 |
 |
|
|
|
A not negligible spin-off of large-scale population-based surveys is their capacity to enable the investigation of complex methodological issues with public health and planning implications. For instance, it has been noted repeatedly that, overall, women are more likely than men to use services for mental health reasons (1–6). It has also been noted on a few occasions that the association between sex (see Note) and the use of services decreases when the social anchorage of individuals, as defined by their current social roles, is taken into account (3,7). This change in the level of association between exposure and outcome suggests that the pattern of social anchorage could be a confounder or a modifier of the relation between sex and the use of services. Both alternatives are conceptually plausible, and they could coexist. However, confounding and interaction have differing implications for the interpretation of findings, the design of studies, and the development of mental health services and public health policy. Confounding is a distortion between the “true” and the observed association of exposure and outcome, resulting from the effect of a covariate that is independently associated with both exposure and outcome. If the pattern of social anchorage were a confounder, it would partly account for the association between sex and the use of services for mental health reasons. For example, if nonworkers were more likely than workers to use services, and if women were more likely than men to be nonworkers, the higher probability that women, compared with men, would use services could be partly explained by their higher probability of being nonworkers. Failure to control for the confounding effect of working status would bias the results and in this case lead to an overestimation of the association between sex and the use of services. Convincing evidence for a strong confounding effect of working status on the association between sex and the use of services might contribute to directing research and public health efforts away from the issue of sex, which at the moment is a concern for researchers and decision makers. While confounding is something to be wary of in research design and data analysis, interaction is a finding to be further investigated. For example, a significant interaction between sex and working status could show that, although women are overall more likely than men to use services for mental health reasons, this inequality differs in workers and nonworkers. Failure to take into account the interaction effect of working status would conceal findings that have the potential to improve the development of comprehensive mental health services and policy. This study aimed to investigate whether the pattern of social anchorage confounds or modifies the association between sex and the use of services for mental health reasons. MethodsThis report is based on a case–control study nested in the CCHS 1.2. We defined cases as users of general medical services for mental health reasons in the year preceding the interview and control subjects as never-users (that is, never in their lifetime) of any services for mental health reasons. Following Regier and others (8), we defined general medical services as general practitioners and clinicians other than psychiatrists. We restricted analyses to respondents aged 18 to 65 years, because some of the social roles under study are less accessible to adolescents (that is, those aged 15 to 17 years) and to seniors (that is, those aged 66 years and over). Indicators of the pattern of social anchorage include the roles of spouse, parent, and worker, as well as the combination and accumulation of these social roles. ORs were adjusted for the need for services as defined by 3 variables: 1) the presence or absence, in the previous 12 months, of one or more psychiatric diagnosis assessed in the CCHS 1.2 (that is, agoraphobia, major depression, mania, panic attack, social phobia, and addiction to alcohol or drugs); 2) the presence or absence of chronic mental illness; and 3) the index of emotional distress assessed according to the K10 (see www.abs.gov.au/Ausstats/abs@.nsf/0/b90fb790f3cb1fe3ca256d120004c6a6?OpenDocument). All analyses were based on weighted data. The systematic assessment of confounding and interaction is a complex exercise that is debated among epidemiologists (9–13). In this study, we applied several statistical analysis techniques borrowed from various approaches. Basic statistics, such as the probability of using general services for mental health reasons and ORs for sex, were estimated for each stratum of social anchorage to provide some insights into the likelihood of confounding and interaction. Large variations in these statistics across the categories of the covariate are an indication of interaction, whereas a global OR falling outside the range of strata-specific ORs is an indication of confounding. We explored 2 approaches to detecting confounding (14). Following the statistical approach based on the concept of collapsibility, we compared ORs estimated with and without the indicator of social anchorage. Confounding is suspected when these ORs differ by more than a set percentage, usually 10% or more (15). Following the epidemiologic approach, which views confounding as the noncomparability of the groups under study, we compared the distribution of the indicators of social anchorage among women and men belonging to the control group. Finally, an interaction term between sex and each indicator of social anchorage was added in turn and tested for significance in a logistic regression model that already contained sex, the indicator of social anchorage under study, and the 3 variables assessing the need for services. Interaction is present if the slopes for the probability of the use of services for women and men are not parallel—the most extreme case occurring when the slopes cross each other. ResultsOverall, the probability of using services for mental health reasons is higher among individuals who lack specific types of social anchors, such as individuals who are parents but not spouses (that is, single parents), individuals who are neither spouses nor workers, and individuals who hold none of the social roles under study (Table 1). When we controlled for the need for mental health services, we found that Canadian women aged between 18 and 65 years are almost 3 times (OR 2.9) more likely than Canadian men of the same age group to have used general services for mental health reasons in the previous 12 months. Is this global OR confounded or modified by the pattern of social anchorage?
The evidence supporting the confounding effect of social anchorage is not convincing. Not surprisingly, given the very large sample size, all indicators of social anchorage (except the role of parent considered alone) were associated with the use of general services, and their distribution showed significant, albeit mostly small, differences across sex. This suggests that all indicators of social anchorage but one (that is, the role of parent considered alone) are potential confounders, since they are independently associated with exposure and outcome. However, the global OR of 2.9 falls within the range of strata-specific ORs for each indicator of the pattern of social anchorage, which suggests that confounding is unlikely (Table 1). For example, the global OR falls between the OR of 2.8 observed in spouses and the OR of 3.0 observed in nonspouses. Further, the decrease in the OR for sex is small (ranging from The evidence supporting the interaction effect of the pattern of social anchorage is substantial. The strata-specific ORs show important variation within some indicators of social anchorage (Table 1). For instance, when the roles of spouse and parent are considered simultaneously, the ORs for sex in individuals who are neither spouses nor parents, spouses but not parents, or both spouses and parents indicate that women are 2.5 to 3.1 times more likely than men to use general medical services for mental health reasons. However, this inequality between women and men subsides in individuals who are single parents (OR 1.4; 95%CI, 0.7 to 2.6). In this group, men are as likely as women to use services. Professional anchorage, described by the role of worker, appears to be a strong modifier of the association between sex and the use of general medical services for mental health reasons: the strata-specific ORs show a large variation, whether the role of worker is considered alone or whether it is considered in combination with the roles of spouse or parent. Overall, inequality between women and men is larger in workers (OR 3.1) than in nonworkers (OR 1.6), mostly because male nonworkers are roughly 3 times more likely than male workers to use these services (11.6% and 4.0%, respectively; Table 1). More specifically, women and men who are neither spouses nor workers (OR 1.4), or who are parents but not workers (OR 1.2), are as likely to use general services for mental health reasons. A large variation in strata-specific ORs is also noticeable when social anchorage is described by the accumulation of social roles. Finally, the interaction term tested in logistic regression is statistically significant for 3 indicators of the pattern of social anchorage: the role of worker, the combined roles of worker and spouse, and the combined roles of worker and parent. Thus the probability of using services is similar in male and female nonworkers (roughly 12%), whereas it is much lower in male workers (4.0%), compared with female workers (10.6%). ConclusionsFindings from this case–control study nested in the CCHS 1.2 suggest that the pattern of social anchorage, especially the role of worker considered alone or in combination with the role of spouse or parent, tends to modify the association between sex and the use of general services for mental health reasons in the Canadian population aged 18 to 65 years. The findings also suggest that no indicator of social anchorage under study confounds this association. In the presence of interaction, controlling for the covariate is not sufficient and will produce misleading results. An interaction term involving the covariate and the exposure must be included in the model, or separate analyses should be conducted. These findings stress the need to take the pattern of social anchorage of individuals into account in future research, but they can also be expressed in terms more meaningful for public health policy. From a public health perspective, this study produces 3 main findings: 1) women and men who are less anchored to Canadian society are more likely to use general medical services for mental health reasons; 2) inequality between women and men in the use of general medical services for mental health reasons is stronger among individuals who are anchored to the society in which they live; 3) among all anchors under study, professional anchorage, whether alone or in combination with other social roles, seems to be a strong marker of inequality between women and men in the use of general medical services for mental health reasons. This study has some limitations. First, general medical services are the health services most often used by individuals consulting for mental health reasons, and it remains to be verified whether the pattern of social anchorage also modifies the association between sex and the use of specialized and alternative mental health services. Second, analyses were restricted to Canadians aged 18 to 65 years, and confounding and interaction by social anchorage may act differently in the younger (that is, those aged 15 to 17 years) or older (that is, those age 65 years and over) population. Finally, control for the need for mental health services was incomplete, since it did not include a measure of disability. Nevertheless, discovering that the pattern of social anchorage modifies the association between sex and the use of general medical services for mental health reasons in the adult population is instructive for both future research and public health. In particular, the finding that the probability of using these services is much lower in male workers (4%) than in female workers (10.6%) further supports the hypothesis that paid work may have different health implications for women and men in terms of exposure to risk and protective factors and in terms of responses to this exposure (16). Funding and SupportThis study was conducted while Dr Drapeau was supported by postdoctoral fellowships from the CIHR and the training program Research in Addictions and Mental Health Policy and Services. NoteThe authors had originally used the term “gender” rather than “sex” throughout this article, arguing that in this paper they refer to women and men as social beings. Since there is no consensus on the distinction between gender and sex, the authors have agreed that gender be edited to sex to reflect CJP style, wherein the term sex is used to signify the objective biological fact of being male or female and gender refers to culturally defined characteristics or traits of maleness and femaleness (often, but not entirely, determined by the sex of individuals). AcknowledgementsAccess to the CCHS-1.2 data was granted by the Social Science and Humanities Research Council and Statistics Canada. Analyses were carried out at the Centre inter-universitaire québécois de statistiques sociales. The research and analyses are based on data produced by Statistics Canada and the opinions expressed do not represent the views of Statistics Canada. References1. Andrews G, Issakidis C, Carter G. Shortfall in mental health service utilisation. Br J Psychiatry 2001;179:417–25. 2. Lefebvre J, Lesage A, Cyr M, Toupin J, Fournier L. Factors related to utilization of services for mental health reasons in Montreal, Canada. Soc Psychiatry Psychiatr Epidemiol 1998;33:291–8. 3. Parslow RA, Jorm AF. Who uses mental health services in Australia? An analysis of data from the National Survey of Mental Health and Wellbeing. Aust N Z J Psychiatry 2000;34:997–1008. 4. Rhodes AE, Goering PN, To T, Williams JI. Gender and outpatient mental health service use. Soc Sci Med 2002;54:1–10. 5. Stuart H. Access to physician treatment for a mental disorder: a regional analysis. Soc Psychiatry Psychiatr Epidemiol 2000;35:61–70. 6. Wang J, Patten SB, Russel ML. Alternative medicine use by individuals with major depression. Can J Psychiatry 2001;46:528–33. 7. ten Have M, Vollebergh W, Bijl RV, de Graaf R. Predictors of incident care service utilisation for mental health problems in the Dutch general population. Soc Psychiatry Psychiatr Epidemiol 2001;36:141–9. 8. Regier DA, Narrow WE, Rae DS, Mandersheid RW, Locke BZ, Goodwin FK. The de facto US mental and addictive disorders services system. Arch Gen Psychiatry 1993;50:85–94. 9. Boivin J-F, Wacholder S. Conditions for confounding of the risk ratio and of the odds ratio. Am J Epidemiol 1985;121:152–8. 10. Greenland S. Absence of confounding does not correspond to collapsibility of the rate ratio or rate difference. Epidemiology 1996;7:498–501. 11. Kupper LL, Hogan MD. Interaction in epidemiologic studies. Am J Epidemiol 1978;108:447–53. 12. Miettinen OS, Cook EF. Confounding: essence and detection. Am J Epidemiol 1981;114:593–603. 13. Salas M, Hofman A, Stricker BH. Confounding by indication: an example of variation in the use of epidemiologic terminology. Am J Epidemiol 1999;149:981–3. 14. Greenland S, Robins JM. Identifiability, exchangeability, and epidemiological confounding. Int J Epidemiol 1986;15:412–8. 15. Sonis J. A closer look at confounding. Fam Med 1998;30:584–8. 16. Messing K, Punnett L, Bond M, Alexanderson K, Pyle J, Zahm S, and others. Be the fairest of them all: challenges and recommendations for the treatment of gender in occupational health research. Am J Ind Med 2003;43:618–29. Author(s)Manuscript received and accepted May 2005. 1. Postdoctoral Fellow, Unité de psychiatrie sociale–Centre de recherche Fernand-Seguin, Montreal, Quebec, and at the Département de psychiatrie, Université de Montréal, Montreal, Quebec. 2. Researcher, Unité de psychiatrie sociale–Centre de recherche Fernand-Seguin, Montreal, Quebec; Professor, Département de psychiatrie, Université de Montréal, Montreal, Quebec. 3. Researcher, Unité de psychiatrie sociale–Centre de recherche Fernand-Seguin, Montreal, Quebec, and at the Département de psychiatrie, Université de Montréal, Montreal, Quebec. Address for correspondence: Dr A Drapeau, Unité de psychiatrie sociale–Centre de recherche Fernand-Seguin, Pavillon 218-Bédard, 7331 Hochelaga, Montreal QC H1N 3V2 e-mail: adrapeau.hlhl@ssss.gouv.qc.ca
1 | 2
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||