Abstract
Importance and Objective:
Previous studies have linked suicide risk to religious participation, but the vast majority have used ecologic, cross-sectional or case-control data. We use longitudinal cohort data to examine the relationship between service attendance and suicide, and the joint associations of suicide with service attendance and religious affiliation.
Design, Setting, Participants, Main Outcome and Measure:
We evaluated associations between service attendance and suicide in a large long-term prospective cohort, the Nurses’ Health Study, in an analysis that included 89,708 women. Religious service attendance was self-reported in 1992 and 1996. We used proportional hazard models to examine the association between service attendance and suicide with follow-up from 1996 until 2010, adjusting for demographic covariates, lifestyle factors and medical history, depressive symptoms, and social integration measures. We performed sensitivity analyses to examine the influence of unmeasured confounding.
Results:
Attending religious services once per week or more was associated with an approximately five-fold lower rate of suicide, compared with never attending services (Hazard Ratio [HR]=0.16, 95% confidence interval: 0.06–0.46). Service attendance once or more per week, versus less often, was associated with a HR=0.05 (0.006–0.45) for Catholics, but only HR=0.34 (0.10–1.10) for Protestants (p-value for heterogeneity=0.05). Results were robust in sensitivity analysis and to exclusions of persons who were previously depressed or had a history of cancer or cardiovascular disease. There was evidence that social integration, depressive symptoms, and alcohol consumption partially mediated the relationship among those occasionally attending services.
Conclusions and Relevance:
In this cohort of US women, frequent religious service attendance was associated with a significantly lower rate of suicide.
Keywords: Religious service attendance, suicide, mental health
Introduction
Suicide is one of the ten leading causes of death in the United States; (1) it is the fourth leading cause of death for persons aged 18–65; and the risk increases later in life. (2–4) Despite improvements in mental health, the suicide rate in the US (12 per 100,000) is about same as it was over one hundred years ago.(1–3) The major world religions have strong traditions prohibiting suicide, and various forms of religious participation have been thought to be linked with suicide risk.(4, 5) Empirical research on the topic was made prominent by the sociologist Emile Durkheim’s 1897 work Suicide.(6) Durkheim noted that, within Europe, suicide rates were higher in Protestant regions than Catholic regions and attributed the lower suicide rates among Catholics to greater religious integration and less individual autonomy in beliefs.
Although there have been numerous studies of the relationship between various forms of religious participation and suicide(2, 4–12), most of these have serious methodological limitations. Many of the studies, including Durkheim’s, have used ecologic data and cannot control for individual level confounding;(13) many other studies employ survey data but, as a result, can only examine suicide ideation or suicide attempts(10), rather than suicide itself. The relationship between service attendance and suicide and ideation has also been studied using case-control designs.(8) Research with cohort designs is difficult due to the relatively low baseline rate of suicide. To the best of our knowledge, only one study to date has used cohort data to examine the relationship between service attendance and suicide itself,(14) but was not able to control for depressive symptoms which is related to both suicide and service attendance.(2, 4, 15, 16)
We propose to address this question using the Nurses’ Health Study, a large prospective cohort study among U.S. women with repeated measurements of religious service attendance, and detailed information on dietary, lifestyle, social, psychological, and medical risk factors. We also examine the joint associations of suicide with both service attendance and religious affiliation.
Methods
Study design
The Nurses’ Health study (NHS) began in 1976 and included 121,700 nurses aged 30–55 years from across the United States.(17) Information on lifestyle and medical history was collected using a questionnaire at baseline and every two years subsequently. In the current study, follow-up for suicide began with the religious service attendance measure in 1996 and continued until suicide, loss of follow-up, or the end of follow-up in June 2010. Participants were excluded who died before 1996 or did not reply to the 1996 questionnaire (n=27,122), or who had 1996 religious service attendance missing (n=7,246). Our study includes 89,708 participants with 1,528,538 person years. We identified 36 total suicides during follow up. The study protocol for the Nurses’ Health Study was approved by the institutional review boards of Brigham and Women’s Hospital and Harvard School of Public Health. Study participants provided written informed consent.
Religious service attendance was self-reported in 1992 and 1996. The questionnaire asked, “How often do you go to religious meetings or services?” Response categories included “more than once a week, once a week, one to three times per month, less than once per month, never or almost never.” The primary exposure variable was attendance 1996, with attendance in 1992 controlled for as a covariate so as to examine the associations between incident attendance and risk of suicide.
Deaths were identified by using state mortality files and the U.S. National Death Index and by reports from next of kin. We defined suicide deaths using the World Health Organization International Classification of Diseases, Eighth Revision (ICD-8), “Suicide and Self-Inflicted Injury” cluster (E950 through E959) definition, including: suicide and self-inflicted poisoning by solid or liquid substances (E950), by gases in domestic use (E951), or by other gases and vapors (E952); suicide and self-inflicted injury by hanging, strangulation and suffocation (E953), by submersion (E954), by firearms and explosives (E955), by cutting and piercing instruments (E956), by jumping (E957), or by other and unspecified means (E958); and late effects of self-inflicted injury (E959).
We selected covariates that were important predictors for both general health and suicide, including: age (continuous), employment status (part-time, full-time, unemployed or retired), family history of alcoholism (yes/no), body mass index (<20,20–24.9,25–29.9,30–34.9,≥35 kg/m2), physical activity (quintiles); caffeine intake (g/day), alcohol intake (g/day), smoking status (never, former, current in categories of 1–14, 15–24, and ≥25 cigarettes/day), depressive symptoms (yes/no), history of diabetes, hypertension, cancer or hypercholesterolemia (yes/no), family income (quintiles), live alone (yes/no), geographic region (north, south, middle), religious service attendance in 1992 (never, less than once/week, once or more/week), and the Berkman-Syme social integration score(18). Social integration was derived using the following six components (excluding service attendance component): Marital status, other group participation, number of close friends, number of close relatives, number of close friends see at least once per month, number of close relatives seen at least once per month. Indicator variables were used for any missing covariate information for categorical variables and median imputation was used for missing continuous covariates. Covariate measurements prior to the religious attendance exposure were taken as confounders and the first available measurements subsequent to religious attendance exposures were taken as mediators. For mediators, we considered depressive symptoms in 2000 measured using the Center for Epidemiologic Studies Depression-10,(19) alcohol consumption in 1998, and social integration in 2000.
Statistical analyses
We examined the association of religious service attendance with suicide using multivariate Cox proportional hazard models. We calculated hazard ratios and their 95% confidence intervals comparing frequency of religious services attendance (once or more per week, or less than once per week) vs. never attending. To accommodate the relatively small number of events, as a sensitivity analysis we also used “exact logistic regression” which was developed to address small numbers of events (20), and made adjustment for depressive symptoms and social support. We further stratified the analysis by religious affiliation (Catholic vs. Protestant).
We applied mediation analysis methods(21, 22) to examine proportions of the association between religious service attendance in 1996 and subsequent suicide that were mediated by depressive symptoms, alcohol intake, and social integration. Methods for mediation assume baseline covariates suffice to control for exposure-outcome, mediator-outcome, and exposure-mediator confounding.
We conducted several sensitivity analyses to assess the robustness of the results. To minimize the influence of reverse causation, we additionally performed further analyses excluding women who were depressed in 1996 (score of ≤52 on the five-item Mental Health Inventory, self-reported depressive symptoms, or use of antidepressant), and also women who in 1996 had a history of cancer (except non-melanoma skin cancer) or a cardiovascular condition (myocardial infarction, coronary artery bypass graft surgery, percutaneous transluminal coronary angioplasty, or stroke). We used sensitivity analysis to assess how substantial unmeasured confounding would need to be to explain away the observed associations.(23, 24) All statistical analyses were performed under SAS Unix system.
Results
Among 89,708 women at 1996 baseline, 17,028 women attended more than once per week, 36,488 attended once per week, 14,548 attended less than once per week, and 21,644 never attended (Table 1). The majority of our study participants were Catholic or Protestant. Women who attended religious services more frequently tended to have fewer depressive symptoms (Table 1). They were less likely to be current smokers and more likely to be married. There were 36 suicide events during follow-up, with suicide incidence declining with increasing service attendance (Figure 1, Supplementary Table 1).
Table 1.
Age-adjusted baseline characteristics of 89708 participants of the Nurses’ Health Study by religious service attendance in 1996.
| Religious service attendance in 1996 | ||||
|---|---|---|---|---|
| Never (n=21644) | < Once/week (n=14548) | Once/week (n=36488) | >Once/week (n=17028) | |
| Age in years, mean (SD) | 61.5(7.1) | 61.2(7.3) | 62.4(7.1) | 63.5(6.9) |
| Race/ethnic white, (%) | 97.9 | 97.0 | 97.7 | 96.9 |
| Employment status, (%) | ||||
| Not employed outside the home/Retired | 26.5 | 24.0 | 26.4 | 29.1 |
| Employed part-time | 15.2 | 18.9 | 20.3 | 21.1 |
| Employed full-time | 42.2 | 40.7 | 38.0 | 34.8 |
| Missing | 16.1 | 16.4 | 15.2 | 14.9 |
| Family history of alcoholism, (%) | 21.2 | 18.7 | 18.5 | 19.0 |
| Body mass index, kg/m2 | 26.4(5.5) | 26.6(5.2) | 26.4(5.0) | 26.5(5.1) |
| Physical activity, METs/wka | 19.8(24.5) | 20.4(24.8) | 19.5(24.4) | 19.6(24.2) |
| Alcohol intake, g/day | 6.7(11.2) | 5.3(9.0) | 4.6(8.3) | 3.4(7.3) |
| Caffeine intake, g/day | 266.3(214.5) | 249.6(204.8) | 237.4(204.6) | 216.2(199.7) |
| Smoking status, n (%) | ||||
| Never | 33.6 | 40.7 | 46.6 | 57.6 |
| Former | 45.6 | 44.6 | 42.6 | 36.6 |
| Current, 1–14 cigarettes/day | 7.2 | 6.3 | 5.1 | 2.9 |
| Current, 15–24 cigarettes/day | 8.7 | 6.1 | 4.1 | 2.2 |
| Current, 25 or more cigarettes/day | 4.8 | 2.4 | 1.5 | 0.7 |
| History of hypertension, (%) | 35.0 | 34.5 | 33.3 | 33.3 |
| History of diabetes, (%) | 5.7 | 5.5 | 5.3 | 5.1 |
| History of hypercholesterolemia, (%) | 44.5 | 44.9 | 45.4 | 45.3 |
| Antidepressant use, (%) | 8.5 | 7.8 | 6.4 | 6.6 |
| Religious service attendance > once per week in 1992, (%) | 2.3 | 9.3 | 54.3 | 22.7 |
| Married, (%) | 78.4 | 80.2 | 84.6 | 84.0 |
| Close friends you see 1/month | 2.7(1.0) | 2.9(1.0) | 3.0(1.0) | 3.2(1.1) |
| How many close friends | 3.2(1.2) | 3.4(1.1) | 3.5(1.2) | 3.6(1.2) |
| Close relatives you see 1/month | 2.1(1.0) | 2.4(1.1) | 2.5(1.1) | 2.5(1.2) |
| How many close relatives | 2.9(1.3) | 3.2(1.3) | 3.3(1.3) | 3.4(1.3) |
| Social integration index without religious service attendance componentb | 4.7(2.7) | 5.4(2.7) | 5.8(2.7) | 6.3(2.8) |
| Catholic,% | 29.4 | 28.3 | 53.7 | 41.1 |
| Protestant, % | 61.7 | 61.4 | 42.6 | 51.6 |
| Ashkenazi Jewish, % | 3.8 | 6.5 | 0.6 | 0.3 |
| Eastern (Buddhist, Hindu), % | 0.3 | 0.3 | 0.0 | 0.1 |
| Muslim, % | 0.0 | 0.0 | 0.0 | 0.0 |
| Other religious heritage, % | 2.4 | 1.3 | 1.0 | 1.1 |
| Live alone, % | 11.6 | 11.0 | 9.1 | 9.9 |
| Geographic region, % | ||||
| North | 36.1 | 35.9 | 35.4 | 32.2 |
| South | 13.1 | 10.9 | 9.4 | 13.5 |
| Middle | 39.1 | 41.0 | 43.5 | 42.7 |
| Other | 11.7 | 12.2 | 11.7 | 11.6 |
| Family income (dollars/year) | 67510.5 (27282.5) | 67124.0 (27210.2) | 63068.7 (23926.4) | 61307.5 (23963.9) |
Values are means (SD) or percentages and are standardized to the age distribution of the study population.
METs/wk = metabolic equivalent of tasks per week.
Construction of the social integration index is described in Am J Epidemiol 1979;109:186–204.
Figure.
Suicide Incidence From 1996 to 2010 by Frequency of Religious Service Attendance in 1996
Using a multivariate Cox proportional hazard model, compared with women who never attended religious services, women who attended once per week or more in 1996 had an approximately five-fold lower rate of suicide with multivariate adjusted hazard ratio of HR=0.16 (95% CI: 0.06–0.45) (Table 2). When further adjustment was made for the social integration score (other than service attendance) the hazard ratio was essentially unchanged (HR=0.16; 95%CI: 0.06–0.46). Results were also similar after excluding those who were depressed in 1996 as defined by a score of ≤52 on the five-item Mental Health Inventory, self-reported depressive symptoms, or use of antidepressants, or after excluding those who had a history of cancer or of a cardiovascular condition in 1996 (Supplementary Table 2). The odds ratio from exact logistic regression to address the small number of suicide events was likewise similar, OR=0.17 (95% CI: 0.06–0.44). Results were also similar across strata of attendance in 1992 (Supplementary Table 3).
Table 2.
Multivariate adjusted Hazard Ratio and its related 95% Confidence Interval for suicide events (1996–2010) by frequency of religious service attendance in 1996.
| Religious service attendance in 1996 | |||
|---|---|---|---|
| Never | Less than once /week | Once or more /week | |
| Total suicide events (1996–2010, total n=36) | 18 | 11 | 7 |
| Adjusted for age | Reference | 0.91 (0.43–1.94) | 0.15 (0.06–0.37) |
| Adjusted for age and lifestyle demographic variables, and religious service attendance in 1992a | Reference | 0.88 (0.38–2.07) | 0.16 (0.06–0.45) |
| Further adjusted for other aspects of social integration in 1992b | Reference | 0.85 (0.36–2.00) | 0.16 (0.06–0.46) |
Adjusted for age; employment status (not employed outside the home or retired, employed part-time outside the home, employed full-time outside the home); family history of alcoholism (yes/no); body mass index (<20, 20–24.9, 25–29.9, 30–34.9, ≥35 kg/m2); physical activity in metabolic equivalent of tasks per week (quintiles); caffeine intake (g/day); alcohol intake (g/day); smoking status (never, former, and current in categories of 1–14, 15–24, and ≥25 cigarettes per day); depressive symptoms (yes/no), history of diabetes, hypertension, cancer or hypercholesterolemia (yes/no), family income (quintiles, dollars/year), live alone (yes/no), geographic region (north, south, middle), and religious service attendance in 1992 (never, less than once/week, once or more/week).
Further adjusted for other aspects of social integration in 1992, including marital status(not married, married), relative and friends score, relatives and friends frequency index, and social group participation (none, any number of hours).Social integration score, including score and frequency of relatives and friends were derived based on the definition from Am J Epidemiol 1979;109:186–204. For adjustment, we derived social integration score without religious service attendance components.
Even after contemporaneous adjustment in 1996 for all other individual components of the social integration score, the association between service attendance, comparing once or more per week versus less often, was similar and substantial (HR=0.17; 95%CI: 0.07–0.47), and was larger than the effect size for a one standard deviation change in social integration score, either with service attendance excluded (HR=0.82; 95%CI: 0.52–1.28) or with it included (HR=0.67; 95%CI: 0.41–1.10). When using a median split dichotomization, the effect size for service attendance was also larger than that of any other single components of the social integration score (Supplementary Table 4). The effect size for service attendance was similar to or larger in magnitude than that of every other covariate in the adjusted model (Supplementary Table 5).
For an unmeasured confounder to explain away the hazard ratio estimate of HR=0.16 (95%CI: 0.06–0.46), the unmeasured confounder would have to both increase the likelihood of service attendance and decrease the likelihood of suicide by 10.5-fold, above and beyond the measured confounders; weaker confounding would not suffice. To bring the estimate’s upper confidence limit of 0.46 above one, the unmeasured confounder would still have to both increase the likelihood of service attendance and decrease the likelihood of suicide by 3.3-fold, above and beyond the measured confounders.
There was evidence that the association between attendance and suicide differed by Catholic versus Protestant denomination: the service attendance hazard ratio for Catholics for attending once per week or more, versus less often, was HR=0.05 (0.006–0.48) which was approximately seven-fold smaller (p-value for heterogeneity = 0.05) than the analogous hazard for Protestants, HR=0.34 (0.10–1.10) (Table 3).
Table 3.
Joint associations between religious service attendance in 1996 and Protestant versus Catholic identity.
| Religious affiliation | ||||
|---|---|---|---|---|
| Frequency of religious service attendance | Protestant | Catholic | ||
| event/Years | HR (95% CI) | event/Years | HR (95% CI) | |
| Never or < Once/week | 14/260486 | 1.0 (Reference) | 8/129319 | 0.97 (0.38–2.46) |
| ≥ once/week | 5/292735 | 0.34 (0.10–1.10) | 1/322692 | 0.05 (0.006–0.45) |
| HRs (95%CI) for service attendance within strata of religious affiliation | 0.34 (0.10–1.10) | 0.05 (0.006–0.48) | ||
Model adjusted for age, employment status (not employed outside the home or retired, employed part-time outside the home, employed full-time outside the home); family history of alcoholism (yes/no); body mass index (<20, 20–24.9, 25–29.9, 30–34.9, ≥35 kg/m2); physical activity in metabolic equivalent of tasks per week (quintiles); caffeine intake (g/day); alcohol intake (g/day); smoking status (never, former, and current in categories of 1–14, 15–24, and ≥25 cigarettes per day); depressive symptoms (yes/no), family income (quintiles, dollars/year), live alone (yes/no), geographic region (north, south, middle), and history of diabetes, hypertension, cancer or hypercholesterolemia (yes/no), religious service attendance in 1992 (never, less than once/week, once or more/week), other aspects of social integration in 1992,including marital status(not married, married), relative and friends score, relatives and friends frequency index, and social group participation (none, any number of hours).Social integration score, including score and frequency of relatives and friends were derived based on the definition from Am J Epidemiol 1979;109:186–204. For adjustment, we derived social integration score without religious service attendance components.
P for multiplicative interaction=0.05
RERI= −0.26 (−1.33, 0.81), p for additive interaction=0.63
We used mediation analysis to investigate whether depressive symptoms in 2000, alcohol consumption in 1998, and subsequent social integration in 2000 mediated the association between religious service attendance and suicide. No single mediator seemed to explain much of the association (Table 4). When adjustment was made for depressive symptoms, alcohol consumption and social integration simultaneously, the hazard ratio for those attending less than once per week versus not at all changed from 0.85 (0.36–2.00) to 0.94 (0.40–2.23), accounting for about half of the association; however, for those attending more than once per week versus not at all, the hazard ratio remained essentially unchanged with HR=0.16 (0.06–0.46, Table 4).
Table 4.
Adjustment for potential mediators for the association between religious service attendance in 1996 and suicide.
| Religious service attendance in 1996 | |||
|---|---|---|---|
| Never | Less than once /week | Once or more /week | |
| Multivariate adjusted model from Table 3 model c | Reference | 0.85 (0.36–2.00) | 0.16 (0.06–0.46) |
| Further adjusted for mediators: | |||
| + Social Integration score in 2000 | Reference | 0.85 (0.36–2.02) | 0.16 (0.06–0.46) |
| + Alcohol consumption in 1998 | Reference | 0.86 (0.36–2.05) | 0.16 (0.06–0.45) |
| + Depressive symptoms or antidepressant use in 2000 | Reference | 0.92 (0.39–2.17) | 0.17 (0.06–0.48) |
| + Alcohol consumption in 1998, social Integration score in 2000,and depressive symptoms or antidepressant use in 2000 | Reference | 0.94 (0.40–2.23) | 0.16 (0.06–0.46) |
Model adjusted for age, employment status (not employed outside the home or retired, employed part-time outside the home, employed full-time outside the home); family history of alcoholism (yes/no); body mass index (<20, 20–24.9, 25–29.9, 30–34.9, ≥35 kg/m2); physical activity in metabolic equivalent of tasks per week (quintiles); caffeine intake (g/day); alcohol intake (g/day); smoking status (never, former, and current in categories of 1–14, 15–24, and ≥25 cigarettes per day); depressive symptoms (yes/no), family income (quintiles, dollars/year), live alone (yes/no), geographic region (north, south, middle), and history of diabetes, hypertension, cancer or hypercholesterolemia (yes/no), religious service attendance in 1992 (never, less than once/week, once or more/week), other aspects of social integration in 1992,including marital status(not married, married), relative and friends score, relatives and friends frequency index, and social group participation (none, any number of hours).Social integration score, including score and frequency of relatives and friends were derived based on the definition from Am J Epidemiol 1979;109:186–204. For adjustment, we derived social integration score without religious service attendance components.
Mediator alcohol consumption in 1998 was modeled as >0.9 g/day vs. ≤ 0.9/day. Median level=0.9 g/day
Mediator depressive symptoms in 2000, antidepressant use was modeled as yes vs. no.
Mediator social integration score in 2000 was modeled as >3 vs. ≤3. Median level=3.
Discussion
In this large prospective cohort of 89,708 U.S. nurses with 1,528,538 person years of follow up, we found a substantial inverse association between frequent religious service attendance and risk of suicide. Compared with women who had never attended religious services, women who attended once or more per week had a five-fold lower risk of suicide; results were robust across various exclusions, methods of analysis, and in sensitivity analysis.
Our results are consistent with other literature suggesting an inverse association between religious participation and suicide.(7–14) In contrast to other studies, however, we used longitudinal cohort data, rather than cross-sectional or ecologic data; we made extensive adjustment for confounding; we were able to examine incident rather than prevalent service attendance by controlling for past attendance; and we were able to examine suicide itself rather than suicide ideation or attempts. We were also able to control for depressive symptoms, which is related to both suicide and service attendance.(15, 16) Although other aspects of social integration are also associated with reduced suicide,(25, 26) when we compared the effect size of religious attendance with social integration components, it was religious service attendance itself that seemed most prominent among these associations. Subsequent social support seemed to mediate little of the association between attendance and suicide.
In examining the potential pathways from service attendance to suicide, depressive symptoms, social integration, and alcohol consumption seem to explain some of the association for those attending less than once per week, but little of the association for those attending once per week or more. The mediation estimates were limited by the potential for suicide to occur before the mediators were measured and by the use of mediators measured at a single point in time. Nevertheless, the evidence seems to indicate an association independent of these pathways. It is likely that service attendance is itself associated with the belief that suicide is wrong and that this belief may be the primary pathway. This hypothesis has been proposed previously with some preliminary empirical evidence for it.(12, 27) Religious reasons sometimes given for prohibitions against suicide include: life being a gift from God, suicide being against the natural order, suicide causing injury to the community, suicide encouraging others to follow similar course, and death being this life’s greatest evil.(28, 29) That most of the association in this analysis was not explained by depressive symptoms, alcohol consumption, or social integration is consistent with moral beliefs being the dominant pathway. Future research could also examine other potential mediators such as hopelessness, or meaning and purpose in life.
Our results suggested that the inverse association of service attendance with suicide was stronger for Catholics than Protestants. The result is of historic and sociological significance since, in Durkheim’s 1897 work Suicide, a book that both strongly influenced research on religion and health and shaped the discipline of sociology itself, Durkheim had noted that within Europe suicide rates were higher in Protestant regions than Catholic.(6) Durkheim’s analysis, and many since, used ecologic data and has thus been criticized on these grounds.(4, 7, 30) While Durkheim’s analysis was replicated using individual level data in Switzerland with similar results,(11) our analyses here suggest that the lower suicide rates among Catholics versus Protestants may in fact apply only to those regularly attending services. Concerning those regularly attending services, the inverse association between service attendance and suicide was stronger for Catholics than Protestants.
Although the Nurses’ Health Study was not targeted to a particular religious group, the study sample in fact consists mainly of Caucasian Christians, and entirely of U.S. female nurses. The suicide rate in our sample was about half of that on average among US women (2). Our results might thus not be generalizable to the general US population, to men, to other races, to other countries, or areas with limited religious freedom.(31) Our results are consistent with previous research among African Americans on suicide ideation and behaviors (32, 33), although suicide rates are in general lower among African Americans (2) and patterns across denominations may also differ for African Americans.
Even though religious service attendance has been used commonly in previous published studies, and tends to be the strongest predictor of health,(4) religiosity is multidimensional, and different aspects of religion and spirituality may therefore be differently associated with suicide. Service attendance was collected through a self-reported questionnaire and moreover may be subject to measurement error and possible over-reporting, though the relative ordering of frequency might still be preserved. Further research could examine other religious practices, mindfulness practices, other aspects of spirituality and religiosity, other race/ethnic and demographic groups, and other forms of social participation.
Our study made use of observational data. Although we adjusted for major confounders for the association between religious service attendance and suicide, the results may still be subject to unmeasured confounding by personality, impulsivity, feelings of hopelessness, or other cognitive factors. However, in sensitivity analysis, for an unmeasured confounder to explain away the effect of service attendance on suicide it would have to both increase the likelihood of service attendance and decrease the likelihood of suicide by over 10-fold, above and beyond the measured covariates. Such substantial confounding by unmeasured factors seems unlikely given adjustment for an extensive set of covariates, and the known risk factor associations for suicide.(34, 35) The results comparing Catholics and Protestant could also be subject to differential misclassification of suicide by affiliation(36), with Catholics potentially under-reporting suicide events due to suicide being considered a mortal sin. Suicide events were identified using U.S. National Death Index as well as reports from next of kin. In some cases, suicide events may be missed if the family members choose not to share. While such misclassification might affect Catholic-Protestant comparisons, it may be less likely to affect hazard ratios for service attendance within strata of affiliation. Our results were based on a relatively small number of events which can make modeling difficult; however, results were similar when using a technique, exact logistic regression, intended to address small event numbers. The small number of events also creates considerable uncertainty in the religious service attendance estimates stratified by affiliation.
Our study was intended to overcome methodological limitations in previous research. Strengths of our study include a large sample size, long duration of follow-up, prospective cohort study design, and repeated measures of religious service attendance, and extensive confounding control. We have clear temporality of the covariates, exposure, and outcome. We were able to adjust for baseline religious service attendance level to avoid reverse causation, and to estimate the effect of “incident exposure”; we were also able to adjust for depressive symptoms, and social support as potential confounders.
Our results do not imply that health care providers should “prescribe” attendance at religious services. However, for patients who are already religious, service attendance might be encouraged as a form of meaningful social participation; religion and spirituality may be an under-appreciated resource that psychiatrists and clinicians could explore with their patients, as appropriate.
Conclusion
In this large prospective long-term cohort study of US women, frequent religious service attendance was associated with substantially lower suicide risk.
Supplementary Material
Acknowledgements
Drs. VanderWeele, Li, Tsai, and Kawachi report no competing interests. The Nurses’ Health Study was funded by U.S. National Institutes of Health (NIH) UM1 CA186107. The analysis and paper was supported by a research grant from the Templeton Foundation. The funding agency played no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. Shanshan Li and Tyler J VanderWeele had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors have made (a) substantial contributions to the conception and design; or the acquisition, analysis, or interpretation of the data, (b) the drafting of the article or critical revision for important intellectual content, (c) final approval of the version to be published, and (d) agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. This study was approved by the institutional review boards at Harvard T. H. Chan School of Public Health.
Footnotes
Previous meeting presentation: None.
References
- 1.Jones DS, Podolsky SH, Greene JA. The Burden of Disease and the Changing Task of Medicine. The New England Journal of Medicine. 2012; 366(25):2333–2338. [DOI] [PubMed] [Google Scholar]
- 2.Nock MK, Borges G, Bromet EJ, Cha CB, Kessler RC, Lee S. Suicide and Suicidal Behavior. Epidemiologic reviews. 2008;30(1):133–154. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kachur SP, Potter LB, James SP, Powell KE. Suicide in the United States. National Center for Injury Prevention and Control. Govt Reports Announcements & Index (GRA&I), Issue 26, 2004. [Google Scholar]
- 4.Koenig HG. Handbook of religion and health (2nd ed.) Oxford; New York; Oxford University Press; 2012. [Google Scholar]
- 5.Gearing R, Lizardi D. Religion and Suicide. J Relig Health 2009;48(3):332–341. [DOI] [PubMed] [Google Scholar]
- 6.Durkheim E Suicide: a study in sociology. (1st Free Press paperback ed.) New York: Free Press; 1966. [Google Scholar]
- 7.Stack S Suicide: A 15-year review of the sociological literature part II: Modernization and social integration perspectives. Suicide Life-Threat. Behav 2000; 30(2):163–176. [PubMed] [Google Scholar]
- 8.Nisbet AP, Duberstein RP, Conwell RY, Seidlitz RL. The Effect of Participation in Religious Activities on Suicide Versus Natural Death in Adults 50 and Older. The Journal of Nervous and Mental Disease. 2000;188(8):543–546. [DOI] [PubMed] [Google Scholar]
- 9.Thompson MP, Ho C, Kingree J. Prospective associations between delinquency and suicidal behaviors in a nationally representative sample. Journal Of Adolescent Health. 2007;40(3):232–237. [DOI] [PubMed] [Google Scholar]
- 10.Rasic DT, Belik SL, Elias B, Katz LY, Enns M, Sareen J. Spirituality, religion and suicidal behavior in a nationally representative sample. Journal of Affective Disorders 2009;114(1):32–40. [DOI] [PubMed] [Google Scholar]
- 11.Spoerri A, Zwahlen M, Bopp M, Gutzwiller F, Egger M. Religion and assisted and non-assisted suicide in Switzerland: National Cohort Study. International Journal of Epidemiology. 2010;39(6):1486–1494. [DOI] [PubMed] [Google Scholar]
- 12.Dervic K, Oquendo MA, Grunebaum MF, Ellis S, Burke AK, Mann JJ. Religious affiliation and suicide attempt. American Journal of Psychiatry 2004;161(12):2303. [DOI] [PubMed] [Google Scholar]
- 13.Wray M, Colen C, Pescosolido B. The Sociology of Suicide. Annual Review of Sociology, 2011;37(37):505–528. [Google Scholar]
- 14.Kleiman EM, Liu RT. Prospective prediction of suicide in a nationally representative sample: religious service attendance as a protective factor. The British journal of Psychiatry: the journal of mental science. 2014;204:262–266. [DOI] [PubMed] [Google Scholar]
- 15.Harris EC, Barraclough B. Suicide as an outcome for mental disorders. A meta-analysis. The British journal of Psychiatry : the journal of mental science. 1997;170:205–228. [DOI] [PubMed] [Google Scholar]
- 16.Maselko J, Hayward RD, Hanlon A, Buka S, Meador K. Religious Service Attendance and Major Depression: A Case of Reverse Causality? American Journal of Epidemiology. 2012;175(6):576–583. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Colditz GA, Hankinson SE. The Nurses’ Health Study: lifestyle and health among women. Nature Reviews Cancer. 2005;5(5):388. [DOI] [PubMed] [Google Scholar]
- 18.Berkman L, Syme S. Social Networks, Host Resistance, and Mortality: A Nine-Year Follow-up Study of Alameda County Residents. American Journal of Epidemiology. 1979;109(2):186. [DOI] [PubMed] [Google Scholar]
- 19.Andresen E, Malmgren J, Carter W, Patrick DL. Screening for depression in well older adults- evaluation of a short-form of the CES-D. American Journal Of Preventive Medicine. 1994;10(2):77–84. [PubMed] [Google Scholar]
- 20.Hosmer DW, Lemeshow S, Sturdivant RX. Applied logistic regression. Wiley; (Third edition). 2013. [Google Scholar]
- 21.VanderWeele T Explanation in causal inference: methods for mediation and interaction. Oxford University Press; 2015. [Google Scholar]
- 22.MacKinnon DP. Introduction to statistical mediation analysis. Lawrence Erlbaum Associates, New York: 2008. [Google Scholar]
- 23.Ding P, VanderWeele T. Sensitivity Analysis Without Assumptions. Epidemiology 2016;27(3):368–377.. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.VanderWeele T Unmeasured confounding and hazard scales: sensitivity analysis for total, direct, and indirect effects. Eur J Epidemiol. 2013;28(2):113–117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Tsai AC, Lucas M, Kawachi I. Association between social integration and suicide among women in the United States. JAMA Psychiatry. 2015;72(10):987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Tsai AC, Lucas M, Sania A, Kim D, Kawachi I. Social integration and suicide mortality among men: 24-year cohort study of U.S. Health Professionals. Annals of Internal Medicine. 2014;161(2):85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Oquendo AM, Dragatsi KD, Harkavy-Friedman FJ et al. Protective Factors Against Suicidal Behavior in Latinos. The Journal of Nervous and Mental Disease. 2005;193(7):438–443. [DOI] [PubMed] [Google Scholar]
- 28.Aquinas Thomas. Summa theologica : ad manuscriptos codices a Francisco Garcia, Gregorio Donato, lovaniensibus ac duacensibus theologis, Joanne Nicolai ac Thoma Madalena diligentissime collata, variis indicibus aucta. Lutetiæ Parisiorum: J.-P. Migne; 1859. [Google Scholar]
- 29.Church Catholic. Catechism of the Catholic Church. Libreria Editrice Vaticana Loyola University Press, Vatican City: Chicago, Ill. 2 Ed 2000. [Google Scholar]
- 30.Selvin HC. Durkheim’s Suicide: Further Thoughts on a Methodological Classic. Émile Durkheim, Greenwood Press, Westport, Conn: 1965, pp 113–136. [Google Scholar]
- 31.Hayward R, Elliott M. Cross-National Analysis of the Influence of Cultural Norms and Government Restrictions on the Relationship Between Religion and Well-Being. Rev Relig Res. 2014;56(1):23–43. [Google Scholar]
- 32.Taylor JR, Chatters ML, Joe MS. Religious Involvement and Suicidal Behavior Among African Americans and Black Caribbeans. The Journal of Nervous and Mental Disease. 2011;199(7):478–486. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Chatters L, Taylor RJ, Lincoln K, Nguyen A, Joe S. Church-Based Social Support and Suicidality Among African Americans and Black Caribbeans. Archives of Suicide Research. 2011;15(4):337–353. [DOI] [PubMed] [Google Scholar]
- 34.Kessler RC, Berglund P, Borges G, Nock M, Wang PS. Trends in suicide ideation, plans, gestures, and attempts in the United States, 1990–1992 to 2001–2003. JAMA. 2005;293(20):2487. [DOI] [PubMed] [Google Scholar]
- 35.Nock M, Borges G, Bromet E, et al. Cross-national prevalence and risk factors for suicidal ideation, plans and attempts. The British Journal of Psychiatry. 2008;192(2):98–105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Fv Poppel, Day LH. A test of Durkheim’s theory of suicide - without committing the ‘ecological fallacy.’ Emile Durkheim. American Sociological Review. 1996;61(3):500. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.