|Home | About | Journals | Submit | Contact Us | Français|
To examine the risk and correlates of mortality following death of a spouse and whether mortality risk varies by gender.
Prospective cohort study (1993–2000) of 1693 Mexican Americans aged 65 years and older that were married at baseline. Mortality was confirmed by matching records with the National Death Index or through proxy report. Risk of death related to incidence of widowhood was estimated using proportional hazard regression and adjusted for age, education, US nativity, financial strain, social support, health behaviors, medical conditions, disability, and depressive symptoms.
In the unadjusted Cox hazard analysis, widowed men are significantly more likely to die (HR = 2.32, CI = 1.48–3.61), but loss of spouse has no significant effect on the subsequent risk of death for widowed women (HR = 1.50, CI = .90–2.49). After adjustment for covariates known to influence survival, the association between widowhood and mortality in men remained significant but the magnitude of the association decreased by 26%, which suggests a partial mediation effect of these factors on survival. The trajectory of the survival curve shows that the risk of death associated with widowhood is highest within the first two years.
Widowhood in older Mexican American men is a risk factor for mortality.
Widowhood is one of the most significant negative life events at any age. Spousal loss often manifests as increased risk of depression, disability, and cognitive impairment (1–4). One of the most consistent findings in the literature is that spouses are at high risk for mortality within 6 to 12 months after the death of their spouse (5–11). These studies also agree that both men and younger spouses are at higher risk for mortality than women and older spouses after their spouse dies. Whether the widowhood effect on mortality varies by race/ethnicity is unclear. Few studies have explored this question and the results have been mixed. One study found no differences by race (12), while another large study of white and black couples found substantial variation by race. White couples were more prone to long term negative effects of conjugal loss, while black couples did not show an appreciable widowhood effect (13).
Explanations for the higher mortality risk of widowed spouses have focused on SES and to a limited extent on health and health behaviors. The role of socioeconomic circumstances of the surviving spouse has had mixed results. One study found an association between individuals with the lowest SES and highest mortality risk (14). Other studies showed higher SES individuals at higher risk of post-widowhood mortality (12, 15). The authors of the latter studies conjectured that the more highly educated may have higher quality relationships making the death more stressful, or that cultural and structural characteristics of their social support network may lead to less available social support. In part supporting the role of social support as a buffer to distress of widowhood, some evidence has been presented that mortality risk is negatively associated with the number of children (16). Health related explanations are limited, but a pair of studies concluded that prior health status including chronic disease and disability (10, 11) explained part of the association between widowhood and mortality. Depression is hypothesized to exacerbate the mortality risk of widowhood because it has been shown to be independently associated with bereavement and death (17). Biological reaction to the distress of conjugal loss has been documented recently in a study of elderly respondents. Bereavement was negatively associated with antibody response to influenza vaccination, and being married with high marital satisfaction was associated with higher response to influenza exposure (18). Finally, health risk behaviors are associated with widowhood, particularly health regulation behaviors such as medication compliance (19), and some studies have found evidence that shared health behaviors explain some of the association of mortality (11, 16) but another study found no significant effect (10).
Most of the previous research has lacked baseline data on individuals prior to the loss of their spouse which leaves the possibility open that the higher risk of mortality is due, in part, to similar health profiles of spouses (20, 21). Another limitation of previous research is that few studies have included measures of health status and only one study incorporated health variables as time dependent covariates (11). By including time dependent measures of health that affect survival such as health behaviors, chronic disease, disability, and depression, the association of widowhood with mortality may be explored in greater detail and may point to potential interventions to lower post-widowhood mortality. Another limitation of previous research is that many studies were composed primarily of non-Hispanic White respondents with very little representation of minorities. Therefore, this study examines the risk of mortality following death of a spouse after adjusting for time dependent measures of health, health behaviors, social support, and financial strain using data from older Mexican Americans followed over seven years. This study uses a unique database of a large sample of Mexican Americans so that the results can be compared to other findings in the literature on Non-Hispanic Whites. Consistent with previous research, we expect that men will have a higher risk of death related to conjugal loss than women.
This study used data from the Hispanic Established Populations for the Epidemiologic Studies of the Elderly, 1993–2000 (Hispanic EPESE) (22). The Hispanic EPESE is a multistage random sample of older Mexican Americans, aged 65 years and older, residing in Arizona, California, Colorado, New Mexico, and Texas. About 85% of the older, Mexican American population resides in these five states. The baseline data for the Hispanic EPESE included 3,050 respondents 65 years and older. Participants were reinterviewed in 1995, 1998, and 2000. At baseline, there were 1693 married individuals, 955 widows, and the remaining sample was divorced, single, or separated. This study followed the 1693 respondents that were married in 1993 over seven years to track incidence of widowhood.
Mortality was confirmed by matching records with the National Death Index or through proxy report. Survival time was recorded in months beginning with the start of the study in 1993 and ending with the death of the respondent or the last follow up in 2000. The mortality outcome for 107 respondents could not be determined by a successful match with the National Death Index or through proxy report so those cases are excluded from analysis. None of these cases were widowed over the study period. Over seven years of follow-up, 482 respondents died from the subsample of 1693 married respondents at baseline. Continuously married respondents accounted for 441 deaths and respondents that became widowed over the study period accounted for the remaining 41 deaths. Table 1 provides the number of respondents that were alive, dead, or lost to follow-up (LTF) over the seven year study by widowhood status.
Widowhood was a binary time-dependent variable indicating the respondent’s widowhood status over the study period. At each follow up, respondents reported if their marital status changed and the date on which the change occurred. Time spent as a widow was calculated in months. Over the course of the study period, 239 respondents became widowed. Table 2 presents data for the weeks that the 239 respondents were widowed until death or the end of the study.
Time invariant covariates included demographic characteristics of the respondents in 1993. Age is continuous in years and ranges from 65 to 94 at baseline. Education is continuous in years and ranges from 0 to 17. A binary question asked respondents if they were born in the United States.
Time dependent covariates were included over four waves of the HEPESE. Financial strain assesses how much difficulty the respondents have in meeting monthly payments on bills. Social support was measured with two items that asked how often the respondent could count on and talk about problems with family or friends when needed. Health behaviors were measured with binary indicators for current smoking or drinking alcohol. Body mass index, defined as weight in kilograms divided by height in meters squared, was measured as a continuous indicator. Chronic disease was assessed by asking respondents if a doctor had ever told them they had any of the following conditions: heart disease, stroke, diabetes, hip fracture, cancer, hypertension, and arthritis. The index was created by summing the conditions and ranged from 0 to 7. Past studies have shown good agreement between self-reported medical conditions and conditions documented by the patients’ physicians and proxy (23, 24). A performance based measure of lower body mobility was assessed with a standing balance, a timed eight-foot walk, and timed repeated chair stands (25). The three measures were summed and ranged from a low of 0 (unable to perform) to a high of 12. Depressive symptoms were measured with the Center for Epidemiologic Studies Depression scale (CES-D) (26) and ranged from 0 to 60. Characteristics of the sample are presented in Table 3.
To determine the association between widowhood and mortality, only incidence of widowhood over the study period was measured in the analyses. Cox proportional hazard models were used to model the association of widowhood and mortality over the course of seven years. Respondents that became widowed during the study period were compared to married respondents. Our data is left-truncated because widowed subjects are not at risk of death associated with widowhood until he/she becomes widowed. Therefore, in our analyses, widowhood was treated as a time-dependent covariate to study hazard of death. For widowed subjects, they contributed to the partial-likelihood of non-widow from entry to study to the time of widow, and contributed to the partial-likelihood of widow from time of widow to death or end of study. The first analysis tracked the age adjusted survival curve over seven years stratified by gender and widowhood. Then, hierarchical Cox proportional hazards models were used to assess the association of widowhood and mortality adjusted for both time varying and time invariant covariates. These models were presented separately for men and women. A final test charted the trajectory of the survival curve for widowed men and women. There is the possibility that married couples could be enrolled conjointly in the study; however, nonindenpendence of observations is not violated because all analyses are stratified by gender and adjusted for cluster effects. Baseline missing data was multiple imputed with five datasets which could reach 96.7% efficiency for the amount of missing in our data (27). The imputation model included all available respondent characteristics in our analytical model. For missing covariates in the follow-up, the last measured value is carried over in the analyses of time-dependent models. All analyses were performed with SAS, version 9 (SAS Institute, Inc., Cary, NC). All tests were two-tailed and P ≤ .05 was considered statistically significant.
Figure 1 presents the age adjusted survival rate for the sample stratified by gender and marital status. The groups are divided into male widowed, male non-widowed, female widowed, and female non-widowed. Overall, the survival rate decreases over time for each group. By the end of the study period, married women have the highest survival rate, followed by widowed women, married men, and widowed men. Over seven years, the difference in the survival rate between married women and widowed women is 6.4% and is 22.1% between married men and widowed men. Widowed women and married men have about the same survival rate over time (67.2% vs. 71.2%).
Table 4 shows the results from the proportional hazards regression of mortality and widowhood for men with hierarchical adjustment for time invariant and time dependent covariates. The unadjusted hazards ratio shows that widowed men are at higher risk of mortality than married men (HR = 2.32, CI = 1.48–3.61). Adjustment for other covariates reduces the risk by 26% (HR =2.06, CI = 1.31–3.23). As expected, alcohol consumption, smoking, body mass index, chronic conditions, lower body mobility, and depressive symptoms are statistically significant in their relation to mortality.
Table 5 shows the results from the proportional hazards regression of mortality and widowhood for women with hierarchical adjustment for time invariant and time dependent covariates. The unadjusted hazards ratio is presented first with widowed women at higher risk of death compared to married women but this risk is not statistically significant (HR = 1.50, CI = .90–2.49). Adjustment for other covariates reduces the risk by 20% but the hazard ratio remains statistically insignificant.
Figure 2 presents the age adjusted cumulative hazard rate for death for widowed men and women as a function of time widowed. As in previous analyses, men have a higher risk of death over time than women. For both genders, risk of death increases over time but the gap between men and women increases. The trajectory of risk begins to flatten after 24 months, and it is flat by 33 months for both men and women. This graphic suggests that the highest risk of death associated with widowhood is within 24 months. There appears to be little increased risk of death after 33 months.
The aim of this study was to assess the association of widowhood and mortality among older Mexican Americans. The results suggest that widowhood puts men at higher risk for mortality, but it does not seem to have the same risk for women. Overall, our findings are consistent with other studies by finding increased risk of death for husbands. The primary difference with our finding is that the length of time that husbands are at risk of death is extended. Most studies find the risk of death levels out within 24 months, but our study of Mexican Americans revealed that the risk of death after losing a spouse levels out by 33 months which may suggest that Mexican American men have a longer risk of death related to widowhood than non-Hispanic whites. The longer trajectory of risk may also indicate that Mexican American men may take longer to adjust to the loss of a spouse.
Our study confirms the importance of a spouse for survival, at least for men. Marriage confers many health benefits over being single, divorced, or widowed such as lower depressive symptoms, better physical health, and higher life satisfaction (28–31), but these advantages vary by gender with husbands benefiting more from marriage than women (32, 33). A recent study of twins found that survival is extended by having a spouse and close ties with friends and family (34). Gender differences in the health benefits of social networks reveal that men tend to have few friends and rely primarily on their spouse for support, while women have broader social ties (35). Therefore, one hypothetical explanation for the association of widowhood with increased mortality of the surviving spouse is the loss of potential health benefits of marriage. In part, this may explain why studies consistently find higher mortality risk related to widowhood for men because the marital relationship is more important to maintaining health for men than it is for women.
Most studies on Mexican American spouses have found that wives experience more burden related to their husbands’ health than husbands do related to their wives’ health (36, 37). The implication is that wives have more caregiving burden than husbands. Therefore, although the death of a husband may be depressing, it may act to lessen the caregiving burden for wives. Conversely, husbands consistently show little evidence of caregiving burden for their wives so their death may signal the beginning of self-care for which they may be ill-equipped.
Although one of the strengths of this study was inclusion of several time dependent indicators of health, health behaviors, social support, and financial strain, these measures did not completely explain the association of widowhood and mortality for men. Respondents died of other causes that could not be included in the analysis such as respiratory illness, accidents/suicide, digestive illness, and unknown causes of death. Future research that can control for a broader range of diseases may provide an explanation for the association of widowhood and mortality, especially because a majority of widows die shortly after their spouse. For example, clinical studies have found an association of loneliness with impaired cellular immunology, which in turn predicts vulnerability to infection (38). It is also possible that the stress of widowhood worsens the severity and compromises the management of pre-existing chronic conditions. In that scenario, acute complications from these conditions could lead to greater mortality.
Recent research has indicated that distress due to widowhood is conditional on marital quality and differs substantially by gender with wives generally experiencing higher distress than husbands (39, 40). Therefore, these results should be interpreted with caution given that this database was not able to account for marital quality or spousal support. Absence of personality traits and other psychosocial characteristics is also a limitation of this study. An additional limitation is that incidence of widowhood was low throughout the seven year follow-up, which could affect the statistical significance. Finally, because of lack of blood data, we could not explore the potential link between increased blood stress markers with widowhood. In light of these limitations, future studies should examine change in health behavior markers after widowhood such as decreased physical activity, adherence to medical appointments, medication compliance, and flu vaccinations. Are there changes in the level of biomarkers for stress such as cortisol and interleukin-6? Is there a decrease in the level of social engagement? Providing answers to these questions in future studies will be key to the design, development, and testing of interventions to improve post-widowhood survival and health-related quality of life. An example of such interventions is discussion of hospice use with patients, particularly men, whose spouses face impending death. A recent study found that spouses who used hospice care before and after the loss of their spouse showed reduced mortality risk compared to widowed spouses who did not use hospice care (41). In summary, our results provide an insight into a risk factor for death among older Mexican Americans, a fast growing segment of the elderly population in the United States.
Funding Source: This study was supported by grants from the National Institute on Aging (R01AG10939, R01AG21089, T32AG000270), the National Cancer Institute (1 P50 CA105631-02), and the Bureau of Health Professions’ Geriatric Academic Career Award (1 K01 HP 00034-01).
This study was supported by grants from the National Institute on Aging (R01AG10939, R01AG21089, T32AG000270), the National Cancer Institute (1 P50 CA105631-02), and the Bureau of Health Professions’ Geriatric Academic Career Award (1 K01 HP 00034-01).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.