Wearing the Garment of Widowhood: Variations in Time Since Spousal Loss Among Community-Dwelling Older Adults

Abstract

We examine how the passage of time since spousal loss varies by social and demographic characteristics, using data from the University of Alabama at Birmingham Study of Aging. In multivariate analyses, African American race, female sex, lower income, and higher risk of social isolation had significant and independent associations with variation in time since spousal loss. African American women were at highest risk for long-term widowhood. Accurate characterizations of widowhood among community-dwelling older adults must consider variation in the length of time individuals are living as widowed persons and socioeconomic concomitants of long-term widowhood.

INTRODUCTION

Spousal loss occurs within a temporal dimension of the human lifespan. The period of time before and after the loss of a spouse is an important aspect of the spousal loss experience (Bonanno, 2009; Burke, Shrout, & Bolger, 2007; Carr, 2004; Caserta, Lund, & Diamond, 1989; Choi, 1995). In previous research we examined how pre-widowed persons differed from other older adults, how women incur a higher frequency of spousal loss, and how widows are subject to negative health outcomes following the subsequent loss of another family member (Williams, Baker, & Allman, 2005; Williams, Baker, Allman, & Roseman, 2006; Williams, Sawyer, Roseman, & Allman, 2008). In this article, we characterize the concept of time since spousal loss to expand our understanding of the temporal scope of the spousal loss experience. While some studies use current “age” as an indicator of duration and impact of widowhood, in this paper, we focus instead on the continuous length of time the individual has been living as a widowed person (Holden, Kim, & Novak, 2010).

There is ongoing interest in the period of time following the loss of a spouse, as individuals react to the transition from a married to widowed status. The bulk of literature on the temporal dimension of spousal loss focuses on the time course of morbidity or mortality following the bereavement event (Holden, Kim, & Novak, 2010; Strobe, Schut, & Strobe, 2007). There are significant racial variations in duration of marriage, as well as important gender differences in the length of widowhood (Carnelley, Wortman, Bolger, & Burke, 2006; Carr, 2004; Holden, Kim, & Novak, 2010; Strobe, Schut, & Strobe, 2007; Williams, Sawyer, Roseman, & Allman, 2008).

Although recent research compares levels of adaptation and resilience among short-term and long-term widows (Hahn, Cichy, Almeida, & Haley, 2011), to the best of our knowledge, no studies have fully characterized or provided detailed descriptions of social and demographic variations by time since spousal loss among older adults. The purpose of this article is to examine the variation in an array of key social and demographic constructs by time since spousal loss among African American and White community-dwelling adults age 65 years and older, controlling for physical, emotional, and cognitive status.

METHODS

We examined time since spousal loss using data from the University of Alabama at Birmingham (UAB) Study of Aging, a population-based, prospective, observational study of community-dwelling adults age 65 years and older, recruited between 1999 and 2001. Participants were initially recruited from a random sample of Medicare beneficiaries living in five counties in central Alabama, stratified by county, race, and sex. The study oversampled African Americans, men, and rural residents to balance race, sex and residence categories. A second in-home assessment was conducted 48 months following the baseline assessment. At the baseline interview, time since spousal death was not reported but was available at the 48-month assessment. Participants reporting a current widowed status at the 48-month follow-up were included in this analysis. The study protocol was reviewed and approved by the UAB Institutional Review Board.

Sociodemographic Measures

Time since spousal loss in years was self-reported during the 48-month interview. Age was a continuous variable. Other socio-demographic variables were dichotomized: sex, race: African American versus White; residence: urban versus rural residence; education: <12 years versus ≥ 12 years; and household income: <$12,000 versus ≥ $12,000 per year. Race/gender subgroups were: African American female, African American male, White female and White male. We defined transportation difficulty as a positive response to: “Over the past 4 weeks, have you had any difficulty getting transportation to where you want to go?” or “Do you limit your activities because you don’t have transportation?” We measured Social Isolation Risk using categories defined by the Lubben’s Social Network Scale, which examines the individual’s living arrangements, size of social networks, and reciprocal social supports, as well as frequency and emotional closeness of social contact family and friends (Lubben, 1988).

Health and Function Measures

We assessed psychological distress, using the 15-item Short form of the Geriatric Depression Scale (GDS) (Sheikh & Yesavage, 1986), and the five-item Short form of the Arthritis Impact Measurement Scales (AIMS 2) Anxiety subscale (Meenan, Mason, Anderson, Guccione, & Kazis, 1992). To measure cognitive function we utilized a variant of a 30-item assessment patterned after the Folstein Mini-Mental State Examination (Folstein, Folstein, & McHugh, 1975). To assess self-reported health, we used the SF-12 Health Survey (Stewart, Hays, & Ware, 1988). Comorbidity burden was operationalized as a cumulative score derived from the diseases of the Charlson Comorbidity index without consideration of severity (Charlson, et al., 1986). We used the UAB Life-Space Assessment to measure mobility based on the distance through which a person reported moving during the four weeks preceding the assessment. Frequency of movement and use of assistance (from equipment or persons) were also assessed to compute a composite Life-Space score, ranging from 0 to 120, with higher scores reflecting greater, more frequent, and independent mobility (Baker, Bodner, & Allman, 2003).

Statistical Analysis

For purposes of analysis, we aggregated data into time since spousal loss categories, each approximating 25% of the sample: (1) <1–5 years, (2) 6–10 years, (3) 11–22 years, (4) >22 years ago. We used CHI-Square and ANOVA to examine bivariate associations between time since spousal loss categories and socio-demographic, health, and function variables. We created series of multiple linear regression models to identify significant and independent predictors of time since spousal loss. Finally, we used a series of multivariate logistic regression models to identify significant and independent predictors of long-term elapse of time since spousal loss.

Sample

Of 1000 participants in the original cohort, 217 died, 7 withdrew, 18 were in a nursing home, and 43 could not be contacted at the 48-month assessment. Persons had to be able to answer questions by themselves to be offered the in-home assessment. Of the 686 participants eligible, 624 (91%) completed the 48-month in-home assessment. Two hundred and sixty participants (41.6%) reported a current widowed status and serve as the basis of this report.

RESULTS

In this sample, time since spousal loss ranged from <1 year to 60 years with a mean of 15.5 years. Overall, 50.4% of participants had been widowed 10 years or less. Table 1 displays the mean years since spousal loss by quartiles.

TABLE 1.

Time Since Spousal Loss Quartiles in Years

	N (%)	*Mean (SD)
All	N=260 (100.0)	15.5 (13.3)
≤5 years	N=64 (24.6)	2.7 (1.5)
6–10 years	N=67 (25.8)	8.3 (1.4)
11–22 years	N=62 (23.8)	16.0 (3.3)
≥23 years	N=67 (25.8)	34.6 (9.7)

^*P=<.001

Bivariate Associations

In the bivariate analysis (Table 2), we found significant dose-response trends across quartiles for sex, race, income, life-space mobility and transportation difficulty, with females and African Americans exhibiting the longest elapsed times since the death of a spouse. Although we had anticipated that time since spousal loss would be a function of age, education, and residence, our bivariate analysis did not find a significant association between these variables and time since spousal loss. Additionally, we found no significant differences among spousal loss categories on measures of psychosocial and physical health. In bivariate analyses using race-sex categories (Table 3), we found time since spousal loss was distributed unevenly among community dwelling elders. Compared to other widowed persons in our sample, African American females exhibited the longest mean length of time since the onset of widowhood.

TABLE 2.

Sociodemographic, Health and Function Correlates by Time Since Spousal Loss Among Older Adults

			Length of Widowhood
	All Widows		<1–5 years		6–10 years		11–22 years		>22 years
	n = 260	%	n = 64	%	n = 67	%	n = 62	%	n = 67	%	p value
Female	206	79.2%	43	67.2%	48	71.6%	52	83.9%	63	94.0%	<.001
African American	150	57.6%	28	43.7%	34	50.7%	38	61.3%	50	74.6%	.002
Education <12th grade	136	52.3%	27	42.2%	38	56.7%	32	51.6%	39	58.2%	.254
*Income <$12,000	128	51.8%	22	35.5%	31	49.2%	31	52.5%	44	69.8%	.002
Rural resident	136	52.3%	28	43.7%	36	53.7%	32	51.6%	40	59.7%	.335
^Transportation difficulty	53	20.4%	07	11.0%	12	18.0%	15	24.2%	19	28.3%	.012
^Depression (any)	31	11.9%	7	10.9%	9	13.4%	4	06.5%	11	16.4%	.353
^Anxiety (any)	65	25.0%	16	25.0%	14	20.9%	15	24.2%	20	29.9%	.147
	Mean	(SD)	Mean	(SD)	Mean	(SD)	Mean	(SD)	Mean	(SD)
Age	75.5	(06.3)	74.2	(05.8)	75.2	(05.9)	75.9	(06.8)	76.7	(06.6)	.147
Widowed years	15.5	(13.3)	02.7	(01.5)	08.2	(01.4)	16.0	(03.3)	34.6	(09.7)	<.001
+Life space	52.6	(22.9)	59.1	(23.9)	53.3	(22.5)	52.1	(21.5)	46.3	(22.1)	.015
^Self-reported health	03.3	(0.73)	03.3	(0.71)	03.3	(0.73)	03.2	(0.69)	03.4	(0.76)	.353
+Social isolation	03.2	(0.97)	03.2	(0.92)	03.4	(0.92)	03.2	(1.01)	03.0	(0.99)	.197
+MMSE	24.6	(04.5)	25.9	(03.4)	24.3	(04.4)	24.5	(05.1)	23.8	(04.8)	.052
^Comorbidities	03.0	(01.8)	03.2	(01.8)	02.9	(01.7)	03.2	(01.9)	02.9	(01.8)	.673
^GDS	02.4	(02.3)	02.2	(02.3)	02.5	(02.3)	02.3	(02.1)	02.7	(02.3)	.494

^*Missing income data changes the denominator:

All Widows n= 247; <1–5 years n = 62; 6–10 years n = 63; 11–22 years n=59; >22 years n=63

⁺Lower scores indicate worse health, function, or less favorable psychosocial circumstances

^{^}Higher scores indicate worse health, function, or less favorable psychosocial circumstances

TABLE 3.

Mean Time Since Spousal Loss in Years in Years

	N (%)	*Mean (SD)
All	N=260 (100.0)	15.5 (13.3)
AA female	N=114 (43.8)	20.3 (14.5)
W females	N=92 (35.4)	13.4 (12.3)
AA males	N=36 (13.8)	11.0 (08.6)
W males	N=18 (06.9)	05.7 (04.1)

^*P=<.001

Multivariate Associations

Tables 4 and 5 display the results of regression models, identifying variables with significant and independent associations with time-since-spousal loss. For each model, we first entered the sociodemographic variables followed by the addition of psychosocial and health/function variables. We tested a race/sex interaction term in the sociodemographic models and it was not significant.

TABLE 4.

Linear Regression Analyses for Sociodemographic Characteristics of Health and Function on Time Since Spousal Loss Among Older Adults

	Unstandardized Coefficients		Standardized Coefficients		95.0% CI	95.0% CI
	B	Std. Error	Beta	t	Lower	Upper	p value
Age	.023	.012	.126	1.86	−.001	.047	.065
Race	.523	.154	.234	3.47	.230	.853	.001
Sex	.848	.176	.306	4.81	.501	1.195	<.001
Education <12th grade	.051	.029	.140	1.78	−.006	.107	.077
*Income<$12,000	.365	.167	.162	2.19	.036	.694	.030
Rural/Urban resident	.254	.138	.113	1.83	−.019	.527	.068
Transportation difficulty	.332	.169	.125	1.97	.000	.665	.050
+Life Space Mobility	−.001	.004	−.028	−3.45	−.009	.006	.731
^Self-Reported Health	−.097	.107	−.063	−.905	−.309	.114	.366
+Social Isolation risk	−.141	.074	−.116	−1.91	−.286	.005	.058
^Comorbidities	−2.27	.040	.000	−.001	−.079	.079	1.00
+Cognitive function	−.028	.021	−.104	−1.37	−.069	.013	.176
^Depression	−.007	.035	−.014	−.199	−.076	.062	.843
^Anxiety	.029	.018	.110	1.62	−.006	.064	.107

⁺Lower scores indicate worse health, function, or less favorable psychosocial circumstances

^{^}Higher scores indicate worse health, function, or less favorable psychosocial circumstances

TABLE 5.

Logistic Regression Analyses for Sociodemographic Characteristics of Health and Function on Time Since Spousal Loss Among Older Adults

	Unstandardized Coefficients		Standardized Coefficients		95.0% CI	95.0% CI
	B	Std. Error	Beta	t	Lower	Upper	p value
Age	.023	.012	.126	1.856	.001	.047	.065
African American Race	.533	.154	.234	3.469	.230	.835	.001
Female Sex	.848	.176	.306	4.812	.501	1.195	<.001
Education <12th grade	.051	.029	.140	1.775	−.006	.107	.077
Income <$12,000	.365	.167	.162	2.186	.036	.694	.030
Rural/Urban resident	.254	.138	.113	1.833	−.019	.527	.068
^Transportation difficulty	.332	.169	.125	1.972	.000	.665	.050
+Life Space Mobility	−.001	.004	−.028	−.345	−.009	.006	.731
^Self-Reported Health	−.097	.107	−.063	−.905	−.309	.114	.366
+Social Isolation risk	−.141	.074	−.116	−1.908	−.286	.005	.058
^Comorbidities	−2.272	.040	.000	−.001	−.079	.079	1.000
+Cognitive function	.028	.021	−.104	−1.356	−.069	.013	.176
^Depression	−.007	.035	−.014	−.199	−.076	.062	.843
^Anxiety	.029	.018	.110	1.619	−.006	.064	.107

⁺Lower scores indicate worse health, function, or less favorable psychosocial circumstances

^{^}Higher scores indicate worse health, function, or less favorable psychosocial circumstances

In a multiple linear regression including only sociodemographic factors, race, sex, and age were associated significantly with time since loss. With the addition of health/function and psychosocial variables, age was no longer significant. Table 4 shows that African American race, female sex, and low income were significantly and independently related to time since spousal loss. Transportation difficulty was marginally significant (P=.05).

Table 5 displays the results of a multivariable logistic regression model examining sociodemographic, health/function and psychosocial predictors of long-term (>22 years) widowhood. In the sociodemographic model, age, race, and sex were significant, independent sociodemographic predictors of longer time since loss. With the addition of health/function and psychosocial variables, age was no longer significant. African American race, and female sex, lower income, and higher risk of social isolation risk were significantly, and independently associated with time since spousal death.

DISCUSSION

This study explores the concept of time since spousal loss among community-dwelling older adults living in the Southeast U.S. Socio-demographic characteristics had a significant association with the length of time that an individual resides in the status of widowhood. Controlling for age, women were significantly more likely than men to be widowed, with the odds of long-term widowhood 7 times greater for women than men. This sex-based patterning of the duration of widowhood occurred within a social and cultural context that further differentiated the widowhood experiences of men and women, giving rise to distinctive configurations of socio-cultural and psychosocial features. Traditionally, among women, widowhood is a firmly-held cultural expectation and a well defined social role, often lasting more than 20 years and typically shaping the final stage of the life course. For men, widowhood is more likely of shorter duration, often less than 10 years, and characterized by less well defined norms and fewer clearly-articulated expectations. Thus, the experience of widowhood is very different for women than for men, with women typically enduring longer exposure to widowed life and potentially suffering from greater limitations to personal autonomy imposed by social and cultural mores of appropriate behavior. This sex-based patterning of life after spousal loss also suggests that the social support needs of widowed men and women vary and that supportive services should be configured accordingly.

Because of the well documented gender variation in longevity in the U. S. (Newman & Brach, 2001), society has grown comfortable with the notion of extended widowhood for women. However, social conceptions of widowhood often are premised on the age factor, conjuring up images of frail elderly women living out the remainder of their days without the company of a spouse (Utz, Reidy, Carr, Nesse, & Wortman, 2004). Yet, the absence of a significant, independent association of age with time since spousal loss in this analysis challenges us to think beyond age when considering the time span of widowhood in society today. This is especially the case for people of color whose martial trajectories assume different shapes than those of Whites.

The odds of long-term widowhood were 3 times greater for African Americans than Whites, and regardless of age, African American women were living in the widowed state longer than White women, White men, or African American men. That can be attributed, in part, to longer life expectancy for African American women than African American men and to fewer available partners after spousal loss (LaVeist, Sellers, Brown, & Nickerson, 1997). For these reasons, it is important to consider the unique historical, social, and cultural context of aging and widowhood among women of color (Conway-Turner, 1999). Early marriage combined with early spousal death among African American populations may result in a cohort of long-term widows who do not fit the prototype of the mainstream older adult traditionally depicted in spousal loss literature (Rodgers, 2004).

In this analysis, continued exposure to the widowed state was longer for those with the lowest level of income, with the odds of long-term widowhood 2 times greater for those with low income compared to the more affluent. It is well documented that widowhood is associated with economic deficits, especially among women. Older women who depend on the economic resources of spouses are more likely to incur financial instability after the death of a spouse. However, the financial hardship of widowhood rarely abates over time, giving rise to persistent burden and exerting ongoing stress.

Others have previously described the risk of impoverishment among older African American women, in general, and widows, in particular (Wilson-Ford, 1991). The financial strain of long-term widowhood can be more pronounced for women of ethnic and racial minority status who may be less closely tied to the formal economy of wages and pensions. There is research suggesting that older African American women with lower household incomes are at higher risk for extreme social isolation (LaVeist, Sellers, Brown, & Nickerson, 1997).

Although maintaining independence is important to older widows, they often need to sustain reciprocal supportive relationships with both the immediate and extended family members (Jenkins, 2003). Cross-cultural studies have identified variations in the symbolic meaning of widowed status (Sudha, Suchindran, Mutran, Rajan, & Sarma, 2006) and increased awareness of the possibility that that long-term widows with decreased capacity to contribute to the needs of the extended family may incur a loss in social status and prestige and experience a decline in interaction and co-residence with offspring and other family members (Cattell, 2003; Glaser, Grundy, & Lynch, 2003; McNally, 2003).

Long long-term widowhood is associated with an array of social deficits with the potential to limit social engagement and social contact (Bennett, 2005). In this analysis long-term widowed persons had 1.5 greater odds of social isolation, a condition where contact with others is minimal, infrequent, or lacking in closeness. The issue of social isolation is a critical one because of the health promotion role of social support and social connectedness across the life course, in general, and among older women, particular (Shearer & Fleury, 2006). Although research on the impact of social support on a woman’s adjustment to recent widowhood is inconclusive, (Laditka & Laditka, 2003; Miller, Smerglia, & Bouchet, 2006), the health and wellbeing of long-term widows may be diminished by social isolation (Wegner, Davies, Shahtahmasebi, & Scott, 1996).

One potential contributing factor to social isolation is transportation difficulty. In this analysis the presence of transportation difficulty was marginally associated with time since spousal death. Those individuals who with the longest exposure to widowhood reported more difficulty acquiring needed transportation and greater difficulty in participating in activities due to lack of transportation. Among older adults, transportation difficulty has been shown to be related to marital status, financial issues, inadequate social support, poor physical and emotional health, mobility and/or cognitive limitations (Park et al., 2010) with consequences for access to important goods and services in the area of nutrition and health care (Hendy, Nelson, & Greco, 1998; Horton & Johnson, 2010).

While recent research found that social isolation among widowed persons does not vary by age, and that older widows are not more socially isolated than younger widows (Cornwell, Laumann, & Schumm, 2008), the work of Utz and colleagues (2002) points to qualitative differences between short-term and long-term widowed persons in the type of social contacts in which they engage.

Our research suggests that it is not age, but social processes associated with the length of widowhood among subsets of widows that render them more vulnerable to social isolation. While it is generally thought that that the biopsychosocial sequelae of spousal loss are a manifestation of instability in economic and social processes at work in the life of the recently widowed person (Dupre & Meadows, 2007; Zick & Smith, 1991), our findings suggest that it is a mistake to assume that such processes have re-stabilized in long-term widowhood even when the increased health risks of early widowhood have disappeared.

CONCLUSION

The garment of widowhood is not cut from one cloth. Rather is composed of many fabrics, reflecting the diversity and heterogeneity of the widowhood experience in society today and rendering it difficult to depict widowhood in a uniform way. An accurate characterization of the experience of widowhood among community-dwelling older adults must take into account variation in the length of time that individuals are living as widowed persons. Qualitative research is needed to explore contextual factors related to the lived experiences over time of widows (Porter, 1996), as well as the impact of long-term widowhood on the health and wellbeing of women. Such research can increase awareness of and appreciation for the diverse nature of widowhood, in general, and its ongoing salience for older adult, particularly for women of color (Capp-Taber, 2009). Additionally, more work is needed in the area of social processes that lead to social isolation in the long-term widowed women. We still do not have a full understanding of the complex processes through which ongoing exposure to widowed life impacts the depth and breadth of connectedness to social networks and hinders social functioning. Finally, there is an urgent requirement for the development, evaluation, dissemination of multimodal evidenced- based interventions to reduce social isolation among aging women, in general, and long-term widows, in particular (Findlay, 2003).

LIMITATIONS

Research is carried out within a context of available information. In our study, the scope of our analyses was limited by existing data. In the UAB Study of Aging, the marital status of participants was limited by the following categories: never married, currently married, separated, divorced, living as married, and widowed. “Living as married” was not reported by any participants, thus we have no insight on the dimensions of the loss involving a nontraditional partner. It is important to recognize that notions of spousal loss typically do not articulate the experience of disenfranchised widowhood (Doka, 2002; Doka, 1989; Shenk & Fullmer, 1996). It is important to note, as well, that while the findings reported here may reflect the ethnic/racial composition, social norms and cultural practices of the region of the Southeast United States, we believe the insights generated from this study have relevance to widowed adults residing in other locations.