databases which included a diverse immigrant population with substantial refugee immigrant and non refugee immigrant representation from major regions of the world (e.g., Sub Saharan Africa and South Asia). This permitted detection of effect sizes of relevance to public health. Refugee status, refugee sponsorship status and country of birth were objectively measured since these were based on official immigration records. The SNM indicator performed well in terms of predictive validity and reasonably well in terms of concurrent validity. To our knowledge this is the first population based study to examine severe neonatal morbidities among refugee immigrants in an industrialized country. 4.5 CONCLUSION Newborns of refugee immigrant women were found to have a slightly increased risk of any SNM compared to that of non refugee immigrant women but not when compared to non immigrant (majority Canadian born) women, implying that a mild healthy migrant effect with respect to SNM applies to refugee immigrants and non refugee immigrants but less so for refugee immigrants. The increased risk of five SNM subtypes related to the respiratory system among refugee immigrants compared to non refugee immigrants warrants additional investigation. There was no significant difference in the risk of SNM between infants born to non sponsored refugee immigrant women and sponsored refugee immigrant women, suggesting that if a difference in formal support exists between the two groups, it does not appear to impact severe neonatal morbidity as measured here. Sponsorship status may have other health effects that were not examined in this study.

119 101 REFERENCES 1. United Nations High Commissioner for Refugees. War s Human Cost UNHCR Global Trends 2013 [Internet] Jun. Available from: 2. United Nations University. Protracted refugee situations Political, human rights and security implications. New York, New York: United Nations University Press; Vu A, Adam A, Wirtz A, et al. The Prevalence of Sexual Violence among Female Refugees in Complex Humanitarian Emergencies: a Systematic Review and Meta analysis. PLoS Curr [Internet] [cited 2014 Oct 3]; Available from: 4. Kirmayer LJ, Narasiah L, Munoz M, et al. Common mental health problems in immigrants and refugees: general approach in primary care. Can Med Assoc J Sep 6;183(12):E959 E Alder J, Fink N, Bitzer J, Hösli I, Holzgreve W. Depression and anxiety during pregnancy: A risk factor for obstetric, fetal and neonatal outcome? A critical review of the literature. J Matern Fetal Neonatal Med Jan;20(3): Oberlander TF, Warburton W, Misri S, Aghajanian J, Hertzman C. Neonatal outcomes after prenatal exposure to selective serotonin reuptake inhibitor antidepressants and maternal depression using population based linked health data. Arch Gen Psychiatry. 2006;63(8): Corbett M, Oman A. Acute malnutrition in protracted refugee situations: a global strategy UNHCR/WFP [Internet] Jan. Available from: 8. Moss W, Darmstadt GL, Marsh DR, Black RE, Santosham M. Research priorities for the reduction of perinatal and neonatal morbidity and mortality in developing country communities. J Perinatol Aug 13;22(6): Becklumb P. Background paper Canada s immigration program [Internet]. Library of Parliament; 2008 Sep. Available from: e.pdf 10. Citizenship and Immigration Canada. Health care Refugees [Internet] Dec. Available from: healthcare.asp

120 Elgersma S, Dolin B. Assistance for government sponsored refugees, chosen abroad [Internet]. Library of Parliament; 2008 Sep. Report No.: PRB 04 39E. Available from: e.pdf 12. Gagnon A, Rousseau H, Welt M, et al. Appendix 20: Screening during pregnancy: evidence review for newly arriving immigrants and refugees. Can Med Assoc J Sep 6;183(12):E824 E Community Legal Education Ontario. Refugee rights in Ontario, information for front line workers. Social assistance for refugee claimants. [Internet] Apr. Available from: assistance refugees 14. Kuvacic I, Skrablin S, Hodzic D, Milkovic G. Possible influence of expatriation on perinatal outcome. Acta Obstet Gynecol Scand. 1996;75(4): Davis JM, Goldenring J, McChesney M, Medina A. Pregnancy outcomes of Indochinese refugees, Santa Clara County, California. Am J Public Health. 1982;72(7): Vangen S, Stoltenberg C, Johansen REB, Sundby J, Stray Pedersen B. Perinatal complications among ethnic Somalis in Norway. Acta Obstet Gynecol Scand. 2002;81(4): Essen B, Hanson BS, Ostergren PO, Lindquist PG, Gudmundsson S. Increased perinatal mortality among sub Saharan immigrants in a city population in Sweden. Acta Obstet Gynecol Scand. 2000;79(9): Flynn PM, Foster EM, Brost BC. Indicators of acculturation related to Somali refugee women s birth outcomes in Minnesota. J Immigr Minor Health. 2011;13(2): Malamitsi Puchner A, Tzala L, Minaretzis D, Michalas S, Aravantinos D. Preterm delivery and low birthweight among refugees in Greece. Paediatr Perinat Epidemiol. 1994;8(4): Gann P, Nghiem L, Warner S. Pregnancy characteristics and outcomes of Cambodian refugees. Am J Public Health. 1989;79(9): Lalchandani S, MacQuillan K, Sheil O. Obstetric profiles and pregnancy outcomes of immigrant women with refugee status. Ir Med J. 2001;94(3):79 80.

121 King PA, Duthie SJ, Li DFH, Ma HK. Obstetric outcome among Vietnamese refugees in Hong Kong: an age matched case controlled study. Int J Gynaecol Obstet Off Organ Int Fed Gynaecol Obstet. 1990;33: Strobino D, Johns Hopkins Bloomberg School of Public Health. Measures of neonatal morbidity [Internet] [cited 2014 Apr 5]. Available from: Lain SJ, Algert CS, Nassar N, Bowen JR, Roberts CL. Incidence of severe adverse neonatal outcomes: use of a composite indicator in a population cohort. Matern Child Health J Apr;16(3): Claydon JE, Mitton C, Sankaran K, Lee SK. Ethnic differences in risk factors for neonatal mortality and morbidity in the neonatal intensive care unit. J Perinatol. 2007;27(7): Public Health Agency of Canada. Canadian Perinatal Health Report 2008 [Internet] Available from: aspc.gc.ca/publicat/2008/cphr rspc/ 27. Urquia ML, Frank JW, Moineddin R, Glazier RH. Does time since immigration modify neighborhood deprivation gradients in preterm birth? A multilevel analysis. J Urban Health. 2011;88(5): Urquia M, Frank J, Moineddin R, Glazier R. Immigrants duration of residence and adverse birth outcomes: a population based study. BJOG Int J Obstet Gynaecol Apr;117(5): Urquia ML, Frank JW, Glazier RH. From places to flows. International secondary migration and birth outcomes. Soc Sci Med. 2010;71(9): Creatore MI, Moineddin R, Booth G, et al. Age and sex related prevalence of diabetes mellitus among immigrants to Ontario, Canada. Can Med Assoc J. 2010;182(8): Lofters A, Moineddin R, Hwang S, Glazier R. Predictors of low cervical cancer screening among immigrant women in Ontario, Canada. BMC Womens Health. 2011;11(20). 32. Joseph KS, Fahey J. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute for Health Information. Chronic Dis Can. 2009;29(3):

122 United Nations Statistics Division. Composition of macro geographical (continental) regions, geographical sub regions, and selected economic and other groups [Internet] [cited 2014 Apr 3]. Available from: Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol Aug 1;162(3): Petersen MR, Deddens JA. A comparison of two methods for estimating prevalence ratios. BMC Med Res Methodol. 2008;8(9). 36. Snijders TA, Bosker R. Multilevel analysis: an introduction to basic and advanced multilevel modeling. 2nd ed. SAGE Publications Ltd; Urquia ML, Gagnon AJ. Glossary: migration and health. J Epidemiol Community Health. 2011;65(5): Citizenship and Immigration Canada. Evaluation of government assisted refugees (GAR) and resettlement assistance program [Internet] Available from: rap/summary.asp 39. Margioula Siarkou C, Petousis S, Kalogiannidis I, et al. Immigrants present improved obstetric and neonatal outcomes compared to native women. A northern Greek population analysis. J Immigr Minor Health Apr;15(2): Johnson EB, Reed SD, Hitti J, Batra M. Increased risk of adverse pregnancy outcome among Somali immigrants in Washington state. Am J Obstet Gynecol Aug;193(2): Uppal P, Holland AJA, Bajuk B, et al. The association between maternal country of birth and neonatal intensive care unit outcomes. Early Hum Dev Aug;89(8): Ford JB, Roberts CL, Algert CS, Bowen JR, Bajuk B, Henderson Smart DJ. Using hospital discharge data for determining neonatal morbidity and mortality: a validation study. BMC Health Serv Res. 2007;7(1):188.

123 TABLES Table 4.1: Characteristics of deliveries ( ) to refugee immigrant, non refugee immigrant and non immigrant women (N) (proportion of total column, unless otherwise indicated) Refugee Immigrants (N=29,765) 105 Non Refugee Immigrants (N=230,914) Non Immigrants (N=860,617) Characteristic N % N % N % Count of SNM subtypes None 28,385 (95.4) 221,276 (95.8) 817,397 (95.0) 1 1,380 (4.6) 9,638 (4.2) 43,220 (5.0) Maternal age at birth (years) (0.9) 1,060 (0.5) 39,839 (4.6) ,855 (9.6) 22,272 (9.7) 124,115 (14.4) ,678 (25.8) 67,728 (29.3) 238,682 (27.7) ,233 (34.4) 82,151 (35.6) 289,766 (33.7) ,815 (22.9) 47,007 (20.4) 141,187 (16.4) 40+ 1,906 (6.4) (4.6) 27,028 (3.1) Parity (previous births) none 9,336 (31.4) 98,000 (42.5) 395,614 (46.0) 1 9,725 (32.7) 88,302 (38.3) 303,891 (35.3) 2 5,589 (18.8) 31,009 (13.4) 111,130 (12.9) 3 or more 5,105 (17.2) 13,536 (5.9) 49,679 (5.8) Language ability English 17,402 (58.5) 136,520 (59.1) French 759 (2.6) 2,414 (1.1) Not available Both English & French 608 (2.0) 6,775 (2.9) Neither 10,996 (37.0) 85,200 (36.9) Education level 0 9 years 8,750 (29.4) 31,493 (13.6) years 11,217 (37.7) 56,192 (24.3) 13+ years a 3,282 (11.0) 25,310 (11.0) Not available Trade certificate, non university diploma 3,796 (12.8) 32,127 (13.9) Bachelors, Masters, Doctorate 2,720 (9.1) 85,792 (37.1) Region of maternal birth Sub Saharan Africa 8,716 (29.3) 12,174 (5.3) South Asia 8,749 (29.4) 83,303 (36.1) Latin America & Caribbean 3,035 (10.2) 25,717 (11.2) Western & Central Asia 2,673 (9.0) 14,241 (6.2) Not available East Asia 1,651 (5.5) 33,240 (14.4) Eastern Europe 1,931 (6.5) 15,985 (6.9) Southern Europe 1,875 (6.3) 5,933 (2.6) Southeast Asia, Oceania Islands 838 (2.8) 25,541 (11.1) North Africa 161 (0.0) 3,572 (1.6) Industrialized b 118 (0.4) 11,054 (4.8) Duration of residence at birth (years) 4 c 14,239 (47.8) 127,906 (55.4) 5 9 8,188 (27.5) 65,341 (28.3) Not applicable ,339 (17.9) 28,144 (12.2)

124 ,850 (6.2) 8,635 (3.7) (0.5) 883 (0.4) a 13+ age category is exclusive of other categories, did not complete a subsequent qualification b Industrialized Countries: North America, Western and Northern Europe, Australia & New Zealand c 4 category includes a small proportion of women who delivered prior to receiving permanent residence and prior to eligibility for OHIP 106

125 Table 4.2: Severe Neonatal Morbidity (SNM) and SNM subtypes (in descending order of risk among refugee immigrants), number of cases, % of births with SNM and unadjusted risk ratios (95% CI) by refugee status % of all births (n cases) Unadjusted RR (95% CI) SNM or SNM Subtype Refugee Immigrant (n=29,765) Non Refugee Immigrants (n=230,914) Non Immigrants (n=860,617) 107 a) Refugees vs. Non Refugees b) Refugees vs. Non Immigrants Any SNM 4.64 (1,380) 4.17 (9,638) 5.02 (43,220) 1.12 ( ) 0.92 ( ) Ventilatory support a 2.60 (774) 2.34 (5,400) 3.05 (26,214) 1.12 ( ) 0.85 ( ) Any intravenous fluids a 1.63 (486) 1.34 (3,084) 1.50 (12,934) 1.23 ( ) 1.09 ( ) Sepsis/septicaemia 1.25 (372) 1.18 (2,724) 1.36 (11,726) 1.06 ( ) 0.92 ( ) Respiratory distress syndrome 1.03 (307) 0.85 (1,958) 1.13 (9,713) 1.22 ( ) 0.91 ( ) Gestational age < 32 weeks 0.94 (280) 0.82 (1,896) 0.82 (7,015) 1.15 ( ) 1.16 ( ) Birthweight <1500 grams 0.86 (256) 0.78 (1,801) 0.67 (5,769) 1.10 ( ) 1.29 ( ) Primary atelectasis respiratory failure 0.39 (116) 0.30 (683) 0.33 (2,881) 1.32 ( ) 1.17 ( ) Central venous or arterial catheter a 0.38 (112) 0.28 (655) 0.46 (3,931) 1.33 ( ) 0.82 ( ) Seizure 0.26 (77) 0.16 (376) 0.20 (1,729) 1.59 ( ) 1.28 ( ) Pneumonia 0.21 (62) 0.18 (407) 0.21 (1,794) 1.18 ( ) 1.00 ( ) Any body cavity surgical procedure a 0.19 (58) 0.21 (482) 0.25 (2,125) 0.93 ( ) 0.79 ( ) Birth Trauma 0.13 (40) 0.11 (246) 0.12 (1,025) 1.26 ( ) 1.13 ( ) Necrotising enterocolitis 0.13 (38) 0.10 (240) 0.11 (943) 1.23 ( ) 1.17 ( ) Resuscitation a 0.12 (36) 0.13 (291) 0.13 (1,117) 0.96 ( ) 0.93 ( ) Hypoxic ischemic encephalopathy 0.11 (33) 0.05 (104) 0.06 (506) 2.46 ( ) 1.89 ( ) Intraventricular hemorrhage (grades 2,3,4) 0.09 (28) 0.08 (179) 0.09 (738) 1.21 ( ) 1.10 ( ) Bronchio pulmonary dysplasia 0.09 (28) 0.05 (127) 0.12 (1,069) 1.71 ( ) 0.76 ( ) Pneumothorax requiring intercostal catheter 0.07 (21) 0.08 (178) 0.12 (1,068) 0.92 ( ) 0.57 ( ) Transfusion of blood or blood products a 0.03 (9) 0.02 (37) 0.01 (118) 1.89 ( ) 2.21 ( ) Cerebral infarction 0.02 (6) 0.01 (34) 0.02 (141) 1.37 ( ) 1.23 ( ) Periventricular leukomalacia 0.02 ( b ) 0.02 (38) 0.02 (180) 1.02 ( ) 0.80 ( ) Bolded values are statistically significant; a Procedures; b suppressed due to counts 5

126 Table 4.3: Any Severe Neonatal Morbidity (SNM) number of cases, proportion SNM (%) and risk ratios (95% CI) comparing refugee immigrants to non immigrants (model 1), refugee immigrants to non refugee immigrants (model 2), unsponsored refugees to sponsored refugees (model 3) for deliveries from Model 1 Model 2 Model 3 Refugee Immigrant Non Refugee Immigrants Non Immigrants Refugee Immigrant Non Refugee Immigrants Non sponsored Refugee Immigrants N 29, , ,643 29, ,085 15,045 10,519 % SNM ( n cases) 4.62 (1,370) 4.17 (9,593) 5.01 (42,793) 4.62 (1,370) 4.17 (9,593) 4.68 (704) 4.58 (482) Risk Ratios (95% CI) Reference 1.10 Reference 1.06 Reference ( ) ( ) (1.00) ( ) (1.00) ( ) (1.00) ICC (%) N/A 1.47 a 0.91 Bolded values are statistically significant Model 1: Risk ratios adjusted for maternal age at delivery and parity Models 2 and 3: Risk ratios account for maternal country of birth (random effect), and adjust for maternal age at delivery, parity, education level, language ability and duration of residence (fixed effects) ICC=intra class correlation coefficient; N/A=not applicable; a covariance parameter estimate for country of birth significant 108 Sponsored Refugee Immigrants

127 SECTION A4.1: SUPPLEMENTARY FILES Table A4.1.1: ICD 10 CA diagnosis and CCI codes for severe neonatal morbidity (SNM) Description ICD 10 CA Diagnosis Code Canadian Classification of Health Intervention Codes Gestational age < 32 weeks (from maternal birth admission file) Birthweight < 1500 grams (from maternal birth admission file) Birth Trauma (intracranial hemorrhage paralysis due to brachial plexus injury, skull or long bone fracture) Broncho pulmonary dysplasia Cerebral infarction Hypoxic ischemic encephalopathy Intraventricular hemorrhage (grades 2,3,4) Necrotising enterocolitis Other respiratory: primary atelectasis respiratory failure Periventricular leukomalacia Pneumonia Respiratory distress syndrome Seizure Sepsis/septicaemia (streptococcus, staphylococcus, E.coli, unspecified Gram negative) P100 TO P103, P130, P132, P133, P140, P141 P271 I63 P915, P9181, P916 P521, P522 P77 P280, P285 P912 P23 (preferred PHAC code used), (codes J12 TO J18 used in LAIN et al, not used here) P220 P90, R56 P36 (preferred PHAC code used), (codes A40, A415, A419, B951, B962 used in LAIN et al, not used here) Any body cavity surgical procedure (substring) 109 1AA52, 1AA87, 1AC87, 1AE87, 1AF87, 1AG87, 1AJ87, 1AK87, 1AN52, 1AN59, 1AN87, 1AP59, 1AP72, 1AP87, 1AW59, 1AW72, 1AW87, 1AX87, 1BA72, 1BA80, 1BA87, 1BB72, 1BB80, 1BB87, 1BD72, 1BD80, 1BD87, 1BF80, 1BG72, 1BG80, 1BG87, 1BK59, 1BM72, 1BM80, 1BM87, 1BN72, 1BN80, 1BN87, 1BP72, 1BP80, 1BP87, 1BQ72, 1BQ80, 1BQ87, 1BS72, 1BS80, 1BS87, 1BT72, 1BT80, 1BT87, 1GA87, 1GA89, 1GB87, 1GB89, 1GD89, 1GE80, 1GE87, 1GE89, 1GE91, 1GH84, 1GJ86, 1GJ87, 1GK87, 1GK89, 1GM80, 1GM86, 1GM87, 1GN92, 1GR87, 1GR89, 1GR91, 1GT78, 1GT87, 1GT89, 1GT91, 1GV87, 1GV89, 1GW87, 1GX80, 1GX86, 1GX87, 1GY70, 1GY72, 1GY86, 1HJ76, 1HJ82, 1HN87, 1HP76, 1HP78, 1HP80, 1HP82, 1HP83,

128 Description ICD 10 CA Diagnosis Code Canadian Classification of Health Intervention Codes 1HP87, 1HR80, 1HR84, 1HR87, 1HS80 (excl. 1HS80G), 1HS90, 1HT80 (excl. 1HT80G), 1HT89, 1HT90, 1HU80 (excl. 1HU80G), 1HU90, 1HV80 (excl. 1HV80G), 1HV90, 1HW78, 1HW79, 1HX80, 1HX87, 1HX80, 1HZ87, 1IA76, 1IA80, 1IA87, 1IB76, 1IB79, 1IB80, 1IB82, 1IB87, 1IC76, 1IC80, 1IC82, 1IC87, 1ID76, 1ID80, 1ID82, 1ID86, 1ID87, 1IF83, 1IJ76, 1IJ80, 1IM76, 1IM80, 1IM82, 1IM83, 1IM87, 1IN83, 1IN84, 1IN87, 1JE57 (excl. 1JE57G), 1JE76, 1JE80, 1JE87, 1JJ76, 1JJ80, 1JK76, 1JK80, 1JK87, 1JW51 (excl. 1JW51G), 1JW57, 1JW76, 1LA84, 1LC84, 1LD84, 1NA72, 1NA74, 1NA76, 1NA77, 1NA80, 1NA84, 1NA86, 1NA87, 1NA88, 1NA89, 1NA90, 1NA91, 1NA92, 1NE80, 1NF76, 1NF78, 1NF80, 1NF82, 1NF84, 1NF86, 1NF87 (excl. 1NF87B), 1NF89, 1NF90, 1NF91, 1NF92, 1NK76, 1NK77, 1NK80, 1NK82, 1NK84, 1NK87 (excl. 1NK87B), 1NM74, 1NM76, 1NM77, 1NM80, 1NM82, 1NM87 (excl. 1NM87B), 1NM89, 1NM91, 1NP72, 1NP73, 1NP86, 1NQ74 (excl. 1NQ74B), 1NQ80, 1NQ84, 1NQ86, 1NQ87 (excl. 1NQ87B), 1NQ89, 1NQ90, 1NT80, 1NT84, 1NT86, 1NT87, 1NV89, 1OA87, 1OB87, 1OB89, 1OD76, 1OD89, 1OE76, 1OE80, 1OE89, 1OJ76 (excl. 1OJ76B), 1OJ87, 1OJ89, 1OK87, 1OK89, 1OK91, 1OT72, 1OT87, 1OT91, 1PB87, 1PB89, 1PC80, 1PC87 (excl. 1PC87D), 1PC89, 1PC91, 1PE57 (excl. 1PE57BD), 1PE80 (excl. 1PE80D), 1PE82, 1PE87 (excl. 1PE87D), 1PE89 (excl. 1PE89D), 1PG76, 1PG77, 1PG80 (excl. 1PG80D), 1PG86, 1PG89, 1PL74 (excl. 1PL74CD), 1PL80, 1PM79, 1PM86, 1PM87 (excl. 1PM87B), 1PM89, 1PM90, 1PM91, 1PM92, 1QE53, 1QE80, 1QE82, 1QE84, 1QE87, 1QE89, 1QG89, 1QM74, 1QM80, 1QM87, 1QM89, 1QM91, 1QN82, 1QT87, 1QT91, 1RB74, 1RB80, 1RB83, 1RB87, 1RB89,1RD89, 1RF51, 1RF72, 110

129 Description ICD 10 CA Diagnosis Code Canadian Classification of Health Intervention Codes 1RF74, 1RF80, 1RF87, 1RF89, 1RM87 (excl. 1RM87B), 1RM89, 1RM91, 1RN87, 1RN89, 1RS74, 1RS80, 1RS86, 1RS87, 1RS89, 1RW87, 1RW88, 1RW91, 1RW92, 1SA74, 1SA75, 1SA80, 1SA89, 1SC74, 1SC75, 1SC80, 1SC87, 1SC89, 1SE53, 1SE89 (excl. 1SE89D), 1SF80, 1SF87, 1SF89, 1SG80, 1SG87, 1SH87, 1SM74, 1SM80, 1SM87, 1SN87, 1SN93, 1SQ53, 1SQ74, 1SQ80, 1SQ87, 1SQ91, 1SQ93, 1SW74, 1SY80, 1SY84, 1SY87, 1SZ87, 1VA53, 1VA74, 1VA75, 1VA80, 1VA87, 1VA93, 1VC74, 1VC80, 1VC87, 1VC91, 1VC93, 1VE80, 1VG53, 1VG55, 1VG72, 1VG73, 1VG74, 1VG75, 1VG80, 1VG87, 1VG93, 1VK80, 1VK87, 1VK89, 1VL80, 1VL87, 1VM80, 1VM87, 1VN80, 1VN87, 1VP74, 1VP80, 1VP87, 1VP89, 1VQ74, 1VQ79, 1VQ80, 1VQ82, 1VQ87, 1VQ91, 1VQ93, 1VS72, 1VS80, 1VX87 Any intravenous fluids 1LZ35CAE6, 1LZ35HAC1, 1LZ35HAC5, 1LZ35HAC6, 1LZ35HAC7, 1LZ35HAE6, 1LZ35HAT7, 1LZ35HAT9, 1LZ35HAZ9, 1LZ35HHC1, 1LZ35HHC5, 1LZ35HHC6, 1LZ35HHC7, 1LZ35HHE0, 1LZ35HHE6, 1LZ35HHT7, 1LZ35HHT9, 1LZ35HHZ9, 1LZ35HRC5, 1LZ35HRC6, 1LZ35HRC7, 1LZ35HRT9, 1LZ35HRZ9 Central venous or arterial catheter 1KV53HACH, 1KV53HAFT, 1KV53LAFT, 2IM28GP, 2LZ28GQPL, 2LZ28GRPL, 2LZ28JAPL, 1KX53HACH, 1KX53HAFT, 1KX53 LAFT, 2LZ28GQPL, 2LZ28GRPL Pneumothorax requiring intercostal catheter Resuscitation Transfusion of blood or blood products 111 1GV52DA, 1GV52DATS, 1GV52HA, 1GV52HAHE, 1GV52HATK, 1GV52LA, 1GV52LATS, 1GV52LAXXE, 1GV54JATS, 1GV55JATS 1HZ30JN, 1HZ30JY, 1GZ30CJ, 1GZ30CJNB, 1GZ30JH 1LZ19HHU1A, 1LZ19HHU1J, 1LZ19HHU2A, 1LZ19HHU2J, 1LZ19HHU3J, 1LZ19HHU4J, 1LZ19HHU5J, 1LZ19HHU6A,

130 Description ICD 10 CA Diagnosis Code Canadian Classification of Health Intervention Codes 1LZ19HHU6J, 1LZ19HHU9A, 1LZ19HHU9J, 1LZ19HMU1, 1LZ19HMU2, 1LZ19HMU9, 1LZ35HAC5 Ventilatory support (mechanical ventilation and/or CPAP) 1GZ31CAEP, 1GZ31CAND, 1GZ31CAPK, 1GZ31CBND, 1GZ31CRND, 1GZ31GPND, 1GZ31JAGX, 1GZ31JAMD, 1GZ31JANC, 1GZ31JAPK 112

131 113 Table A4.1.2: Concurrent validity comparing Severe Neonatal Morbidity (SNM) subtypes (Canada) to Neonatal Adverse Outcome Indicator (NAOI) subtypes (Australia) among all live births (singletons and multiples) with gestational age 24 weeks GA at the birth admission Canada a Australia b Absolute Difference Percent Difference N 1,193, ,775 (CDN AUS) ((CDN AUS) /CDN)*100 Severe neonatal morbidity subtype % % % % Neonatal death 0.16 c 0.22 N/applicable N/applicable Gestational age < 32 weeks 1.05 N/A N/A N/A Resuscitation with intubation or CPR Birthweight < Respiratory distress syndrome Hypoxic ischemic encephalopathy Intraventricular hemorrhage (grade 2) Injury due to birth trauma Ventilatory support d Sepsis/septicaemia (Lain et al, 2012) Sepsis (PHAC, 2008) Seizure Pneumothorax with intercostal catheter Pneumonia (Lain et al, 2012) Pneumonia (P23 only, PHAC 2008) Surgical procedure Central line (incl. Umbilical vein or artery) Transfusion Necrotizing enterocolitis Any intravenous fluids Broncho pulmonary dysplasia Cerebral infarction 0.02 N/A N/A N/A Peri ventricular leukomalacia 0.02 N/A N/A N/A Other respiratory 0.35 N/A N/A N/A ANY SEVERE NEONATAL MORBIDITY e ANY SEVERE NEONATAL MORBIDITY PHAC e Australian proportions reported in Lain et al, 2012 unless otherwise indicated a Ontario (Canada) hospitalizations data from April 1, 2002 to March 31, NAOI index based on Ontario data referred to as SNM indicator b New South Wales (Australia) hospitalizations and midwife data from January 2001 to December Lain et al (2012) refers to index as NAOI c neonatal mortality not included in Canadian indicator since correlated with outcomes interested in predicting (i.e., perinatal mortality, infant mortality) d Value received through personal communication, not as reported in Lain et al, 2012 e proportion as reported in Lain et al, 2012 does not take into account revised value for ventilator support N/A = not available

132 114 Table A4.1.3: Comparing predictive validity of all live births with SNM in Canada and Australia: neonatal (0 27 days) and infant mortality (0 364 days) (as captured in death data from the Ontario Registrar General Death Database ( )) Live births with NAOI/SNM Live births without NAOI/SNM New South Wales, AU Infant Mortality (%) Ontario, CAN Neonatal Mortality % (n) 2.53 (1643) 0.03 (289) 98 Infant Mortality % (n) 3.42 (2,235) 0.08 (936) 41 Predictive ability of NAOI/SNM

133 115 Table A4.1.4: SNM Model 1 (refugee immigrants vs. non immigrants ) coefficients (n=1,113,367) Coefficient Estimate (SE) Risk Ratio (95% CI) Main Effects Refugees vs. Non Immigrants (0.027) 0.94 ( ) Non Refugees vs. Non Immigrants (0.011) 0.83 ( ) Confounders Maternal age (years) (0.022) 1.03 ( ) (0.014) 1.00 ( ) Reference (0.011) 1.03 ( ) (0.013) 1.20 ( ) (0.021) 1.51 ( ) Parity (0.010) 1.52 ( ) 1 (ref) Reference (0.015) 1.01 ( ) 3 or more (0.019) 1.16 ( ) Bolded values are statistically significant

134 116 Table A4.1.5: SNM Model 2 (refugee immigrants vs. non refugee immigrants ) coefficients (n=259,719) Coefficient Estimate (SE) Risk Ratio (95% CI) Main Effects Refugees vs. Non Refugees (0.032) 1.10 ( ) Confounders Maternal age (years) (0.119) 1.01 ( ) (0.037) 0.86 ( ) Reference (0.025) 1.12 ( ) (0.030) 1.41 ( ) (0.042) 1.87 ( ) Parity (0.022) 1.66 ( ) 1 (ref) Reference (0.033) 0.93 ( ) 3 or more (0.042) 0.95 ( ) Education 0 9 years (0.036) 1.14 ( ) years (0.029) 1.14 ( ) 13+ years (0.034) 1.11 ( ) trade certificate, non university diploma (0.031) 1.10 ( ) Bachelors, Masters, Doctorate Reference 1.00 Language Ability neither English or French (0.024) 1.00 ( ) one or both of English and French Reference 1.00 Duration of residence (years) (0.002) 1.01 ( ) Bolded values are statistically significant

135 117 Table A4.1.6: SNM Model 3 (non sponsored refugee immigrants vs. sponsored refugee immigrants) coefficients (n=25,564) Coefficient Estimate (SE) Risk Ratio (95% CI) Main Effects Non Sponsored refugees vs. Sponsored refugees (0.090) 1.06 ( ) Confounders Maternal age (years) (0.319) 0.83 ( ) (0.122) 0.86 ( ) Reference (0.080) 1.06 ( ) (0.089) 1.37 ( ) (0.117) 1.90 ( ) Parity (0.073) 1.58 ( ) 1 (ref) Reference (0.091) 0.97 ( ) 3 or more (0.094) 1.10 ( ) Education 0 9 years (0.124) 1.29 ( ) years (0.116) 1.25 ( ) 13+ years (0.137) 1.03 ( ) trade certificate, non university diploma (0.126) 1.24 ( ) Bachelors, Masters, Doctorate Reference 1.00 Language Ability neither English or French (0.086) 0.97 ( ) one or both of English and French Reference 1.00 Duration of residence (years) (0.007) 1.00 ( ) Bolded values are statistically significant

136 118 Table A4.1.7: Stratification by world region and sub regions of maternal birth SNM risk ratios (RR) (95% CI) comparing refugee vs. non refugee immigrant for deliveries from SNM (n cases=18,945) World region, sub regions and selected maternal countries of birth Total n Refugee % of all births Non Refugee % of all births Unadjusted a Risk Ratio (95% CI) Adjusted b Risk Ratio (95% CI) Africa 24, ( ) 1.11 ( ) Sub Saharan Africa 20, ( ) 1.10 ( ) North Africa ( ) 0.76 ( ) Asia 169, ( ) 1.08 ( ) Western & Central Asia 16, ( ) 1.06 ( ) South Asia 91, ( ) 1.06 ( ) East Asia 34, ( ) 1.07 ( ) S.E. Asia/Oceania Islands 26, ( ) 1.02 ( ) Latin America & Caribbean 28, ( ) 1.01 ( ) Europe 31, ( ) 1.10 ( ) Southern Europe 7, ( ) 0.95 ( ) Bolded values are statistically significant; a Unadjusted ratios: account for maternal country of birth (random effect) in regional and sub regional analyses; b Adjusted ratio: account for maternal country of birth (random effect) in regional and sub regional analyses, adjusted for maternal age at delivery, parity, education at arrival, language ability and duration of residence (fixed effects)

137 119 Table A4.1.8: Coefficient estimates (SE) comparing risk of SNM among refugee immigrants to nonrefugee immigrants, fully adjusted models with specified interaction No Interaction Duration No Interaction (0.0325) Education (0.0325) a p value on interaction term statistically significant (p<0.05) Full model: fcob (random effect), refugee status, maternal age, parity, education category, language ability and duration of residence

138 120 SECTION A4.2: SENSITIVITY ANALYSES Table A4.2.1: Comparing risk of SNM among refugee immigrants (vs. non immigrants and vs. non refugee immigrants) stratified by age at arrival categories (<15, 15 years old and BOTH age groups) SNM % (n) Adjusted Main Effect 15 yrs at arrival Adjusted Main Effect including BOTH <15 years at 15 years at BOTH Coefficient (SE) RR (95% CI) Coefficient (SE) RR (95% CI) arrival arrival Refugees vs (300) 4.62 (1,370) 4.58 (1,670) (0.0269) 0.94 ( ) (0.0244) 0.93 ( ) Non Immigrants 9.00 (18) 5.01 (42,793) 5.01 (42,811) Ref 1.00 Ref 1.00 Refugees vs (200) 4.62 (1,370) 4.58 (1,670) (0.0325) 1.10 ( ) (0.0296) 1.08 ( ) Non Refugees 4.86 (27,925) 4.17 (9,593) a 4.24 (10,949) Ref 1.00 Ref 1.00 Bolded values are statistically significant; Adjusted RR for vs. non immigrants includes maternal age and parity; Adjusted RR for vs. non refugees includes random effect for country of birth and maternal age, parity, education, language and duration of residence (fixed effects); a Chi square p value <0.05 for difference in SMM proportion between women <15 years old at arrival and only women 15 at arrival (tests rows)

139 121 Table A4.2.2: Comparing risk of SNM for refugee immigrants compared to non refugee immigrants, with and without Ontario Health Insurance Plan (OHIP) eligibility at the time of birth Refugee Non Refugee SNM RR Total in paper 29, ,085 SNM % (n) 4.62 (1,370) 4.17 (9,593) 1.10 Without OHIP eligibility at birth Total (n, % of total in paper) 1,829 (6.2) 4,263 (1.8) 6,092 Government Assisted 9 N/A Landed/Claimants 1,796 N/A Dependents 24 N/A SNM with OHIP at birth % (n) 4.68 (1,308) 4.16 (9,436) 1.13 SNM without OHIP at birth % (n) 3.94 (72) 4.74 (202) a 0.82 a significant difference between those without OHIP at birth and population included in manuscript Table A4.2.3: Comparing nesting structures: random effect for country of birth or for mother. Coefficient estimates and unadjusted odds ratios with 95% confidence intervals comparing risk of SNM among refugee immigrants to non refugee immigrants Unadjusted Main Effect Results Levels included Coefficient (SE) OR (95% CI) Model 2 None a (0.0296) 1.11 ( ) Refugees vs. Births within countries of birth b (0.0334) 1.11 ( ) Non Refugees Births within mothers b (0.0323) 1.12 ( ) Bolded values are statistically significant a model estimated using PROC GENMOD (LOG BINOMIAL) b models estimated using PROC NLMIXED (using starting parameters from None model, LOGIT)

140 CHAPTER 5 (MANUSCRIPT 3): THE MIGRATION JOURNEY OF REFUGEE IMMIGRANT WOMEN TO ONTARIO, CANADA AND PRETERM BIRTH IN OFFSPRING 122

141 ABSTRACT 123 Background: Women forced to flee their country of birth (refugees) and experience a prolonged migration journey prior to arriving in Canada due (secondary migrants), may be at higher risk of preterm birth (PTB) compared to those who voluntarily immigrate (non refugees). A population based cohort study was conducted in Ontario, Canada to determine: 1) if the relationship between refugee status and risk of PTB varied with secondary vs. primary migration; 2) if this relationship varied with maternal region of birth and; 3) if secondary migrants transition country contributes to PTB risk. Methods: Linked Ontario immigration ( ) and hospitalizations data ( ) were used. Multilevel cumulative probability models for ordinal outcomes (22 31, and 37 weeks) were used to estimate adjusted cumulative odds ratios (COR) with 95% confidence intervals (95% CI) with a random effect for country of birth. An interaction term between refugee status and secondary migration; models stratified by maternal region of birth; and models restricted to secondary migrants with a random effect for transition country, addressed the objectives of this study. Results: All secondary refugees (vs. secondary non refugees) experienced an excess in PTB risk over and above that experienced by primary refugees (vs. primary non refugees) (COR=1.59 ( ) for the former and COR=1.12 ( ) for the latter). A similar effect was seen among Asian immigrants. Transition country did not contribute to PTB risk among secondary migrants. Discussion: Secondary and primary refugees experienced higher risk of PTB compared to their nonrefugee counterparts, with secondary refugees experiencing the largest excess risk. Future research may benefit from examination of psychosocial risk factors and chronic stress. Key words: preterm birth, refugees, secondary migration

142 5.1 INTRODUCTION 124 Migration has been identified as an important social determinant of health 1 and for many migrants, is a multi step journey. Where, along the spectrum of voluntary (usually described as economic migration) to forced (usually described as refugee migration) 2 a migrant may fall, can also have an important impact on migrant health. Previous research suggests that women who voluntarily migrate to a transitional country before migrating to Canada have upwardly mobile socioeconomic profiles and thus experience a lower risk of preterm birth (PTB) compared to women who voluntarily migrate directly from their country of birth. 3 However, in the case of refugees, departure from their country of birth is often forced because of persecution, war or violence. 4 The amount of people displaced as result of persecution, conflict, generalized violence and human rights violations numbered 51.2 million people in Refugees may also experience a prolonged migration journey, including residence in one or more countries before reaching their final destination. It has been estimated that refugees are spending, on average, 17 years waiting for an alternative solution to their exile. 6 Such journeys may compound an already stressful existence for refugees as their lives remain uncertain and living and working conditions in transition countries may be unfavourable. Therefore, we hypothesize that secondary migration may have a different impact on the preterm birth risk for refugees when compared with non refugees. In terms of perinatal health, preterm birth (PTB) has been identified as an outcome that is particularly sensitive to social disadvantage. Epidemiological studies have identified large, persistent and robust disparities in PTB based on race (i.e., African American vs. white populations) 7,8 as well as large gradients in risk based on income, education and occupation in the United States. 9 Disparities in PTB by socio economic measures have also been reported in Sweden, UK, the Czech Republic and Finland. 10 In Quebec, Canada, significant gradients in risk of PTB were seen by neighbourhood income quintile and maternal education. 11 A recent study conducted in Sweden 12 with war refugees found that risk of PTB was significantly higher among refugees that gave birth in the first year of residence compared to the second year. The mechanism by which social disadvantage increases risk of preterm birth is not well understood but there has been increasing interest in the role of psychosocial stressors. 13 PTB is also a leading cause of perinatal morbidity and mortality. 14 Compared to infants born at term, preterm infants suffer from higher rates of respiratory distress and feeding difficulties in early life, neuro developmental disabilities and recurrent health problems in early childhood as well as cardiovascular and metabolic disorders in adulthood. 15

143 125 This study examines the interplay of secondary migration 16 (i.e., migration to transition country before migration to Canada) and refugee status (refugee immigrant vs. non refugee immigrant) on the risk of preterm birth. Our study had three objectives. The first objective was to determine whether the relationship between refugee status and risk of preterm birth subtypes varied with type of migration (i.e., secondary vs. primary migration). Secondly, we sought to identify whether this relationship varied by maternal region of birth. Finally we examined whether the secondary migrant s country of last permanent residence (CLPR) contributed to risk of PTB. 5.2 METHODS Study design and inclusion/exclusion criteria This was a population based cohort study. Refugee status (refugee immigrants compared to non refugee immigrants) was determined by identifying women who immigrated to Ontario, Canada and received permanent residency between April 1, 2002 and December 31, Our definition of refugee immigrants was consistent with the United Nations definition. 4 Non refugee immigrants had non refugee migration histories. Implicit in this study is that exposures prior to reaching reproductive age likely play an important role in any differences that may exist. To ensure that these exposures occurred prior to immigrating to Canada, women who were <15 years old at the time of arrival were excluded. All women included in this study were eligible for provincial health care insurance. All singleton infants born to these women in any Ontario hospital between April 1, 2002 and March 31, 2011 were included. Multiple deliveries to a given mother were excluded since the etiology for preterm birth is different for multiples than for singletons. 14 Births < 22 weeks gestation or with a birth weight < 500 grams were excluded Data Sources Two administrative databases held at the Institute for Clinical and Evaluative Sciences (ICES) in Toronto, Ontario were linked to conduct this research. See below for a detailed description of each database The Citizenship and Immigration Canada Database (CIC) is the official Canadian immigration registry, composed of all legal immigrants landing (arrival) records, in which the arrival date indicates the date of receiving permanent residency in Canada. Immigrants to Canada arrive in one of three classes: economic, family or refugee class. In this study non refugee immigrants refer to

144 126 economic and family class immigrants. The CIC database also contains information on refugee status, country of birth (COB), country of last permanent residence (CLPR), education level and knowledge of Canadian official languages (English and French) upon arrival to Canada. With the exception of language, which is self reported for non principal applicants, the remaining information is ascertained based on legal documentation provided by immigrants during their application process. The CIC linked database has been used in several epidemiological studies on topics such as diabetes mellitus, cervical screening, health care utilization and perinatal health. 3, The Discharge Abstract Database (DAD) originates from the Canadian Institute of Health Information (CIHI) and contains both maternal and infant hospital admission data. The DAD was used to derive a perinatal specific dataset created by deterministically linking maternal and infant hospitalization records, resulting in maternal infant delivery pairs. Gestational age (GA) at the time of birth contained in the DAD was used to identify babies who did and did not experience the outcome of interest, preterm birth (22 31 weeks and weeks, see below for rationale). This dataset also contains information on maternal age at the time of delivery and self reported parity Outcome of Interest Traditionally preterm birth (PTB) has been defined as birth as any birth less than 37 weeks gestational age. However, since infant health outcomes and survival differ across the range of preterm gestational ages and diverse causal mechanisms are implicated, investigation of PTB sub groups based on gestational age has been recommended. 8 In recognition of this, preterm birth was measured as a three category variable in this study, specifically very preterm (22 31 weeks), moderately preterm (32 36 weeks) and full term ( 37 weeks). The outcome was determined using the gestational age indicated on the DAD hospitalization record. Gestational age has been recorded in the DAD in completed weeks since 2002, and is based on the best clinical estimate in a woman s medical chart. Best clinical estimates of gestational age in the DAD are based on either the first day of the mother s last menstrual period (LMP) or early ultrasound measurements, the latter considered more precise. 21 The use of prenatal ultrasonography is the dominant method used for pregnancy dating in Ontario 19 with annual rates of ultrasonography utilization steadily rising between 1996 and In 2006, each pregnant woman in Ontario had approximately 2.2 ultrasound examinations in either the first or second trimester. 22

145 5.2.4 Exposure, Effect Modifier and Covariates Main Exposure Refugee Status 127 Refugee immigrants were compared to non refugee immigrants Effect Modifier The effect modifier of interest was secondary migration 16 and in this study refers to migrants who resided in a country (regardless of their legal status in that country at the time) other than their COB for 6 months in the 5 years prior to immigration to Canada. Secondary migrants are those that satisfy the above criteria while primary migrants are those that did NOT report residing in a country other than their COB in the 5 years prior to immigration to Canada. Primary migrants came to Canada directly from their COB. See Figure 1 for an illustration of secondary and primary migration Covariates The covariates listed below were conceptualized as having one or more purposes in this study: 1) to describe the population; 2) to stratify analyses and identify possible effect modification; 3) to indicate a contextual environment 23 that may contribute to individual PTB risk (used in this capacity in multi level models) or ; 4) to act as confounder of the relationship between refugee status and risk of PTB and therefore adjusted for in multivariate models. Definitions of all covariates examined and their function in the study are detailed below. Maternal birth place. Country of birth (COB) (describe, stratify, contextual), world sub region of birth (describe, stratify) and world region of birth (stratify). The United Nations country classification system was used to categorize countries into world sub regions and world regions. 24 Country of last permanent residence (CLPR) (contextual) is relevant for secondary migrants and refers to the country, other than the COB, in which an individual resided for 6 months in the 5 years preceding immigration to Canada regardless of their legal status while in that country. The United Nations country classification system was used as above (stratify). 24 Education at the time of arrival (confounder, as it is a determinant for economic immigrants and not a determinant for refugees, and it is also associated with PTB risk) was described as 0 9 years, 10 12

146 years, 13+ years (exclusive of other categories), trade certificate/non university diploma and bachelor/masters/doctorate. 128 Knowledge of official languages at arrival (confounder, same explanation as above) was categorized as knowledge of one or both official Canadian languages (English, French) and knowledge of neither official language. Maternal characteristics included were age at arrival (describe), maternal age at delivery described by 6 categories (15 19, 20 24, 25 29, 30 34, 35 39, 40+ years) (confounder) and duration of residence measured in years (confounder). Maternal obstetrical characteristics included parity (confounder) described by four categories (0, 1, 2 and 3 previous births) Analytic Methods Bivariate analyses were conducted to examine the distribution of two categories of preterm birth (very and moderately preterm), descriptors and confounders by refugee status and further stratified by secondary migration. Generalized linear mixed models were used to estimate the unadjusted and adjusted cumulative odds ratios (COR) 25 and 95% confidence intervals (95% CI) of preterm birth (22 31 weeks and weeks gestation) among refugee immigrants compared to non refugee immigrants within strata of secondary migration. For objective one, models were estimated all refugee immigrants compared to non refugee immigrants and for objective 2 models were further stratified by world sub regions of maternal birth. Regions with reasonably large population sizes were reported. All models accounted for maternal country of birth (as a random effect) and were adjusted for maternal age, parity, education level, language ability and length of residence in Canada (as fixed effects). As indicated above, all models included a random intercept for COB. Including this random intercept allowed us to take into account the potential non independence of the outcome among mothers from the same COB. 26 To ensure that modeling variation at the COB level was appropriate, the statistical significance of the covariance parameter estimate from an intercept only model (no other variables) was estimated by dividing the parameter estimate by its standard error. If this value was > 1.96, the outcome was deemed non independent among mothers from the same country of birth. Further, if significant non independence was found, the intra class correlation coefficient (ICC) was

147 129 calculated for the intercept only model to estimate the proportion of total variation (at the individual and group level) in preterm birth that is explained by variation at the group level (COB). For multi level cumulative odds models, the ICC is calculated using the following formula, ICC= / ( + 2 /3) where is 3.14 and is the covariance parameter estimate for the group level intercept. 27 If the random intercept coefficient was statistically significant ( =0.05), modeling at the COB level was deemed appropriate regardless of the magnitude of the ICC. 26 The ICC was interpreted as the average correlation in PTB risk between women from the same maternal COB. 23,26 The assumption of proportionality for cumulative probability models was assessed by verifying that the odds ratios in two binary logistic models (model 1: weeks vs & 37 weeks; model 2: weeks & weeks vs. 37 weeks) were not significantly different than each other for the exposure of interest. 25 In addition, the Score Test for the proportional odds assumption was also estimated using a cumulative odds model without a random effect for COB. Inability to reject the null hypothesis (p>0.05) indicated that the proportionality assumption was met. 25 For objective 3, models restricted to secondary migrants and including a random intercept for CLPR were also explored. These were estimated to identify whether women that transited through the same CLPR were similar with respect to risk of PTB. A significant ICC (described above) for these models signified that CLPR contributed to PTB risk among secondary migrants. In secondary analysis, the possible role of maternal co morbidities in contributing to preterm birth risk was explored. These co morbidities (likely mediators) consisted of diagnoses reported in the mother s DAD birth admission record and were categorized as systemic morbidities (i.e., hypertension, diabetes) and localized morbidities (i.e., hemorrhage, cervical incompetence). For a full list of these comorbidities please see the appendix to this chapter (Section A5.1 Table A5.1.3). Co morbidity risks among those that delivered preterm in each group were estimated along with risk ratios and 95% confidence intervals comparing refugee immigrants and non refugee immigrants. Co morbidities that had a strong association with preterm birth risk were considered for further mediation analysis. The proportion of all refugee and non refugee immigrant women born or transitioning through the same ( shared ) and different ( unshared ) countries were summarized for the top 10 countries (by frequency) in order to understand the contribution of specific maternal COBs and CLPRs in each group. Refugee CLPRs with populous refugee camps, 28 administered by the United Nations High Commissioner for Refugees (UNHCR) were also noted.

148 5.2.6 Research Ethics 130 This study received research ethics approval from: 1) The Institute for Clinical and Evaluative Sciences and Sunnybrook Health Sciences Centre; 2) St. Michael s Hospital Research Ethics Office; and 3) The University of Toronto Research Ethics Office. 5.3 RESULTS There were 123,856 singleton live births to immigrant women included in our study (Table 5.1). These consisted of 101,122 births to primary non refugee immigrants, 11,661 births to primary refugee immigrants, 9,773 births to secondary non refugee immigrants and 1,300 births to secondary refugee immigrants. Figure 2 summarizes the risk of preterm birth in each preterm birth category (22 31 weeks, weeks) for refugees and non refugees by secondary migration status. This figure illustrates a clear divergence in risk between refugees and non refugees based on secondary migration status, with the largest divergence seen in the moderately PTB category (32 36 weeks). For primary migrants, the risk of moderately PTB for non refugees and refugees was roughly the same (47.6 per 1000 and 43.6 per 1000, respectively). However for secondary migrants the proportion of moderately PTBs was quite different, at 47.9 per 1000 among non refugees and 59.2 per 1000 among refugees. Examination of the distribution of descriptors, confounders by refugee status and secondary migration revealed some important differences between the groups being compared (Table 5.1). Secondary refugee immigrants were more likely to be at a disadvantage compared to the other groups for a few characteristics including: parity of 3 or more at the time of index birth (16.2%) and low educational attainment (0 12 years) at the time of arrival (66.1%). There were stark differences for two covariates by refugee status only (see Table 5.1). In terms of sub region of maternal birth, greater proportions of secondary and primary refugee immigrants were born in Sub Saharan Africa (33.6% and 30.1%) compared to secondary and primary non refugee immigrants (9.4% and 4.4%). In multi level models examining the cumulative odds of preterm birth among refugee immigrants compared to non refugee immigrants, random intercept coefficients for COB were statistically significant for the all the immigrants model (covariance estimate=0.050, standard error (SE)=0.013) and Asian immigrants (covariance estimate=0.042, SE=0.016) but not for Sub Saharan African immigrants (covariance estimate=0.049, SE=0.037). The ICCs for these models were 1.49%, 1.27% and 1.49%, respectively, indicating very minimal correlation in PTB risk between mothers from

149 131 the same COB. Despite the lack of significance of the ICC in the Sub Saharan African model, the COB random intercept was retained for all models for consistency. The proportionality assumption was met for all models (all immigrants, Sub Saharan African immigrants and Asian immigrants) using both the two binary logistic model method and the Score test. Adjusted multilevel cumulative odds regression revealed statistically significant effect modification of the relationship between refugee status and the cumulative odds of preterm birth by secondary migration among all immigrants (p=0.063) (all coefficients in Table 5.2). The adjusted COR (95% CI) for preterm birth among secondary migrants (refugee immigrants vs. non refugee immigrants) was 1.59 ( ) and 1.12 ( ) for primary migrants. In models restricted to births to Sub Saharan African immigrant women (n=9,223) (Table 5.3), secondary migration was not a significant effect modifier (p=0.086) with adjusted CORs for secondary migrants and primary migrants of 2.08 ( ) and 1.31 ( ) respectively. In models restricted to births to Asian immigrants (n=86,304) (Table 5.3), secondary migration did act as an important effect modifier (p=0.021) of the relationship between refugee status and preterm birth. These adjusted CORs (95% CI) were 1.60 ( ) and 1.05 ( ) for secondary migrants and primary migrants, respectively. See Tables A5.1.1 and A5.1.2 in this chapter s appendix for coefficients and standard errors for all variables included in Sub Saharan African and Asian models. There was no significant correlation in PTB risk for secondary migrant women who transitioned through the same CLPR. For the all immigrant and Sub Saharan African secondary migrant models the random intercept coefficients for CLPR were found to be non significant with ICCs of 0.68% and 1.1% respectively. The model restricted to infants born to Asian mothers did not converge. Due to nonsignificance of the CLPR random intercept and convergence problems, models restricted to secondary migrants were not examined further. The exploration of maternal co morbidities as contributors to preterm risk did not identify any specific co morbidity or groups of co morbidities that could be responsible for the increased risk of preterm birth among secondary refugee immigrants (see appendix section A5.1 Table A5.1.4). Descriptive country level analysis summarizing countries that are shared and unshared between refugees and non refugees can be found in the appendix section A5.1 Table A DISCUSSION In this population based cohort study we found that secondary migration interacted significantly with refugee status to increase the risk of PTB. For immigrants as a whole as well as for Asian immigrants, there was a greater increase in risk of PTB among secondary refugee immigrants

150 132 compared to secondary non refugee immigrants than there was between primary refugee immigrants compared to primary non refugee immigrants. The large COR for secondary refugees vs. secondary non refugees is notable (COR=1.6), considering that this is of the same magnitude of some of the most strongly predictive risk factors for preterm birth including cigarette smoking, cocaine use, prepregnancy body mass index and bacterial vaginosis (RR ). 29 In addition, even though there was no significant effect modification by secondary migration among Sub Saharan African immigrants, the CORs between refugee immigrants and non refugee immigrants were of considerable magnitude (COR=2.1 for secondary migrants and 1.3 for primary migrants). A search of peer reviewed literature on the health effects of secondary or serial migration on refugees (or migration to more than one country or multi step migration 16 ) identified only one qualitative study. This study examined Burundian and Liberian refugee families who undertook secondary migration within the United States. This study examined the reasons secondary migration was undertaken and the impact it had on the family. 30 The most common reason cited for secondary migration within the United States was the need for larger and/or more affordable housing, followed by work opportunities and family re unification. Families that moved cited numerous impacts of secondary migration including loss of material and emotional support of local agencies, ethnic and religious communities, friends and support networks. They also cite a number of protective factors including an increased ability to support family in Africa, improved housing, family solidarity, family pride and neighborhood safety. In general, these findings are perhaps relatable to situations of international secondary refugee migration, specifically, in the factors which may make secondary migration to Canada desirable (i.e., opportunities for job security and a higher standard of living) contrasting with the elements of support that are lost. However, in many respects, these findings may not be generalizable to the secondary refugee immigrant women included in our study. The most apparent difference being that the majority of countries in which secondary refugee immigrants transitioned through are not industrialized countries (see appendix section A5.1 Table A5.1.1) and there are unlikely to be resources available and agencies responsible for promoting refugee integration as there are in the United States. In fact, in many transition countries, the host population may be hostile to hosting refugee populations, 31,32 actively preventing their social integration. Despite the lack of recognition of the importance of secondary refugee migration in the peerreviewed literature, in grey literature searches I was able to find reference to a study currently underway that stresses the potential impact of secondary migration on the health of refugee populations. 33 As such, we look forward to research findings released on this topic in the future.

151 5.4.1 Strengths & Limitations 133 To our knowledge this is the first study that has identified secondary migration as an important risk factor for preterm birth among refugees. A large study population of immigrants was used, with objectively ascertained information on legal refugee status and other important socio demographic characteristics of the immigrant population that are important to health. The population based hospital admission data included nearly all immigrant births taking place in the province of Ontario during the study period. Healthcare is universally available to all legal residents and Canadians citizens in Ontario ensuring that cost is not a prohibitive factor in a hospital birth. Less than 2% of all births in Ontario take place outside of the hospital (unattended or attended by a midwife) 34,35 with the vast majority being low risk births. It has been suggested that immigrant women in Canada have a stronger preference for hospital birth than Canadian women so there is likely less of a selection bias with using hospital admission data. Even if there was a differential preference for home birth between refugee immigrants and non refugee immigrants it is likely to be small and not of concern. In addition, there is a very low proportion (<1%) of missing values for most variables in both the immigration and hospitalizations data. An additional strength of the study is that the study period spanned a limited time window ( ) in which immigrants had to both arrive in Ontario and give birth in order to be included in the study. This narrow window of time lessens the possible impact of cohort effects related to immigration waves. Using this time period also permitted us to take advantage of continuous gestational age and parity at the time of birth reported in the hospitalizations database beginning on April 1, This study also had some limitations. With respect to absence of potentially important data, we describe below the relevance of data on behavioural risk factors (i.e., infection and maternal nutrition) as well as various measures of psychosocial stress that may be important in understanding the findings. The lack of data on behavioural risk factors may be counteracted by the potential correlation with maternal education at the time of arrival and country of birth. However, in an attempt to more effectively control for behavioural factors associated with education, women who were < 15 years of age at the time of arrival were excluded from these analyses. The thought being that educational attainment may change less among those that were 15 years and that education at arrival may continue to be a correlate of determinants of PTB after migration. It is still possible that educational attainment increased over time even among this group of immigrant women. There is evidence to suggest that adult refugee immigrants to Canada are less likely than skilled immigrants to participate in post migration education while adult skilled workers (the largest proportion of non refugee immigrants) are the most likely of all groups to participate in post migration education. 36 In the current

152 134 study, within the first five years since immigrating, higher levels of education among refugee and nonrefugee immigrants generally tended to increase risk of PTB (see appendix section A5.1 Table A5.1.7). If refugee immigrants and non refugee immigrants in all arrival education categories improved their education after immigration but refugee immigrants slightly less so than non refugee immigrants based on trends seen in Banerjee & Verma (2009) 36, this should cause PTB risk to slightly increase among refugee immigrants with a slightly higher increase in risk of PTB among non refugee immigrants (based on trends seen in our data in the first five years since immigration) and an overall decrease in the COR. Considering these pieces of information, potential misclassification of education with increased duration of residence in our study may cause reported CORs in our study to be biased slightly away from the null. We also recognize that secondary migration (having lived in a country other than the country of birth for at least six months in the five years prior to immigration) may be a misnomer for some immigrants. It is possible that some secondary migrants in this study lived in more than one other country in the five years prior to immigrating to Canada but reported the most recent country during immigration. In this way these secondary migrants may be more aptly named serial migrants. 16 Additionally, primary migrants may have lived in another country in the five years prior to immigration and therefore may have been secondary or serial migrants in their distant past. However, regardless of this potential misclassification, the secondary migration indicator does capture those who experienced at least one migration event that lasted 6 months in the previous five years. We were also unable to capture the length of time that secondary migrants resided in the stated transition country. Finally the results of this study are not necessarily generalizable to unsuccessful refugee claimants and others without permanent residency status in Canada. Lack of status in Canada likely has important implications for stress and prohibits access to quality health care (without cost) 37,38, the effects of which may have a differing impact on risk of PTB than results shown in this study Interpretation For all immigrant women and immigrant women born in Asia, secondary migration significantly modified the relationship between refugee status and preterm birth among immigrants arriving to and giving birth in Ontario between 2002 and Consistent with significant effect modification, models restricted to all secondary migrants demonstrated significantly greater disparity in the cumulative odds of PTB between refugee immigrants and non refugee immigrants (COR=1.59, 95%: ) compared to models restricted to all primary migrants (COR=1.12, 95% ). For Asian immigrants, the cumulative odds of PTB was significantly higher among secondary refugee immigrants compared to

153 135 secondary non refugee immigrants (COR=1.60, 95% CI: ) but not when comparing primary refugees to primary non refugees (COR=1.05, 95% CI: ). Among immigrant women born in Sub Saharan Africa significant effect modification of refugee status by secondary migration was not present. However, wide confidence intervals for the COR estimated in the secondary migrant stratum suggest insufficient sample sizes may be responsible for absence of effect modification rather than the true absence of effect. Importantly, the cumulative odds of PTB among women born in Sub Saharan Africa was significantly higher among refugee immigrants compared to non refugee immigrants in both strata of secondary migration but with a larger point estimate in the secondary migrant stratum (COR=2.08, 95% CI: ) than in the primary migrant stratum (COR=1.31, 95% CI: ). Random intercepts included for COB suggest that the correlation in PTB risk between women from the same COB was small, but significant (range of ICCs: %). In other words, the contextual environment of the maternal country of birth and its associated characteristics (e.g., health care access) exerted some influence on the risk of PTB among women born in that COB. Given the larger excess risk of PTB between refugees and non refugees in the secondary migrant stratum, the possible contribution of the CLPR was explored. Like COB, theoretically, PTB risk may be similar among women who transited through the same CLPR due to similar exposures or experiences based in that country (i.e., exposure to refugee camps) that influence PTB risk. However, this was shown to not be the case with non significant ICCs in models restricted to secondary migrants. The potential contribution of maternal co morbidities to explain the increased risk of PTB among secondary refugees compared to secondary non refugees was also explored. No individual comorbidity of group of co morbidities was significantly elevated among preterm births to secondary refugee immigrants compared to secondary non refugee immigrants. In fact, most co morbidities that had high enough counts to be reported had proportions that were lower among secondary refugee immigrants. This supplementary analysis was limited by the small number of preterm births among secondary refugee immigrants. Also, there may be an under reporting bias present for these diagnoses given this is a limitation of using individual diagnoses in hospitalization databases. 39 In summary, secondary migration is an important risk factor for PTB among refugee immigrants, the strength of which appears to vary by region of maternal birth. Among secondary migrants, there was little correlation in PTB risk among women that transitioned through the same CLPR, suggesting the CLPR did not contribute to risk of PTB among secondary migrants.

154 5.4.3 Possible explanations for increased risk 136 Despite extensive epidemiologic research into numerous candidate causes of preterm birth, scientific understanding of the etiology of preterm birth is limited 10 and few strongly predictive risk factors have been identified. 8 Behavioural risk factors that have shown the most consistent associations are for cigarette smoking (RRs from 1.2 to 2.0), cocaine use (RR 2.0), pre pregnancy body mass index (related to diet and nutrition) (RR 1.5) and bacterial vaginosis (RR 2). 29 A substantial amount of research has also considered the effects of alcohol consumption, caffeine, illicit drug use, diet and nutrition including supplements as well as physically strenuous employment or physical activity. 8,29 Many of these behavioural risk factors are implicated in risk of PTB among socially disadvantaged populations. 29 A few small, mainly descriptive studies have shown that these risk factors are prevalent among refugee immigrants. In Canada and the US refugee studies have reported hunger and malnutrition during the migration process and low pregnancy weight gain among asylum seekers Infectious diseases have also shown to be common among refugees in Canada. 40,41,43 It has also been stated, however, that behavioural risk factors such as these do not entirely explain increased risk of PTB among socially disadvantaged populations. 44 Considering the more often cited healthy behaviours of immigrant women to Canada, 45,46 it seems unlikely that the excess risk among refugees in our study can be completely attributed to potential infections and nutritional deficiency which also may already be partially adjusted for by maternal education or considered through the contextual influence of the COB. In light of the limited ability of behavioural risk factors to explain increases in PTB among socially disadvantaged populations, there has been increasing scientific interest in the constellation of factors that encompass psychosocial stress. 8,13,14,29 The Institute of Medicine s review of preterm birth 29 defines stress as demands that tax or exceed the adaptive capacity of an organism that result in psychological and biological changes. In terms of specific stressors, the most consistent evidence is for major life events (e.g., a death in the family), exposure to chronic stress (e.g., accumulation of multiple stressors over the life course), maternal anxiety, racism, domestic or personal violence and the intendedness of pregnancy. 29 Within the context of our study findings, all of these psychosocial risk factors may be disproportionately present among refugee immigrants. A recently published study examining stressrelated disorders among Malian refugees in Burkina Faso (n=118 males; n=290 females) found that 83% of participants had a family member killed in the war and 83% experienced personal injury or physical damage. 47 A systematic review examining gender based violence (GBV) in complex emergencies noted that rates of GBV were high and that intimate partner violence was much more common than rates of war time rape. 48 Unintended pregnancies may be particularly problematic for refugees given that

155 137 conflict situations may limit contraceptive supplies 49 and potentially high rates of GBV may lead to increases in unintended pregnancy. It seems possible that increased presence of these (and other) psychosocial risk factors could be responsible for the generally increased risk of PTB among refugee compared to non refugee immigrants in this study. The substantial increase in PTB risk among secondary refugees compared to secondary nonrefugees suggests additional risk factors that are unique to secondary refugees and separate from those experienced by primary refugees (who share experiences and exposures from the COB). The fact that there is no significant correlation in PTB risk among women that transition through the same CLPR, may suggest a risk factor that broadly affects all secondary refugees regardless of the CLPR. A subtle but important difference between secondary and primary refugees is the elapsed time between departure from their COB and arrival in Canada, with primary refugees experiencing virtually no lag and secondary refugees experiencing a lag of undetermined length. Considering both the psychosocial stress that likely affects both primary and secondary refugees and this lag time aspect, it seems reasonable to suggest that primary refugees experience of psychosocial stress could be shorter in duration (thus, acute) or at least attenuated by arrival in Canada, compared to that of secondary refugees whose environment is destabilized for longer. Conversely, among secondary refugees, the psychosocial stress that accompanies refugee status combined with the prolonged experience of stress as a secondary migrant perhaps related to lack of a durable solution 50 may interact to produce chronic stress. This exposure to chronic stress and possibly weakened adaptive resources (i.e., reduced coping skills and social support) 10,51,52 among secondary refugees may explain the substantially increased risk of PTB in this group. A large body of evidence suggests that maternal chronic stress contributes to increased risk of PTB. 51 It has been hypothesized that chronic stress can initiate early labour through the heightened activity of chemical messengers involved in the stress response. Presence of these chemical messengers influences higher production of corticotrophin releasing hormone (CRH) by the placenta, which in turn plays an important role in the preparation for and the initiation of events that are required for birth CONCLUSION This study found that: 1) refugee immigrants who migrated elsewhere prior to arrival in Canada had a 1.6 times greater risk of PTB than non refugee immigrants that also migrated elsewhere prior to arrival in Canada, and 2) the CLPR that secondary migrant women transitioned through, prior to coming to Canada, had no influence on PTB risk. One may speculate that chronic stress resulting from forcible displacement, psychosocial risk factors that are not resolved quickly (due to secondary migration) and a

156 138 weakened adaptive response may be responsible for excess risk of PTB among secondary refugees. This hypothesis should be confirmed with detailed assessment of chronic stress and confounding factors along with PTB risk among secondary refugees. However considering the magnitude of elevated risk and the sheer global volume of refugees in protracted situations (6.4 million at the end of 2012), 6 it may be advisable to evaluate if changes in policies and procedures can help mitigate the increased risk of PTB in this group. For example, international and national policies that encourage faster local integration or identification of resettlement opportunities 50 for those refugees that have fled to a transition country may have a role in reducing risk of PTB among other adverse outcomes. It may be valuable to develop, test and implement programs and services to reduce chronic stress and enhance resiliency 53 among secondary refugees. From a health care perspective, clinicians and health care workers that serve refugee communities may benefit from discussing the migration and related health histories of their clients and implement strategies to prevent preterm birth such as appropriate, timely and high quality prenatal care.

157 REFERENCES International Organization for Migration (IOM). Migration: a social determinant of the health of migrants [Internet]. IOM Migration Health Department; Available from: 2. Castles S. Towards a sociology of forced migration and social transformation. Sociology. 2003;37(1): Urquia ML, Frank JW, Glazier RH. From places to flows. International secondary migration and birth outcomes. Soc Sci Med. 2010;71(9): United Nations University. Protracted refugee situations Political, human rights and security implications. New York, New York: United Nations University Press; United Nations High Commissioner for Refugees. War s Human Cost UNHCR Global Trends 2013 [Internet] Jun. Available from: 6. United Nations High Commissioner for Refugees. Displacement The new 21st century challenge, UNHCR Global Trends 2012 [Internet] Available from: 7. Blumenshine P, Egerter S, Barclay CJ, Cubbin C, Braveman PA. Socioeconomic disparities in adverse birth outcomes. Am J Prev Med Sep;39(3): Buck Louis GM, Platt RW. Reproductive and perinatal epidemiology. New York, New York: Oxford University Press; Parker JD, Schoendorf KC, Kiely JL. Associations between measures of socioeconomic status and low birth weight, small for gestational age, and premature delivery in the United States. Ann Epidemiol. 1994;4(4): Kramer MS, Goulet L, Lydon J, et al. Socio economic disparities in preterm birth: causal pathways and mechanisms. Paediatr Perinat Epidemiol. 2001;15(s2): Luo Z C, Wilkins R, Kramer MS, Fetal and Infant Health Study Group of the Canadian Perinatal Surveillance System. Effect of neighbourhood income and maternal education on birth outcomes: a population based study. CMAJ. 2006;174(10):

158 Liu C, Urquia M, Cnattingius S, Hjern A. Migration and preterm birth in war refugees: a Swedish cohort study. Eur J Epidemiol Jan 14;29: Dunkel Schetter C. Psychological science on pregnancy: Stress processes, biopsychosocial models, and emerging research issues. Annu Rev Psychol. 2011;62: Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. The Lancet. 2008;371(9606): Saigal S, Doyle LW. An overview of mortality and sequelae of preterm birth from infancy to adulthood. The Lancet. 2008;371(9608): Urquia ML, Gagnon AJ. Glossary: migration and health. J Epidemiol Community Health. 2011;65(5): Creatore MI, Moineddin R, Booth G, et al. Age and sex related prevalence of diabetes mellitus among immigrants to Ontario, Canada. Can Med Assoc J. 2010;182(8): Lofters A, Moineddin R, Hwang S, Glazier R. Predictors of low cervical cancer screening among immigrant women in Ontario, Canada. BMC Womens Health. 2011;11(20). 19. Urquia M, Frank J, Moineddin R, Glazier R. Immigrants duration of residence and adverse birth outcomes: a population based study. BJOG Int J Obstet Gynaecol Apr;117(5): Urquia ML, Frank JW, Moineddin R, Glazier RH. Does time since immigration modify neighborhood deprivation gradients in preterm birth? A multilevel analysis. J Urban Health. 2011;88(5): Joseph KS, Fahey J. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute for Health Information. Chronic Dis Can. 2009;29(3): You J, Alter D, Stukel T, et al. Proliferation of prenatal ultrasonography. CMAJ. 2010;182: Diez R. A glossary for multilevel analysis. J Epidemiol Community Health. 2002;56(8): United Nations Statistics Division. Composition of macro geographical (continental) regions, geographical sub regions, and selected economic and other groups [Internet] [cited 2014 Apr 3]. Available from: Agresti A. Categorical data analysis. 3rd ed. Wiley; 2012.

159 Merlo J, Chaix B, Yang M, Lynch J, R\aastam L. A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. J Epidemiol Community Health. 2005;59(6): Snijders TA, Bosker R. Multilevel analysis: an introduction to basic and advanced multilevel modeling. 2nd ed. SAGE Publications Ltd; Fifty Largest Refugee Camps A story map presented by Esri [Internet]. [cited 2014 Apr 3]. Available from: camps/ 29. Behrman RE, Butler AS. Preterm birth: causes, consequences, and prevention. National Academies Press; Weine SM, Hoffman Y, Ware N, et al. Secondary migration and relocation among African refugee families in the United States. Fam Process. 2011;50(1): Kirui P, Mwaruvie J. The dilemma of hosting refugees: A focus on the insecurity in north eastern Kenya. Int J Bus Soc Sci. 2012;3(8): Mercy Corps. Analysis of host community refugee tensions in Mafraq, Jordan [Internet] Oct. Available from: Unsafe havens? Health risks for refugees University of Cambridge [Internet] [cited College of Midwives of Ontario. The facts about home birth in Ontario [Internet]. Available from: OURCES_2012.pdf 35. Danielle Bachove. Are home births safe? McLeans [Internet] Aug 26 [cited 2014 Apr 3]; Available from: try this at home/ Apr 30]. Available from: havens health risks forrefugees 36. Banerjee R, Verma A. Determinants and effects of post migration education among new immigrants in Canada [Internet] Feb. Report No.: 11. Available from: %20Banerjee%20&%20Verma.pdf

160 Rousseau C, Ricard Guay A, Laurin Lamothe A, Gagnon AJ, Rousseau H. Perinatal health care for undocumented women in Montreal: When sub standard care is almost the rule. J Nurs Educ Pract [Internet] Jan 8 [cited 2014 Apr 6];4(3). Available from: Wilson Mitchell K, Rummens J. Perinatal outcomes of uninsured immigrant, refugee and migrant mothers and newborns living in Toronto, Canada. Int J Environ Res Public Health May 31;10(6): Ford JB, Roberts CL, Algert CS, Bowen JR, Bajuk B, Henderson Smart DJ. Using hospital discharge data for determining neonatal morbidity and mortality: a validation study. BMC Health Serv Res. 2007;7(1): Kahler LR, Sobota CM, Hines CK, Griswold K. Pregnant women at risk: An evaluation of the health status of refugee women in buffalo, New York. Health Care Women Int Jan;17(1): Oxman Martinez, J, Hanley, J. Health and social services for Canada s multicultural population: challenges for equity. Ottawa: Heritage Canada; United States Department of Health and Human Services. Bhutanese refugee health profile [Internet] Jan. Available from: health profile.pdf 43. Pottie K, Janakiram P, Topp P, McCarthy A. Prevalence of selected preventable and treatable diseases among government assisted refugees. Can Fam Physician. 2007;53: Wadhwa PD, Entringer S, Buss C, Lu MC. The contribution of maternal stress to preterm birth: issues and considerations. Clin Perinatol Sep;38(3): Project for an Ontario Women s Health Evidence Based Report. Social determinants of health and populations at risk Chapter 12 [Internet] Feb. Available from: SDOHandPopsatRisk.pdf 46. Toronto Public Health, Access Alliance Multicultural Health and Community Services. The global city: newcomer health in Toronto [Internet] Nov. Available from: standards/map/f iles/pdf/global_city.pdf

161 47. Carta MG, Oumar FW, Moro MF, et al. Trauma and stressor related disorders in the Tuareg refugees of a camp in Burkina Faso. Clin Pract Epidemiol Ment Health CP EMH. 2013;9: Stark L, Ager A. A systematic review of prevalence studies of gender based violence in complex emergencies. Trauma Violence Abuse Jul 1;12(3): Reproductive health response in crises consortium. Emergency contraception for conflict affected settings: a reproductive health response in conflict consortium distance learning module [Internet]. Available from: United Nations High Commissioner for Refugees The UN Refugee Agency, Northern Europe. Durable Solutions [Internet] [cited 2014 Apr 3]. Available from: we do/durable solutions.html 51. Latendresse G. The interaction between chronic stress and pregnancy: preterm birth from a biobehavioral perspective. J Midwifery Women s Health. 2009;54(1): Orr ST. Social support and pregnancy outcome: a review of the literature. Clin Obstet Gynecol. 2004;47(4): Gagnon AJ, Carnevale F, Mehta P, Rousseau H, Stewart DE. Developing population interventions with migrant women for maternal child health: a focused ethnography. BMC Public Health. 2013;13(1):471.

162 TABLES Table 5.1: Study characteristics for primary non refugee immigrants, primary refugee immigrants, secondary non refugee immigrants and secondary refugee immigrants Primary Migrants (N=112,783) Secondary Migrants (N=11,073) Characteristic (n, % of column unless otherwise indicated) Non Refugee (N=101,122) Refugee (N=11,661) Non Refugee (N=9,773) Refugee (N=1,300) Maternal age at birth (years) (0.8) 210 (1.8) 65 (0.7) 29 (2.2) ,423 (14.3) 1,687 (14.5) 1,114 (11.4) 211 (16.2) ,361 (35.0) 3,448 (29.6) 2,832 (29.0) 389 (29.9) ,140 (32.8) 3,759 (32.2) 3,464 (35.4) 413 (31.8) ,766 (14.6) 2,071 (17.8) 1,901 (19.5) 217 (16.7) 40+ 2,653 (2.6) 486 (4.2) 397 (4.1) 41 (3.2) Parity (previous births) none 53,518 (52.9) 3,990 (34.2) 4,035 (41.3) 452 (34.7) 1 35,708 (35.3) 3,944 (33.8) 3,601 (36.8) 397 (30.5) 2 8,890 (8.8) 2,134 (18.3) 1,474 (15.1) 240 (18.5) 3 2,986 (3.0) 1,585 (13.6) 662 (6.8) 211 (16.2) Language ability English 65,015 (64.3) 7,482 (64.2) 7,062 (72.3) 728 (56.0) French 1,037 (1.0) 367 (3.2) 122 (1.3) 45 (3.5) Both English & French 3,566 (3.5) 342 (2.9) 639 (6.5) 39 (3.0) Neither 31,504 (31.2) 3,470 (29.8) 1,950 (20.0) 488 (37.5) Education level 0 9 years 9,344 (9.2) 3,297 (28.3) 1,149 (11.8) 420 (32.3) years 18,584 (18.4) 3,614 (31.0) 1,687 (17.3) 440 (33.9) 13+ years a 10,682 (10.6) 1,368 (11.7) 809 (8.3) 123 (9.5) Trade certificate, non university diploma 14,664 (14.5) 1,752 (15.0) 1,465 (15.0) 176 (13.5) Bachelors, Masters, Doctorate 47,848 (47.3) 1,630 (14.0) 4,663 (47.7) 141 (10.9) Region of maternal birth Sub Saharan Africa 4,398 (4.4) 3,510 (30.1) 921 (9.4) 437 (33.7) South Asia 42,769 (42.3) 3,208 (27.5) 2,6246 (26.9) 287 (22.1) Latin America & Caribbean 8,193 (8.1) 1,757 (15.1) 455 (4.7) 137 (10.5) Middle East/North Africa/Central Asia 7,356 (7.3) 1,177 (10.1) 2,707 (27.7) 196 (15.1) Eastern Europe 5,505 (5.4) 422 (3.6) 1,051 (10.8) 79 (6.1) Southeast Asia, Oceania Islands 10,768 (10.7) 154 (1.3) 667 (6.8) 72 (5.5) 144

163 East Asia 15,188 (15.0) 1,214 (10.4) 470 (4.8) 34 (2.6) Southern Europe 2,058 (2.0) 188 (1.6) 397 (4.1) 31 (2.4) Industrialized b 4,887 (4.8) 31 (0.3) 473 (4.8) 27 (2.1) Region of last permanent residence Middle East/North Africa/Central Asia 4,068 (41.7) 316 (24.4) Sub Saharan Africa 394 (4.0) 310 (24.1) South Asia 1,015 (10.4) 227 (17.6) Industrialized c 2,501 (25.6) 198 (15.3) Not Applicable Not Applicable Southeast Asia, Oceania Islands 178 (1.8) 80 (6.2) Eastern Europe 429 (4.4) 52 (4.0) Southern Europe 465 (4.8) 42 (3.3) Latin America & Caribbean 327 (3.4) 35 (2.7) East Asia 374 (3.8) 32 (2.5) Mean age at arrival (standard deviation) 27.1 (4.9) 27.8 (5.7) 28.2 (5.1) 27.3 (5.7) Mean duration of residence 2.2 (1.9) 2.0 (2.0) 2.1 (1.9) 2.1 (2.1) (standard deviation) Duration of residence at birth (years) 4 c 86,905 (86.0) 10,026 (86.0) 8,548 (87.5) 1,099 (84.5) ,217 (14.1) 1,635 (14.0) 1,225 (12.5) 201 (15.5) a. 13+ age category is exclusive of other categories, did not complete a subsequent qualification; b Industrialized Countries = North America, Western and Northern Europe, Australia & New Zealand; c 4 category includes a small proportion of women who delivered prior to receiving permanent residency and prior to eligibility for the Ontario Health Insurance Plan (OHIP) 145

164 146 Table 5.2: Cumulative odds coefficient estimates (standard errors) and cumulative odds ratios (95% confidence intervals) for preterm birth subtypes (22 31 weeks, weeks), all immigrants multivariable model Coefficient Estimate (SE) Cumulative Odds Ratio (95% CI) Main Effects Secondary Refugees vs. Secondary Non Refugees (0.121) 1.59 ( ) Primary Refugees vs. Primary Non Refugees (0.048) 1.12 ( ) Secondary Refugees vs. Primary Refugees (0.117) 1.31 ( ) Secondary Non refugees vs. Primary Non Refugees (0.051) 0.93 ( ) Confounders Maternal age (years) (0.118) 1.39 ( ) (0.043) 0.95 ( ) Reference (0.033) 1.17 ( ) (0.040) 1.48 ( ) (0.066) 2.09 ( ) Parity (0.029) 1.24 ( ) 1 Reference (0.047) 0.92 ( ) (0.067) 0.87 ( ) Education 0 9 years (0.049) 1.13 ( ) years (0.038) 1.17 ( ) 13+ years (0.045) 1.12 ( ) Trade certificate, Non university diploma (0.040) 0.98 ( ) Bachelors, Masters, Doctorate Reference 1.00 Language ability neither English or French (0.033) 0.98 ( ) one or both of English and French Reference 1.00 Duration of residence (years) (0.007) 1.01 ( ) Bolded values indicate statistical significance.

165 Table 5.3: Adjusted cumulative odds ratios (ACOR) and 95% confidence intervals (95% CI) for preterm birth subtypes (22 31 weeks, weeks), Sub Saharan African immigrants (n=9,223) and Asian immigrants (n=86,304) ACOR (95% CI) Main Effects Sub Saharan African a (n=9,223) Asian a (n=86,304) Secondary Refugees vs. Secondary Non Refugees 2.08 ( ) 1.60 ( ) Primary Refugees vs. Primary Non Refugees 1.31 ( ) 1.05 ( ) Secondary Refugees vs. Primary Refugees 1.39 ( ) 1.39 ( ) Secondary Non refugees vs. Primary Non Refugees 0.88 ( ) 0.91 ( ) Bolded values indicate statistical significance. CORs account for maternal country of birth (random effect), and adjust for maternal age at delivery, parity, education level, language ability and duration of residence (fixed effects) a random intercept co efficient for country of birth significant 147

166 FIGURES Figure 5.1: Illustrative examples of the migration journey for secondary refugees, primary refugees, secondary non refugees and primary non refugees to Canada SOMALI SECONDARY REFUGEE SOMALI PRIMARY REFUGEE BRITISH SECONDARY NON REFUGEE BRITISH PRIMARY NON REFUGEE TRANSITION THROUGH KENYA TRANSITION THROUGH UNITED ARAB EMIRATES CANADA Figure 5.2: Secondary migration and refugee status interact to increase the risk of very (22 31 weeks) and moderate (32 36 weeks) preterm birth Risk /1 000 singleton livebirths Figure 5.2: Secondary migration and refugee status interact to increase the risk of very and moderate preterm birth 47.9/ / / /1 000 NON REFUGEE 59.2/ / / /1 000 REFUGEE moderate preterm birth (32 36 weeks) very preterm birth (22 31 weeks) Secondary Immigrants Primary Immigrants 1 148