Intrinsic Association of Widowhood With Mortality in Community-Dwelling Older Women and Men: Findings From a Prospective Propensity-Matched Population Study

Abstract

Objectives.

Widowhood is associated with increased mortality. However, to what extent this association is independent of other risk factors remains unclear. In the current study, we used propensity score matching to design a study to examine the independent association of widowhood with outcomes in a balanced cohort of older adults in the United States.

Methods.

We used public-use copies of the Cardiovascular Health Study data obtained from the National Heart, Lung, and Blood Institute. Of the 5,795 community-dwelling older men and women aged 65 years and older in Cardiovascular Health Study, 3,820 were married and 1,436 were widows or widowers. Propensity scores for widowhood, estimated for each of the 5,256 participants, were used to assemble a cohort of 819 pairs of widowed and married participants who were balanced on 74 baseline characteristics. The 1,638 matched participants had a mean (± standard deviation) age of 75 (±6) years, 78% were women, and 16% African American.

Results.

All-cause mortality occurred in 46% (374/819) and 51% (415/819) of matched married and widowed participants, respectively, during more than 11 years of median follow-up (hazard ratio associated with widowhood, 1.18; 95% confidence interval, 1.03–1.36; p = .018). Hazard ratios (95% confidence intervals) for cardiovascular and noncardiovascular mortalities were 1.07 (0.87–1.32; p = .517) and 1.28 (1.06–1.55; p = .011), respectively. Widowhood had no independent association with all-cause or heart failure hospitalization or incident cardiovascular events.

Conclusions.

Among community-dwelling older adults, widowhood was associated with increased mortality, which was independent of confounding by baseline characteristics and largely driven by an increased noncardiovascular mortality. Widowhood had no independent association with hospitalizations or incident cardiovascular events.

THE death of a spouse is a major life event and has been shown to be associated with mortality and other adverse outcomes (1–4). However, it is unclear to what extent these effects are confounded by the sociopsychological and physiological burdens associated with widowhood. Therefore, the objective of this study was to determine the independent association of widowhood with various outcomes in a propensity-matched cohort of community-dwelling older adults who would be balanced on all measured baseline characteristics.

METHODS

Study Design and Participants

Funded by the National Heart, Lung, and Blood Institute, the Cardiovascular Health Study (CHS) is an ongoing community-based epidemiologic study of 5,888 Medicare-eligible adults aged 65 years and older (5). The purpose of CHS is to investigate risk factors for cardiovascular morbidity and mortality in older adults. CHS participants were recruited from Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh County, Pennsylvania. The rationale and design of the CHS have been previously reported (5). Briefly, an original cohort of 5,201 participants was recruited between 1989 and 1990, which was supplemented by a second cohort of 687 African American participants recruited between 1992 and 1993 (5).

For the current analysis, we used a public-use copy of the CHS data obtained from the National Heart, Lung, and Blood Institute. Of the 5,888 original CHS participants, 5,795 consented to be included in the deidentified public-use copy of the data set and are included in our analysis. After excluding participants who were divorced (n = 303) and never married (n = 230), a total of 5,256 married and widowed male and female participants were included in the current analysis. Of these, 1,436 (27%) were widows or widowers. From these 5,256 participants, we assembled a cohort of 819 pairs of propensity-matched widowed and married persons who would be balanced on all measured baseline characteristics.

Baseline Characteristics

Data on marital status and other sociodemographic characteristics, health behaviors, self-rated health, comorbid conditions, medications, vital signs, markers of inflammation, and other biochemical covariates were collected at baseline and have been previously described in details (6,7). Missing values for continuous variables were imputed based on values predicted by age, sex, and race.

Outcome Measurements

Primary outcomes for this study were all-cause mortality and all-cause hospitalization during 13 years of follow-up (median, 11 years and interquartile range, 5 years). Secondary outcomes included cause-specific mortality and hospitalizations as well as incident cardiovascular events among those without those conditions at baseline. For example, for incident heart failure (HF), those with baseline prevalent HF were excluded. Data on mortality were obtained through active surveillance of the contacts of the participant and through obituary searches at the four study sites (8,9). Cause of death was determined by the CHS Events Committee. Data on hospitalization were self-reported and verified using CMS Medicare Utilization Files, and primary discharge diagnoses were verified by CHS Events Committee using International Classification of Diseases, Ninth Revision codes by medical record review (10). All incident cardiovascular events were centrally adjudicated, the process of which has been described in detail in previous literature (11,12).

Assembly of the Balanced Study Cohort

Because of significant imbalances in baseline characteristics between married and widowed individuals (Table 1), we used propensity score matching to assemble a cohort of participants so that the two groups would be well balanced on all measured baseline characteristics (13,14). We estimated propensity scores for widowhood for each of the 5,256 participants using a nonparsimonious multivariable logistic regression model (15–20). In the model, widowhood was the dependent variable, and the 74 baseline characteristics displayed in Figure 1 were entered as covariates. The propensity score for widowhood for a participant would be the conditional probability of that participant being a widow, given his or her measured baseline characteristics (13,14).

Table 1.

Baseline Characteristics of Participants of Cardiovascular Heart Study, By Widowhood, Before and After Propensity Score Matching

% (n) or Mean (SD)	Before Matching			After Matching
	Married (n = 3,820)	Widowed (n = 1,436)	p Value	Married (n = 819)	Widowed (n = 819)	p Value
Sociodemographics
Age, y	73 (5)	75 (6)	<.001	75 (6)	75 (6)	.90
Female (%)	47 (1,782)	82 (1,177)	<.001	78 (641)	77 (630)	.50
African American	11 (409)	23 (328)	.001	16 (128)	15 (126)	.95
Living alone (%)	1 (28)	38 (549)	<.001	3 (28)	4 (30)	.79
College education or higher	46 (1,737)	34 (490)	<.001	37 (299)	35 (288)	.61
Income ≥$25k	46 (1,747)	16 (225)	<.001	21 (170)	19 (158)	.43
Current smoker	10 (388)	14 (204)	<.001	14 (113)	13 (107)	.71
Alcohol/week	2.7 (6.6)	1.7 (5.1)	<.001	1.8 (5.6)	1.9 (5.5)	.62
Body mass index, kg/m2	27 (4)	27 (4)	.93	27 (4)	27 (4)	.98
Social support score	8 (2)	8 (3)	.16	8 (3)	8 (3)	.58
Comorbid conditions
Coronary heart diseases	20 (773)	20 (281)	.59	20 (163)	19 (155)	.65
Heart failure	3.9 (148)	7.0 (100)	<.001	4.8 (39)	6.3 (52)	.19
Hypertension	57 (2,163)	63 (909)	<.001	61 (496)	62 (508)	.56
Diabetes	15 (588)	18 (257)	.028	18 (151)	18 (143)	.65
Stroke	4 (158)	5 (69)	.29	4 (32)	5 (39)	.45
COPD	12 (452)	14 (203)	.024	13 (107)	13 (104)	.88
Cancer	15 (554)	15 (210)	.91	15 (124)	16 (129)	.79
Arthritis	50 (1,891)	56 (798)	<.001	53 (430)	53 (435)	.84
Self-reported fair to poor general health	23 (880)	30 (430)	<.001	27 (222)	28 (227)	.83
Physical and mental functioning
ADL score	0.10 (0.43)	0.19 (0.60)	<.001	0.14 (0.50)	0.14(0.52)	.92
IADL score	0.31 (0.67)	0.47 (0.89)	<.001	0.39 (0.70)	0.41 (0.82)	.60
Depression score	4.2 (4.3)	5.8 (5.0)	<.001	5.6 (5.2)	5.4 (4.8)	.42
MMSE score	27.7 (2.5)	27.0 (3.2)	<.001	27.2 (2.8)	27.2 (2.9)	.73
Medications
ACE inhibitors	7 (266)	8 (114)	.22	7 (55)	7 (56)	1.00
Beta blockers	14 (523)	11 (158)	.010	13 (102)	12 (98)	.82
Calcium channel blockers	13 (513)	14 (203)	.51	14 (117)	14 (114)	.89
Statins	2 (85)	2 (33)	.87	2 (17)	2 (19)	.87
Thiazide diuretics	11 (400)	13 (190)	.005	13 (104)	13 (102)	.94
NSAIDs	12 (450)	15 (210)	.006	14 (118)	15 (119)	1.00
Vital signs
Pulse rate, beats/min	67 (11)	69 (11)	<.001	69 (12)	68 (11)	.41
Systolic blood pressure, mm Hg	136 (22)	139 (22)	<.001	137 (23)	138 (22)	.57
Diastolic blood pressure, mm Hg	71 (11)	70 (12)	.005	70 (11)	70 (12)	.88
Other laboratory measures
Hemoglobin, g/dL	14.2 (1.3)	13.6 (1.3)	<.001	13.8 (1.3)	13.8 (1.3)	.51
Creatinine, g/dL	0.99 (0.38)	0.92 (0.36)	<.001	0.92 (0.34)	0.93 (0.40)	.86
Glucose, mg/dL	111 (35)	112 (36)	.20	112 (40)	113 (37)	.89
Total cholesterol, mg/dL	209 (38)	216 (40)	<.001	215 (39)	216 (40)	.77
Triglyceride, mg/dL	142 (78)	135 (73)	.009	139 (67)	138 (79)	.74
Albumin, g/dL	4.01 (0.29)	3.96 (0.29)	<.001	4.00 (0.29)	4.00 (0.29)	.92
Interleukin-6, pg/mL	2.1 (1.7)	2.4 (2.2)	<.001	2.3 (2.0)	2.3 (1.7)	.50
Fibrinogen, mg/dL	320 (64)	332 (70)	<.001	328 (64)	328 (67)	.81
Serum insulin, mcU/mL	17 (25)	18 (30)	.21	18 (30)	18 (29)	.84
C-reactive protein, mg/L	4.6 (7.7)	5.2 (9.3)	.022	4.8 (6.7)	5.0 (9.8)	.66

Note: ACE = angiotensin-converting enzyme; ADL = activity of daily living; COPD = chronic obstructive pulmonary disease; IADL = instrumental activity of daily living; MMSE = Mini-Mental State Examination; NSAID = nonsteroidal antiinflammatory drug.

Figure 1.

Love plot displaying absolute standardized differences of 74 baseline characteristics between married and widowed older adults in the Cardiovascular Health Study, before and after propensity score matching.

Because propensity score models are sample-specific adjusters and are not intended to be used for out-of-sample prediction or estimation of coefficients, measures of fitness and discrimination are not important for the assessment of the model's effectiveness (21–24). The efficacy of propensity score models is best assessed by estimating postmatch absolute standardized differences between baseline covariates that directly quantifies the bias in the means (or proportions) of covariates across the groups, expressed as a percentage of the pooled standard deviations. We therefore calculated pre- and postmatch absolute standardized differences and presented those findings as Love plots (25,26). An absolute standardized difference of 0% would indicate no residual bias and less than 10% would be considered of inconsequential bias. Using a greedy matching protocol described elsewhere, we were able to match 819 pairs of married and widowed participants who had similar propensity scores (21–24).

Statistical Analysis

For descriptive analyses, we used Pearson chi-square and Wilcoxon rank-sum tests for the prematch data and McNemar’s test and paired sample t test for postmatch comparisons as appropriate. Kaplan–Meier survival and Cox proportional hazard analyses were used to determine the association of widowhood with outcomes during 13 years of follow-up. We conducted a formal sensitivity analysis to quantify the degree of a hidden bias due to potential imbalance of an unmeasured covariate that would need to be present to invalidate our main conclusions (27). We then repeated our analysis in all 5,256 prematch participants using four different Cox regression models: (a) unadjusted; (b) age–sex–race adjusted; (c) multivariable adjusted, using all covariates used in the propensity score model; and (d) propensity score adjusted. Selected subgroup analyses were conducted to determine the heterogeneity of the association of baseline widowhood and mortality. All statistical tests were two sided, and tests with p value < .05 were considered significant. SPSS for Windows (Version 18) was used for all data analysis.

RESULTS

Baseline Characteristics

Overall, matched participants had a mean age 75 (±6) years, 78% were women, and 16% were African American. Imbalances in baseline characteristics before matching and balances achieved after matching between widowed and married patients are displayed in Table 1 and Figure 1. Before matching, widowed participants were more likely to be older, African American, live alone, have lower education and income, be current smokers, and have a higher prevalence of various comorbidities. However, widowed persons were also more likely to be female. After matching, standardized differences for all measured covariates were less than 10% (most were <5%), suggesting substantial covariate balance across the groups (Figure 1).

Association Between Widowhood and Mortality

All-cause mortality occurred in 46% and 51% of matched married and widowed CHS participants, respectively, during 13 years of follow-up (matched hazard ratio [HR] when widowed participants were compared with married participants, 1.18; 95% confidence interval [CI], 1.03–1.36; p = .018; Table 2 and Figure 2). In the absence of hidden confounders, a sign-score test for matched data with censoring provides strong evidence (p = .015) that older adults with spousal loss clearly had an increased risk of all-cause mortality than those who were married. A hidden binary covariate that is a near-perfect predictor of mortality would need to increase the odds of widowhood by 4.14% to explain away this association.

Table 2.

Association of Widowhood With Mortality

	% (Events/Total)		Absolute Risk Increase (%)	Hazard Ratios (95% Confidence Interval)
	Married	Widowed
All cause
Unadjusted*	43 (1,641/3,820)	53 (757/1,436)	+10	1.43 (1.31–1.56); p < .001
Age, sex, and race adjusted	—	—		1.33 (1.21–1.47); p < .001
Multivariable adjusted*†	—	—		1.22 (1.10–1.37); p < .001
Propensity matched	46 (374/819)	51 (415/819)	+5	1.18 (1.03–1.36); p = .018
Cardiovascular
Unadjusted*	18 (693/3,820)	22 (312/1,436)	+4	1.38 (1.21–1.58); p < .001
Age, sex, and race adjusted	—	—		1.28 (1.10–1.49); p = .002
Multivariable adjusted*†	—	—		1.17 (0.98–1.39); p = .08
Propensity matched	22 (176/819)	22 (177/819)	0	1.07 (0.87–1.32); p = .52
Noncardiovascular
Unadjusted*	25 (946/3,820)	31 (442/1,436)	+6	1.45 (1.30–1.63); p < .001
Age, sex, and race adjusted	—	—		1.36 (1.20–1.55); p < .001
Multivariable adjusted*†	—	—		1.25 (1.08–1.44); p = .003
Propensity matched	24 (197/819)	29 (236/819)	+5	1.28 (1.06–1.55); p = .011

Notes: ^*Results from prematched data.

^†Adjusted for all variables in Figure 1.

Figure 2.

Kaplan–Meier plot for all-cause mortality by widowhood in older adults in the Cardiovascular Health Study.

Prematch and postmatch associations of spousal loss with all-cause mortality by gender are displayed in Figure 3. The association of widowhood and all-cause mortality was homogenous across various demographic categories (Figure 4). Among the 5,256 prematched participants, unadjusted, age–sex–race–adjusted, and multivariable-adjusted HRs for mortality associated with widowhood were 1.43 (95% CI, 1.31–1.56; p < .001), 1.33 (95% CI, 1.21–1.47; p < .001), and 1.15 (95% CI, 1.02–1.29; p = .021), respectively (Table 2). Associations of widowhood with cause-specific mortality are displayed in Table 2.

Figure 3.

Kaplan–Meier plots for all-cause mortality by widowhood and sex in older adults in the Cardiovascular Health Study, before and after propensity score matching.

Figure 4.

Association of widowhood and all-cause mortality in subgroups of older adults in the Cardiovascular Health Study.

Association Between Widowhood and Hospitalization

All-cause hospitalization occurred in 86% and 86% of matched married and widowed CHS participants, respectively (matched HR when widowed participants were compared with married participants, 1.10; 95% CI, 0.99–1.22; p = .080; Table 3). Among the 5,256 prematched participants, unadjusted, age–sex–race–adjusted, and multivariable-adjusted HRs for hospitalization associated with widowhood were 1.14 (95% CI, 1.07–1.22; p < .001), 1.13 (95% CI, 1.05–1.21; p = .001), and 1.10 (95% CI, 1.01–1.20; p = .034), respectively (Table 3). Unadjusted, age–sex–race–adjusted, multivariable-adjusted, and matched HRs for associations of widowhood with cause-specific hospitalizations are displayed in Table 3.

Table 3.

Association of Widowhood With Hospitalizations

	% (Events/Total)		Absolute Risk Increase (%)	Hazard Ratios (95% Confidence Interval)
	Married	Widowed
All cause
Unadjusted*	84 (3,194/3,820)	86 (1,234/1,436)	+2	1.14 (1.07–1.22); p < .001
Age, sex, and race adjusted	—	—		1.13 (1.05–1.21); p = .001
Multivariable adjusted*†	—	—		1.10 (1.01–1.20); p = .034
Propensity matched	86 (700/819)	86 (707/819)	0	1.10 (0.99–1.22); p = .08
Pneumonia
Unadjusted*	21 (782/3,820)	23 (324/1,436)	+2	1.30 (1.15–1.49); p < .001
Age, sex, and race adjusted	—	—		1.24 (1.07–1.44); p = .004
Multivariable adjusted*†	—	—		1.17 (0.99–1.38); p = .06
Propensity matched	22 (180/819)	23 (192/819)	+1	1.16 (0.95–1.42); p = .15
Stroke
Unadjusted*	13 (498/3,820)	15 (212/1,436)	+2	1.28 (1.09–1.50); p = .003
Age, sex, and race adjusted	—	—		1.18 (0.98–1.41); p = .08
Multivariable adjusted*†	—	—		1.01 (0.82–1.25); p = .93
Propensity matched	15 (120/819)	14 (113/819)	−1	0.99 (0.77–1.28); p = .95
Acute myocardial infarction
Unadjusted*	10 (389/3,820)	10 (142/1,436)	0	1.09 (0.90–1.33); p = .36
Age, sex, and race adjusted	—	—		1.37 (1.10–1.70); p = .005
Multivariable adjusted*†	—	—		1.37 (1.07–1.76); p = .013
Propensity matched	10 (80/819)	11 (89/819)	+1	1.19 (0.88–1.61); p = .25
Cancer
Unadjusted*	14 (533/3,820)	13 (190/1,436)	−1	1.04 (0.88–1.23); p = .66
Age, sex, and race adjusted	—	—		1.19 (0.99–1.43); p = .06
Multivariable adjusted*†	—	—		1.05 (0.84–1.29); p = .68
Propensity matched	12 (97/819)	12 (100/819)	0	1.07 (0.81–1.41); p = .65
Heart failure
Unadjusted*	12 (464/3,820)	15 (211/1,436)	+3	1.39 (1.18–1.63); p < .001
Age, sex, and race adjusted	—	—		1.30 (1.08–1.56); p = .006
Multivariable adjusted*†	—	—		1.17 (0.95–1.45); p = .16
Propensity matched	15 (119/819)	16 (129/819)	+1	1.14 (0.89–1.47); p = .29

Notes: ^*Results from prematched data.

^†Adjusted for all variables in Figure 1.

Association Between Widowhood and Incident Cardiovascular Events

Of the 1,547 matched participants without prevalent HF at baseline, incident HF occurred in 23% of both matched married and widowed participants (matched HR, 1.04; 95% CI, 0.84–1.28; p = .731; data not shown in table). Among the 5,008 prematch participants without prevalent HF at baseline, unadjusted, age–sex–race–adjusted, and multivariable-adjusted HRs for mortality associated with widowhood were 1.30 (95% CI, 1.14–1.49; p < .001), 1.21 (95% CI, 1.04–1.41; p = .013), and 1.15 (95% CI, 0.96–1.37; p = .130), respectively (data not shown in table).

The unadjusted significant association of widowhood with incident stroke among the 5,029 participants without baseline stroke (HR, 1.30; 95% CI, 1.11–1.53; p = .001) became nonsignificant after adjustment for age, sex, and race (HR, 1.05; 95% CI, 0.88–1.26; p = .592) or in the multivariable model (HR, 0.85; 95% CI, 0.65–1.11; p = .243). Widowhood had no association with incident stroke among the 1,567 matched participants without baseline stroke (HR, 0.88; 95% CI, 0.68–1.14; p = .345). Widowhood had no significant unadjusted or adjusted associations with any other incident cardiovascular events.

DISCUSSION

Summary and Relevance of the Key Findings

Findings from the current study demonstrate that among community-dwelling older men and women, baseline widowhood had an intrinsic association with mortality, which was independent of 74 measured baseline demographic, clinical, subclinical, and biochemical characteristics, and was mostly driven by an increase in noncardiovascular mortality. Widowhood also had significant age–sex–race–adjusted associations with cardiovascular mortality, all-cause and cause-specific hospitalizations, and incident HF but lacked an intrinsic association with these outcomes. These findings provide important insights about the prognostic impact of widowhood on outcomes in community-dwelling older adults and suggest that the impact of widowhood may not be explained by commonly measured confounders and mediators of poor outcomes.

Potential Explanation and Mechanism of the Key Findings

As in any observational study, the significant independent association of widowhood with increased mortality in our study has three potential explanations: residual bias by measured confounders, bias by unmeasured confounders, or a true intrinsic association. The concern about residual bias due to imbalances in any of the 74 measured baseline characteristics is attenuated by the findings from our propensity-matched cohort in which widowed and married individuals were balanced on those characteristics. Although our sensitivity analysis suggests that the observed association may be relatively sensitive to an unmeasured confounder, sensitivity analysis cannot determine if such an unmeasured confounder exists. To be a confounder, an unmeasured baseline characteristic would need to be a near-perfect predictor of mortality, would need to increase the odds of widowhood by more than 4%, and could not be strongly associated with any of the 74 measured baseline characteristics, a possibility that seems highly unlikely. Therefore, an unmeasured confounder cannot fully explain away the findings from our matched cohort.

The independent effect of widowhood on mortality is unlikely to be direct and is more likely mediated via other risk factors. Although all these measured risk factors were balanced at baseline in our matched cohort, widowed older adults may have had a higher incidence of these conditions during follow-up (28). This is evident from the higher prematch baseline prevalence of many such risk factors, such as hypertension, diabetes, and HF among widowed individuals (Table 1). Although it possible that older widowed individuals had a higher prevalence of these conditions, these conditions would be unlikely to have caused spousal death. On the other hand, if widowed older adults developed these conditions after widowhood, it is plausible that prebaseline postwidowhood psychosocial stress may have contributed to the higher baseline prevalence of these conditions. Widowhood is known to be associated with increased risk of depression and decline in cognitive and physical function, which may act as additional mediators (29). Similarly, chronic economic hardship and the erosion of environmental, psychological, and social support resources associated with widowhood may lead to deleterious physical and emotional health outcomes (30–32).

The lack of a strong and significant intrinsic association with hospitalization in our matched cohort may suggest that widowhood-associated mortality was due to sudden cardiac death that precluded hospitalization. However, findings from our cause-specific mortality analyses demonstrate that widowhood had no intrinsic association with cardiovascular mortality. Because noncardiovascular death is seldom sudden, it suggests that widowed older adults were less likely to seek medical care or more likely to receive home health or hospice care leading to death outside of the hospital. These findings also suggest that the association of widowhood with outcomes is complex and poorly understood, and these associations need to be prospectively examined, preferably using designs that would allow repeated-measures analyses.

Comparison With Findings From Relevant Published Literature

Findings of the current analysis are consistent with those from two recent meta-analyses on marital status and risk of mortality (33,34). In one meta-analysis based on 40 cohort studies, compared with married older adults, older widowed individuals had a higher risk of death (33). In another meta-analysis of 87 studies, being married was associated with reduced mortality (34). In that study, larger social networks and higher levels of perceived social support were also associated with reduced mortality. However, to the best of our knowledge, the current study is the first to examine the association between widowhood and outcomes in a propensity-matched cohort of community-dwelling older adults.

Clinical and Public Health Importance

Because widowhood is an unmodifiable risk factor, unadjusted associations are important. A careful review of the baseline characteristics (Table 1) provides insights into the unadjusted poor outcomes for widowed older adults. More than 80% of widowed individuals were women. Despite this apparent gender advantage, widowed older adults were disadvantaged by their older age, poorer sociodemographic and functional status, and higher morbidity burden, all markers and risk factors for poor outcomes. Spousal loss is often considered a benign normal grief. According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV), depression is a major mental illness, yet grief is considered a normal reaction to loss. Although the association between depression and mortality in older adults has not been clearly established (35), findings from our study suggest a significant intrinsic association between widowhood and mortality. Whether addressing risk factors associated with widowhood would improve outcomes in widowed older adults remains unknown and would need to be examined in well-designed prospective studies.

Potential Limitations and Future Direction

Several potential limitations of this analysis need to be acknowledged. There were no data on the onset and duration of widowhood prior to baseline. Also, we had no data on other incident bereavement events, such as loss of an adult child, parent, or sibling during follow-up. However, such bereavement is likely to have occurred similarly between both groups. It is possible that individuals who were married at baseline became widowed during follow-up and vice versa. However, the resultant regression dilution may have underestimated the true associations observed in our study (36). Because of the small number of men in the current study, these findings may not be generalizable to older men. The loss of study participants during matching is unlikely to have affected our findings as the multivariable regression-based analyses in the prematch data also showed similar results.

Conclusions

In conclusion, among community-dwelling older adults, the state of widowhood is associated with an increased risk of mortality, which was independent of confounding by baseline characteristics and was largely driven by an increased risk of noncardiovascular mortality. Widowhood had age–sex–race–adjusted associations with hospitalization or incident HF, but these associations were not independent.

FUNDING

Dr. B.R.W. is supported by grants from the Veterans Administration Health Services Research and Development (PPO 09-315-1 and IIR 03-026). Dr. A.A. is supported by the National Institutes of Health through grants (R01-HL085561 and R01-HL097047) from the National Heart, Lung, and Blood Institute and a generous gift from Ms. Jean B. Morris of Birmingham, Alabama.