Excess Mortality for the Unmarried in Rural Bangladesh
Citation: Rahman (1993) Excess Mortality for the Unmarried in Rural Bangladesh.
Introduction: "Hither mortality risks for widowed, divorced and never-married young men and women relative to their currently married counterparts have been documented in developed societies for about 100 years. However, there has been no consensus on the factors explaining these differentials" (p. 445). There are two main competing hypotheses: (1) The "selection hypothesis" argues that healthy people are more likely than sick/disabled people to get married. Hence, marriage is not protective; rather, a residual group of sick/disabled people never marries, producing the spurious association between marriage and health. (2) The "protection hypothesis" suggests that marriage has beneficial behavioral and socioeconomic qualities which facilitate health. "The purpose of this study is to investigate mortality differentials by marital status for men and women in [Bangladesh]" (p. 445). Methods: Data are from the Matlab demographic surveillance system in Bangladesh. The subjects are approximately 64,000 men and women (aged 15-45) who were interviewed in the 1974 census of the Matlab area. In 1982, these individuals were followed up to determine marital status and mortality. "Disability status, educational levels and household assets were assessed at the beginning of the study and were assumed to be constant for the period of follow- up" (p. 446). In addition to descriptive information, discrete time hazard models with 1-year time intervals within a logistic regression framework were used for multivariate analyses. Results: Proportions exposed to divorce and widowhood was fairly small for both men (.76% and .25%, respectively) and women (1.63% and 5.10%). Remarriage after divorce was common for males (68%), but somewhat less so for females (46%). Disability was much more prevalent among the never-married and divorced than among the currently married. For instance, 23% of women and 14% of men who did not marry by age 40-44 reported a disability, compared to less than 1% who married for both sexes. "In general, never married and divorced individuals had higher mortality rates than their currently married peers . . ." (p. 448). (There were too few widowers in the study to conclude anything about their mortality risks). In addition, "The excess mortality risk associated with being never married relative to the currently married increased with age for both men and women . . ." (p. 448). Multivariate logistic regression indicated that divorced and never-married females had higher odds of dying than the currently married (O.R. never married/married = 1.67; O.R. divorced/married = 3.03). When controls for disability were introduced, these effects were attenuated but not eliminated. The logistic models for males produced very similar results. Conclusions: The unexplained excess mortality among the never married and divorced may be explained by (1) adverse changes in social interaction, nutrition and behavioral patterns associated with the single state and divorce and/or (2) the inability for unmarried individuals-particularly women-to access enough household resources to meet health needs. But, despite this gap in mortality that is unexplained by age, disability or household assets, Rahman contends that this study provides greater support for the selection hypothesis than the protective hypothesis-that is, "mortality differentials by marital status in rural Bangladesh can be [largely] attributed to selection into marriage and remarriage on the basis of underlying disability status" (p. 455).