Self-reported long-standing psychiatric illness and intake of benzodiazepines. A comparison between foreign-born and Swedish-born people

Transcription

1 O R I G I N A L A R T I C L E S Self-reported long-standing psychiatric illness and intake of benzodiazepines A comparison between foreign-born and Swedish-born people L BAYARD-BURFIELD, ]. SUNDQU1ST, S-E. JOHANSSON * Background: The objective of this paper was to analyse whether different groups of foreign-bom individuals have a higher risk of self-reported, long-standing, psychiatric illness or an increased Intake of benzodiazepines when compared to native Swedes. Methods: The present cross-sectional study is based on eight simple random samples of people aged years who participated in the Swedish Survey of Living Conditions. It was analysed by unconditional logistic regression in order to estimate odds ratios of psychiatric illness, based on 36,890 persons interviewed in and intake of benzodiazepines, based on 9,352 persons interviewed in Furthermore, the population prevalences were also calculated. Results: Migrants from and men from Eastern Europe and non- outside Europe showed an Increased risk of serf-reported, long-standing, psychiatric illness. Moreover, female migrants from Eastern Europe and non-westernized countries outside of Europe and Finnish-born men demonstrated an increased risk of intake of benzodiazepines when compared to native Swedes. Conclusions: Our hypothesis that sodoeconomic factors (low educational level, living alone, poor social network and poor economic resources) could explain the relation between the migrants' country of birth and psychiatric illness proved to be wrong., which is dosely related to the migration process and acculturation stress, is an independent factor on its own associated with an increased risk of long-standing, psychiatric illness. Keywords: foreign-born, intake of benzodiazepines, self-reported psychiatric illness T»he stress of migration and the effort of assimilating to a new society are likely to cause health problems. Feelings of alienation, social degradation, discrimination, xenophobia and social and cultural distance or isolation increase vulnerability to mental illness. 1 The number of immigrants in the Swedish population has steadily increased. Between 1975 and 1996, the number of foreign-born people living in increased from 550,451 (6.7%) to 943,804 (10.7%). The largest foreignborn group in 1996 was Finnish-born labour migrants (n=203,371; 21.5% of the foreign-bom population) followed by Southern European labour migrants (n=95,329; 10.1% of the foreign-born population). The amount of immigrant Scandinavians and Europeans has remained fairly constant at around 500,000. The increasing number of foreign-born people in die past 30 years is explained by refugees (and later their closest relatives) * L Bayard-Burfield 1, J. Sundqutet 1-2, S-E Johansson Department of Community Medkine, Malmo University Hospital, Lund University, 2 Stanford Center for Research In Disease Prevention, Stanford University School of Medicine 3 Department of Welfare and Social Statistics, Statistics, Stockholm, Correspondence: Dr Louise Bayard-Burfield, Lund University, Department of Community Medicine, Malmo University Hospital, S Malmo,, tel , fax from war and oppression in Eastern Europe, Asia, Africa and South America. Recent research into immigrants' health has led to more knowledge of the diversity amongst foreign-born migrants regarding migration experiences, socioeconomic conditions and psychiatric health. In a study based on 938 attempted suicides between 1990 and 1994 by Swedes and foreign-born people living in a defined geographical area, foreign-born people had a higher risk of attempted suicide than Swedish-born people when adjusted for age. Older foreign-born women had the highest attempted suicide rates of all. 7 In studies of the mental health of immigrants in Canada and the UK, foreign-born women generally show worse mental health than foreign-born men. 8 " 10 Young foreign-born people had an increased risk of being hospitalized in a psychiatric department and had shorter hospitalization times compared with Swedes. 11 This could imply that migrants have an unsatisfied need for health care. There are many studies of foreign-born people and their mental health, many of them looking at suicide rates for a specific nationality, 1^ but we have not been able to find any studies of mental health in foreign-born people in a large random sample. The associations between psychiatric illness and socioeconomic factors, such as low educational level, civil

2 EUROPEAN JOURNAL OF PUBLIC HEALTH VOL NO. 1 I status (living alone), poor social network and poor economic resources, are well documented, but surprisingly few studies have investigated these socioeconomic factors as possible confounders in studies examining the association between country of birth and psychiatric illness. For example, the association between social network and morbidity and mortality is strong and consistent, irrespective of the various theoretical models and indices which are used. 20 ' 21 Therefore, social network and civil status are important confounding factors in studies of the association between psychosocial factors and mental health problems. The use of benzodiazepines is often seen in insomnia, anxiety and addiction to drugs, thus indicating difficulties in coping with everyday life. The intake of benzodiazepines could also indicate psychiatric problems. Based on data from an annual survey of living conditions representative of the Swedish population, we wanted to analyse the relation between country of birth (foreignborn versus Swedish-born people), self-reported, long-standing, psychiatric illness and intake of benzodiazepines. We hypothesized that, when compared to Swedes, foreign-born persons, particulary women, have a higher risk of self-reported, long-standing, psychiatric illness and an increased risk of taking benzodiazepines, but that the differences would, to a large extent, be explained by low educational level, living alone, poor social network and poor economic resources. To test these hypotheses, we evaluated the independent relations of educational level, civil status, social network, economic resources and country of birth on self-reported, long-standing, psychiatric illness and taking benzodiazepines. METHODS This study is based on eight simple random samples from the Swedish Survey of Living Conditions (in Swedish Undersokningen om Levnadsfdrh&llanden, ULF). The survey was conducted by Statistics and the data was collected in face-to-face interviews. Approximately 6,000 individuals, in the age range years and living in, are interviewed each year. 36,890 individuals aged years during the years are included in this study. The sample consists of 9312 individuals in the same agespan. The response rate was on average 80%. The sample is used when analysing the prevalence of selfreported, long-standing, psychiatric illness and the sample when analysing the prevalence of taking benzodiazepines (self-reported data). Outcome variables Self-reported, long-standing, psychiatric illness Long-standing illness was defined as suffering from any long-standing illness, after-effects of an injury or any disability or weakness. The variable was dichotomized into respondents who reported long-standing, psychiatric illness including the diagnoses by ICD9 and all others. Taking benzodiazepines Taking benzodiazepines was dichotomized into taking benzodiazepines (regularly or occasionally) or not. Explanatory variables Age Age was analysed in the following groups: 25-34, 35 44, 45-54, and years. The Swedes and foreign-born people were divided into five groups. The classification was partly geographical and partly cultural. Swedish-born people comprised the first group. People born in were in the second group. They could be characterized as migrant labourers. The third group was people born in, i.e. the USA, Canada, Australia, New Zealand, Japan, the Baltic states and Europe (with the exception of, Southern Europe and Eastern Europe). This group included migrant labourers who arrived before 1967, when the Swedish Government enacted a law that put an end to migration of labour. However, certain experts, university students and their families and close relatives and spouses of labour migrants were accepted by the Swedish immigration authorities for settlement after Southern Europeans formed the fourth group. They are migrant workers from rural areas of former Yugoslavia and Italy, Greece, Spain and Portugal. The fifth group (all others) consisted of refugees from war and political and religious oppression, i.e. Eastern Europeans, mostly from Poland, Hungary and Czechoslovakia and refugees from Chile, Uruguay, Argentina, Central America, Asia and Africa. Level of education Level of education was divided into three levels: i) less than 10 years of school completed, ii) between 10 and 12 years of school completed and iii) more than 12 years of school completed. comprised two groups: single or married/cohabiting. was based on the variables casual neighbour interaction, contact frequency with siblings and neighbours, to have acquaintances and a close friend and how often seeing this friend. One point was given for each item with a maximum of six points. If acquiring less than three points the social network was considered poor. Although the items included in the index are given the same weights, they are different phenomena of a social network, the main advantage is the reduction of items in the model. The motive for using the index and for the construction of the social network index was that it concurred with other studies. " Economic resources Economic resources was defined as having access to a car or not. STATISTICAL METHOD The prevalences shown in table I are population estimates, as the samples are drawn from the register of the total

3 Self-reported long-standing psychiatric illness Swedish population. They are to some extent underestimated due to non-participation. The non-participants consist of three groups: refusals (70%), not found (20%) and ill (10%). Among the latter two groups there are probably many disadvantaged persons who might have a psychiatric disorder. Furthermore, all-cause mortality in participants and non-participants was analysed in a proportional hazards model adjusted for sex, age, civil status and geographical region. Those who refused had the same mortality risk as the participants. However, those who were ill or not located had an increased mortality. However, relative measures such as odds ratios will probably be less influenced by non-response than absolute measures such as the prevalence. Statistical models The data were analysed using unconditional logistic regression in the main effect models. 25 The results are shown as odds ratios (s) with 95% confidence intervals (CIs). The fit of a model was judged from the deviance. 25 If the log likelihood ratio approximately equalled the degrees of freedom, the fit was considered good. All final models satisfied this demand. and females were analysed separately due to many interactions between sex and the other independent variables. Tables 2 and 3 show odds ratios for each of the independent variables in an age-adjusted model. The independent variables were chosen in advance due to theoretical grounds and their importance in the model was not tested in terms of improved fit. The reliability of the dependent variables and the majority of the otliers have been analysed by reinterviews (test-retest method) giving kappa coefficients between 0.7 and RESULTS Table I presents the prevalence of self-reported, longstanding, psychiatric illness and intake of benzodiazepines in different age groups and in the explanatory variables of the study sample. A total of 865 persons (2.4%) reported having a long-standing, psychiatric illness, % of the women and 2.0% of the men. South European women and men had just over two and three times higher prevalences of self-reported, long-standing, psychiatric illness respectively when compared to Swedish-born respondents (table 1). The differences in prevalence of self-reported, long-standing, psychiatric illness between migrants from all other countries and Swedes were almost as high as for Table 1 The estimated prevalence of self-reported, long-standing, psychiatric illness and intake of benzodiazepines in different subgroups by sex (percentages) n Uvel Number of cases Totals All Age (years) Psychiatric illness Intake of benzodiazepines 18, , Psychiatric illness Intake of benzodiazepines ,

4 EUROPEAN JOURNAL OF PUBLIC HEALTH VOL NO. 1 the South Europeans. There are also interesting differences in prevalence of intake of benzodiazepines in the different country of birth groups. The age-adjusted s of long-standing, psychiatric illness for country of birth and the other independent variables are demonstrated in table 2. Female labour migrants from and female refugees from all other countries exhibited high risks of self-reported, long-standing, psychiatric illness. Male labour migrants born in or Soutiiern European countries and male refugees bom in all other countries exhibited high risks of longstanding, psychiatric illness (table 2). The association between the migrants' country of birth and the outcome factors was weaker in table 2 (long-standing psychiatric illness) than in table 3 (intake of benzodiazepines). The risk of intake of benzodiazepines was higher in women born in all other countries and in men from (table 3). Women with low educational level showed an increased risk of intake of benzodiazepines. We demonstrate the full models adjusting for all the independent variables in table 4 and 5. Male refugees from all other countries and Southern European labour migrants demonstrated high risks of self-reported, longstanding, psychiatric illness. The earlier increased risk of psychiatric illness became non-significant for female refugees from all other countries, i.e. the high risk remaining could be explained by low educational level or the other social and demographic variables (table 4). Women and men with low educational level, living alone, having a poor social network or having poor economic resources (no car) were associated with an increased risk of self-reported, long-standing, psychiatric illness (table 4). In the full models presented in table 5 female refugees from all other countries demonstrated highrisksof taking benzodiazepines, which could not be explained by low educational level or the other social and demographic variables. In addition, Finnish-born men had high risks of taking benzodiazepines when adjusted for all background variables. Living alone, a poor social network or poor economic resources were all associated with increased risks of taking benzodiazepines in both women and men. Table 2 Estimated s with 95% CI for self-reported, long-standing, psychiatric illness in each of the independent variables, age-adjusted models, by sex Uvel Table 3 Estimated s with 95% Cl for intake of benzodiazepines in each of the independent variables, age-adjusted models, by sex Uvel (95% Cl) ( ) ( ) ( ) (-3.18) ( ) ( ) ( ) ( ) ( ) (OJ2-1.97) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (95% Cl) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (2-5) ( )

5 Self-reported long-standing psychiatric illness DISCUSSION This is one of the first reports to document higher levels of self-reported, long-standing, psychiatric illness and intake of benzodiazepines among foreign-born migrants when compared to Swedish people, in two large representative samples of the Swedish population. In a previous study it was revealed that self-reported, long-standing, psychiatric illness was associated with a higher all-cause mortality. We hypothesized that being foreign-bom and female would be associated with increased risks of self-reported, long-standing, psychiatric illness and intake of benzodiazepines, but that die differences would be largely explained by low educational level, living alone, poor social network and poor economic resources (no car). Our hypotheses were not confirmed. Women and men from Southern European countries and men from Eastern Europe and non- outside of Europe demonstrated an increased risk of selfreported, long-standing, psychiatric illness. In none of these cases could this be explained by the background variables. Further, an increased risk of intake of benzodiazepines was seen in refugee women from Eastern Europe and non-westernized countries outside of Europe. Men born in also showed a higher risk of intake of benzodiazepines which could not be explained by low educational level, living alone, a poor social network or poor economic resources (no car). The immigrants coming from Eastern Europe and non- outside of Europe are likely to have left their home coun- Table 4 Esumated s with 95% Cl for self-reported, long-standing, psychiatric illness by sex Level Age (years) Westerniied countries Table 5 Estimated s with 95% CI for intake of benzodiaiepines by sex Level Age (years) Souriiern European countries % ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (1.18-3) ( ) ( ) (0.82-2) ( ) ( ) ( ) ( ) (U5-4.79) (3-4.04) (1.19-2) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (1.02-4) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (1.12-8) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )

6 EUROPEAN JOURNAL OF PUBLIC HEALTH VOL NO. 1 try because of war or political or religious oppression because does not accept labour immigration from these countries. One would have suspected that this group would be suffering from traumatic experiences and would therefore have a high risk of self-reported, long-standing, psychiatric illness. This study reveals higher age-adjusted risks for both women and men from Eastern Europe and non- outside of Europe. In the final models, both women and men have moderately increased s but the increases were only significant for men. We had expected self-reported, long-standing, psychiatric illness and intake of benzodiazepines to be more common in foreign-born women than men: our reasoning was that women seldom take the initiative to migrate and often remain isolated as housewives for many years and, therefore, have more difficulty in coping with the new language, culture and society- However, as the results of our study suggest, it is perhaps men who have the greatest difficulty in adapting to the new country, perhaps due to losing their role as the head of the family, not being able to find work and having difficulties in raising their children in a traditional way. There is a possibility that refugee women under-report long-standing, psychiatric illness while having a high intake of benzodiazepines. Could this imply that doctors ignore these women's psychiatric problems and prescribe drugs instead of listening to their problems? The finding of a significant risk of self-reported, longstanding, psychiatric illness in female and male migrant labourers from is consistent with the fact that they often have repetitive and strenuous jobs and, therefore, suffer from both physical and psychiatric impairment. Kindlund showed that approximately half of all long-standing, certified sick leave among immigrants in could be explained by their occupation and the remainder by factors in the migration process, the latter being particularly true of married immigrants from. Studies comparing Finns living in with the indigenous population of and, have shown that the former both have more psychiatric symptoms and a higher risk of committing suicide. 29>30 This is consistent with our findings which show that Finnish men have a higher age-adjusted risk of self-reported, long-standing, psychiatric illness and a higher risk of intake of benzodiazepines than Swedish men. Other studies on suicide levels in Canada, the USA, Australia and the UK have also revealed that a worse socioeconomic situation for the immigrant population compared to the indigenous population is an important contributing factor for immigrants' suicide. 31 " 34 In the present study we were able to demonstrate that both migrants' country of birth and socioeconomic factors were independent risk factors of self-reported, long-standing, psychiatric illness. A natural resistance to talking about one's mental health and the use of addictive drugs leads us to believe that the respondents are likely to have under-reported actual use of benzodiazepines and the existence of mental illness. We also have to consider that some of the respondents considered themselves to be mentally ill but would not be diagnosed as such if seen by a psychiatrist. It is also noteworthy that young respondents (aged years) reported a low intake of benzodiazepines and a low frequency of psychiatric illness. One would, on the contrary, have expected a greater frequency of psychiatric illness as both suicide attempts and the debut of psychoses often occur in this age group. It is likely though that people with serious psychiatric illness decide not to participate in this kind of study. There are some limitations to this study. For example, since this data emanates from a cross-sectional study, no definite conclusions can be drawn regarding causality for some of the independent variables. It is not possible to exclude an opposite direction of causality - that persons with a long-standing, psychiatric illness live alone, have a poor social network and poor economic resources because of their illness, rather than long-standing psychiatric illness being a result of these as suggested in the literature. 35 ~ 39 For example, a subsample of the Whitehall II Study, in which an occupational cohort of 4,202 male and female civil servants aged years at baseline was followed prospectively over a 5-year period, found that poor social support measured at baseline was a risk factor for long spells of psychiatric illness. However, we have to consider that the group of foreignborn people is very heterogeneous, with many different nationalities and different reasons for migrating. Further, the use of self-reported data might introduce bias. One example is the view that self-reported, long-standing, psychiatric illness is a subjective measure of mental disease which lacks the reliability of the different psychiatric scales used by psychiatrists. However, in this study, we are interested in the layman's opinion of long-standing, psychiatric illness. Non-participants might introduce a bias, particularly the one-third of non-participants who could never be traced or those who were too ill to participate. These limitations are balanced by the strengths of the Swedish Survey of Living Conditions. For example, the large, well-defined, study population, this being a random sample of the whole Swedish population with a low drop-out and a response rate of 80%. This type of survey has a long tradition. The questions are well validated, have been consistent over the years'^ and have been proven to possess high reliability. 26 In the present study, it was possible to adjust for variables such as age, sex, country of birth, educational level, civil status, social network and economic standards. Another strength of this study was the division of the group of foreign-born people into four smaller groups: migrants from,, and all other nationalities (Eastern Europe and non-westernized countries outside of Europe). CONCLUSIONS The major findings in this study were the high levels of self-reported, long-standing, psychiatric illness and intake of benzodiazepines among foreign-born migrants

7 Self-reported long-standing psychiatric illness compared to Swedes, which could not be explained by low educational level, living alone, poor social network or poor economic resources. Other factors associated with the migration process and the acculturation into a different society could be the possible mechanism behind an increased risk of long-standing, psychiatric illness in foreign-born people. This work was supported by grants to Louise Bayard-Burfield from the Council for Health and Health Care Research, Lund/Malm6 and Bror Gardelius' Memorial Fund and by grants to Jan Sundquist from the Swedish Medical Research Council (grant no K9&-27X C) and the Swedish Society of Medicine. The authors also wish to thank Christopher Burfield for his help with the translation. 1 Sundquist J, Iglesias E, Isacsson A. Migration and hearth: a study of Latin American refugees, their exile in and repatriation. Scand J Primary Health Care 1995; 13: Sundquist J. Migration and hearth: epidemiological studies in Swedish primary health care [dissertation]. Studentlitteratur, Lund: University of Lund, Glower GR. The pattern of psychiatric admissions of Caribbean-born immigrants in London. Soc Psychiatr Psychiatr Epidemiol 1989;24: Johansson LM, Johansson S-E, Sundquist J, Bergman B. Suicide among psychiatric in-patients in Stockholm,. Arch Suicide Res 1996;2: Ferrada-Noli M, Asberg M, Ormstad K. Psychiatric care and transcurtural factors in suicide incidence. Nordic J Psychiatr 1996;50: Serten JP, Sijben N. First admission rates for schizophrenia in immigrants to The Netherlands. Soc Psychiatr Psychiatr Epidemiol 1994;29: Bayard-Burfield L, Sundquist J, Johansson S-E, Traskman-Bendz L. Attempted suicide among Swedish-born people and foreign-bom migrants. Arch Suicide Res 1999;5: Noh S, Speechley M, Kaspar V. Depression in Korean immigrants in Canada. I: Method of the study and prevalence of depression. J Nerv Ment Dis 1992,180: Noh S, Speechley M, Kaspar V. Depression in Korean immigrants in Canada. II: Correlation of gender, work and marriage. J Nerv Ment Dis 1992;180: Furnham A, Shlekh S. Gender, generation and social support: correlates of mental health in Asian immigrants. Int J Soc Psychiatr 1993;39: Johansson LM, Sundquist J, Johansson SE, Bergman B. Immigration, moving house and psychiatric admissions. Acta Psychiatr Scand 1998;98: Burke AW. Attempted suicide among the Irish-born population in Birmingham. Br J Psychiatr 1976; 128: Burke AW. Socio-cultural determinants for attempted suicide among West Indians in Birmingham: ethnic origin and immigrant status. Br J Psychiatr 1976;129: Cochrane R. Psychological and behavioural disturbances in West Indians, Indians and Pakistanis in Britain: a comparison of rates among children and adults. Br J Psychiatr 1979;34: Brun A. SCB Befolkningsstatistik del 3 (Population statistics part 3). Stockholm: Statistics 1995: Monk K. Epidemiology of suicide. Epidemiol Rev 1987,-9: Gunnel DJ, Peters TJ, Kammderling RM, Brooks J. Relation between parasuicide, suicide, psychiatric admissions, and sodoeconomic deprivation. BMJ 1995;311: Cullberg J, Stefansson CG, Wennersten E. Psychiatry in low status dwelling areas. Psychiatr Soc Scl 1981;1: Dalgard O. Bomiljo og psykisk helse (Environment and mental health) [dissertation]. Oslo: University of Oslo, 1980 (In Norwegian with English summary). 20 Hanson BS., social support and health in elderty men: a population-based study [dissertation]. Malmo: University of Lund, Syme SL and its relation to morbidity and mortality. In: Isacsson S-O, Janzon L, editors. Proceedings of the Berzelius symposia Sodal Support and Health In Malmo. Stockholm: Almqvist & Wiksell, 1986: Hanson BS, Ostergren PO. Different social network and social support characteristics, nervous problems and insomnia: theoretical and methodological aspects on some results from the population study 'men born in 1914', Malmo,. Soc Sci Med 1987,25: Berkman LF, Syme SL. s, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol 1979; 109: Ostergren PO. Psychosocial resources and health: with special references to social network, social support and cardiovascular disease [dissertation]. Malmo: University of Lund, Hosmer DW, Lemeshow S. Applied logistic regression. Boston: Wiley Intersdence, Waneryd B. Levnadsforhillanden: Sterintervjustudie i undersokningen om levnadsforhallanden 1989 (ULF) (Living conditions: reviews in ULF 1989). Stockholm. Statistics, Bayard-Burfield L, Sundquist J, Johansson SE. Self-reported long-standing psychiatric illness as a predictor of premature all-cause mortality and violent death: a 14-year follow-up study of native Swedes and foreign-bom migrants. Soc Psychiatr Psychiatr Epidemiol 1998;33: Kindlund H. Fortidspensionering och sjukfranvaro 1990 bland invandrare och svenskar (Early retirement pension and sick-leave among immigrants and Swedes in 1990). Swedish Hearth Author 1995;5: Haavio-Mannila E, Stenius K. Mental hearth of immigrants in [dissertation]. Helsingfors: University of Helsingfors, Johansson L-M, Sundquist J, Johansson S-E, Bergman B, Quist J, Traskman-Bendz L Suicide among foreign-born minorities and native Swedes: an epidemiological follow-up study of a defined population. Soc Sci Med 1997;2: Kliver EV, Ward RH. Convergence of suicide rates to those in the destination country. Am J Epidemiol 1988; 127: Monk M. Epidemiology of suicide. Epidemiologic Reviews 1985,9: Burvill PW, Woodings TL, Stenhouse NS, et al. Suicide during of migrants in Australia. Psychol Med 1982; 12: Raleigh VS, Balarajan R. Suicide levels and trends among immigrants in England and Wales. Health Trends 1992;24: Townsend P, Davidson N. Inequalities in health: the Black Report Harmondsworth: Penguin Books Ltd, Romans SE, Walton VA, McNoe B, Herbison GP, Mullen PE. Otago Women's Hearth Survey 30-month follow-up. I: Onset patterns of non-psychotic psychiatric disorder. Br J Psychiatr 1993; 163: Becker T, Thornicroft G, Leese M, et al. Sodal networks and service use among representative cases of psychosis in south London. BrJ Psychiatr 1997;171: Stansfeld SA, Fuhrer R, Head J, Feme J, Shipley M. Work and psychiatric disorder in the Whitehall II Study. J Psychosom Res 1997,43: Stansfeld SA, Rael EG, Head J, Shipley M, Marmot M. Sodal support and psychiatric sickness absence: a prospective study of British civil servants. Psychol Med 1997;27: Lindstrom H. Non-response errors on sample surveys. Stockholm: Statistics, Received 30 April 1998, accepted 18 January 1999