Trajectories of Social Support and Well-Being Across the First Two Years of Widowhood

Abstract

The death of a spouse can be one of the most challenging events an older adult can face, yet many exhibit resilience. The present study examined the trajectories of structural and functional social support components, depression, and life satisfaction across the first 2 years of widowhood. The majority of structural and functional support trajectories exhibited stability across the first two years postloss. However, emotional support and support provided by family members did display a slight decline across time. Depression showed a linear pattern across time (e.g., decline in depressive symptomology) and life satisfaction demonstrated evidence of a 1-year anniversary effect.

The death of a spouse is a major stressor for many older adults. Currently, approximately 13 million people in the United States are widowed, and over 10 million of them are older adults (Janke, Nimrod, & Kleiber, 2008). With this staggering number, research is crucial in facilitating the growth of knowledge needed in this field to allow these older adults to receive the proper support and assistance they need. Overall, the bereavement process differs among individuals and can result in resiliency or a variety of grief outcomes (Bonanno et al., 2002). Even though researchers know losing a spouse in older adulthood is somewhat normative, there still exists a paucity in understanding the immediate postloss trajectories in well-being and social support. Potential factors that may aid older adults in achieving a resilient outcome include structural and functional social support variables. Even though losing a primary source of social support may be considered an aversive event, the majority of widows still manage to maintain normalcy within their lives, leading researchers to question how this process truly takes place (Holm, & Severinsson, 2012).

To understand how individuals adapt to traumatic events, the resiliency framework has been utilized throughout the bereavement field. Resiliency has been broadly defined as successful adaptation despite challenging or threatening circumstances (Masten, Best, & Garmezy, 1990). To date, many studies have looked at resiliency in widowhood and found that older adults can be highly resilient and successfully cope and adapt to the loss of a spouse (Bonanno et al., 2002; Ong, Bergeman, Bisconti, & Wallace, 2006). Resiliency can be broken down into internal (i.e., dispositional) factors, such as personality, and external (i.e., situational) factors, such as social support. Social support is one of the most commonly examined situational factors that influence the stress and coping process, and many researchers have found it to alleviate the deleterious effects of negative life events (Bisconti, Bergeman, & Boker, 2006).

In a prospective study of bereavement, Bonanno et al. (2002) examined depression levels both before and after the death of a spouse. By including preloss depression scores, Bonanno and colleagues (2002) were able to distinguish chronic grief from chronic depression and differentiate between stable but low depression levels or resilient patterns from widows who exhibited improved functioning. An important finding of this study was the emergence of five bereavement outcome trajectories including common grief, chronic grief, chronic depression, depression followed by improvement, and resilience. By tracking levels of depression up to 3 years before the loss and 6 and 18 months after the loss, there was supportive evidence that many bereaved individuals (45.6%) were capable of adapting to the loss and achieving overall resilience. Bonanno et al. (2002) also found significant group differences for instrumental support. More specifically, chronic grievers, depressed-improved, and chronically depressed participants had less instrumental support than that of the resilient participants. However, one of the limitations in the Bonanno et al. study was the relatively large gap in data collection periods. Given the dynamic nature of losing a spouse, it is important to examine outcomes from this stressor both closer to the time of the loss and in more rapid intervals (Bisconti, Bergeman, & Boker, 2004).

In all, social support has been examined as an exchange of resources between individuals, in which the social support is intended to enhance the well-being of the recipient (Shumaker & Brownell, 1984) and is expressed both structurally and functionally. In trying to understand the various facets of social support mechanisms, researchers have viewed social support as a meta-construct consisting of various subcomponents (Haber, Cohen, Lucas, & Baltes, 2007). Structural support includes the size of the social network or the frequency of contact with that network, whereas functional support includes types of emotional, instrumental, or informational support as well as the perceptions and judgments about that support. In an attempt to understand structural support characteristics of social networks across time during widowhood, Morgan, Neal, and Carder (1997) examined first-, second-, and third-year widowed women, and found that networks contained more family members than nonfamily members. Morgan et al. (1997) also found that nonfamily networks were significantly smaller for third-year widows than for first- and second-year widows. Ferraro and Barresi (1982) found support for stability in the interactions with family, friends, and neighbors reported by individuals widowed between 1 and 4 years. However, those widowed more than 4 years reported a decrease in the amount of family interaction.

Although structural components need to be examined, it is also beneficial to include functional support components, which are measures of our evaluations, rather than the network characteristics. The findings on the relationship between functional support components and adjustment to spousal loss in older adults have been conflicting (Norris & Murrell, 1990). More specifically, some researchers have found that individuals who have a high amount of perceived support tend to cope more effectively with life stressors and manage and anticipate various life challenges (e.g., bereavement) leading to proactive coping skills (Uchino, 2009). Conversely, in a study evaluating stressful life events (e.g., widowhood) and psychological distress, Lefrançios, Leclere, Hamel, and Gaulin (2000) examined, but found no support for perceived support as a buffering variable in older adults aged 81–86 years. Perhaps the critical component is not whether social support buffers the stress-outcome relationship, but whether there is a stable trajectory to be found in the first place (Dean, Matt, & Wood, 1992). Thus, an important addition to this literature would be to empirically test the trajectories of social support to evaluate the potential stability or change over time. Previous theoretical perspectives have typically proposed competing and varied effects of widowhood on both structural and functional support components, meaning widowhood will either lead to decreases in support, lead to stability in pre- and postloss levels of support, or an increase in an attempt to compensate for the loss of key supportive behaviors once provided by the late spouse (Dean et al., 1992). Originally, many researchers believed that with the loss of roles, widows might decrease their social participation (Rosow, 1973). However, more recent studies have found support of a compensatory and continuity model compared to a decremental model, suggesting that there may be both an increase in support in the beginning months as well as a level of stability in the social networks of older adults despite the increase in role loss. Therefore, a combination of both a compensatory model and resiliency framework will help guide the current study throughout the evaluation of social support trajectories across the first two years of widowhood.

The actual utilization of support for instrumental and emotional needs has also been included in many social support studies. Guiaux, Van Tilburg, and Van Groenou (2007) found that within the first two years of widowhood, there was an increase in both emotional support received and instrumental support received, which eventually leveled off and decreased after about 2.5 years. Other studies have found that when evaluating emotional versus instrumental support seeking behaviors, widows tend to benefit more from emotional support seeking (Bankoff, 1983; Bisconti et al., 2006). Moreover, in the case of spousal loss, where the situation is irreversible and the emotional ups and downs may prove to be too intense to effectively problem solve (e.g., instrumental support seeking behaviors), it may be more advantageous for widows to strictly engage in emotional support seeking behaviors in the beginning of their bereavement (Thoits, 1986). Additionally, previous studies have also shown widows to be inundated with support immediately following spousal loss, creating a compensatory perspective surrounding the bereavement process (Ferraro, 1984; Lopata, 1973). For example, Kohen (1983) found that following the loss of a spouse there was a dramatic and immediate increase in the probability of having an intimate friend to provide necessary types of support. Through these findings, it may be very beneficial for widows to engage in social activities and utilize their supportive networks after the loss of a crucial support figure. Moreover, although many widows display resilient outcomes (e.g., low levels of grief and depressive symptomology), it may be through the use of various social support components immediately following the loss that they are able to do so.

Another functional component that has been evaluated is the ability to understand if an individual perceives his or her network as available should he or she call on them (i.e., perceived control over support). When evaluating perceived control over supportive behaviors, researchers have found that the level of satisfaction and control over support tends to be very complex and can vary over the months following the loss of a spouse (Scott et al., 2007), which makes it essential to examine it at multiple time points. Kohen (1983) found that control over supportive behaviors might be influenced by the structure of relationships that was established prior to widowhood, instead of the value or helpfulness of that support. In other words, if a widow perceived that type of support as available and accessible, the quality of that relationship may not have weighed on her social support seeking behaviors. Instead there would be a focus on the amount of control the widow could exert on social interactions. However, currently there are few studies that explicitly address the notion of perceived control when evaluating social support. This gap in the social support literature creates a need to evaluate how widows not only perceive their social network’s quality, but also how reliably in control over supportive behaviors widows feel after the loss of a spouse and if that trajectory fluctuates or remains stable across time.

Examining well-being after the loss of a spouse is something that can only accurately be examined longitudinally. One of the most common measures that have been used to understand a widow or widower’s well-being is depression (Carnelley, Wortman, & Kessler, 1999; Norris & Murrell, 1990; Umberson, Wortman, & Kessler, 1992). In the beginning of bereavement research, it was common for practitioners to view people who did not grieve within a specific pattern as having some type of pathology (Wortman & Silver, 1989). The use of depression measurements has therefore been developed extensively to detect levels of psychological function and pathology within critical populations, such as widows and widowers. For example, Carnelley and colleagues (1999) looked at women’s vulnerability to depression after the loss of a spouse compared to stably married women. Findings from this study supported different trajectories of depression (e.g., more depression over 2 years postloss, stable levels of elevated depression), highlighting the need to differentiate between various outcomes such as chronic grief, chronic depression, and resilience.

Another popular measure of well-being both before and during bereavement is life satisfaction (Berg, Hoffman, Hassing, McClearn, & Johansson, 2009). Although satisfaction with life has typically been evaluated through cross sectional methodologies, it has been supported that life satisfaction may fluctuate with the loss of a spouse (Barer, 1994). There has been evidence of great individual differences in the extent of how much widowhood will impact perceptions of life satisfaction. Previous work has found bereavement to contribute to a steeper decline in life satisfaction, especially for men (Cheng & Chan, 2006). However, other studies have found the effect of widowhood on womens’ life satisfaction to be less negative (Berg et al., 2009). Therefore, understanding the trajectories of well-being components during the beginning years of bereavement is imperative to assisting older adults in coping with this loss.

The goal of the present study was to examine the nature of depression, life satisfaction, and social support across the first 2 years of widowhood. In understanding the multidimensionality of social support, we included multiple components of both structural and functional support, which is an improvement over many previous studies of bereavement that have been more unidimensional. Additionally, the first time point of data collection is approximately one month after the death of the spouse. Previous researchers have not collected data so soon after the death (Bonanno et al., 2002) leading to a gap in the literature in terms of understanding the relationship between support and well-being (Bisconti et al., 2006). Having data in the first months of the loss as well as up to the first two years will help elucidate the trajectories of crucial well-being and social support factors that may heavily influence the bereavement process. More importantly, being able to identify the levels of depression, life satisfaction, and social support of those individuals who are not resilient is critical, as setting up future research directions, interventions, health services, and counseling programs is dependent on accurately understanding the relationship between social support and well-being throughout the bereavement process.

Two major hypotheses drive the present study:

1. Depression and life satisfaction will show resilient patterns across the first 2 years of widowhood. More specifically, depression will decrease across time and life satisfaction will remain stable or gradually increase.

2. Due to the fluctuations of support given during the bereavement process, we predicted that widows would show an elevated amount of both functional and structural support components during the initial months of loss, but then would stabilize with the passage of time.

Method

Participants

Participants included 57 widows from the northern Indiana/southwest Michigan area. Key characteristics of the widow sample in this study can be viewed in Table 1. The widows ranged in age from 57 to 83 (M = 71.30; SD = 6.25) and were married from 14 to 63 years (M = 46.5; SD = 4.8). In regards to education, 39.7% of the widows completed high school, 13.8% had vocational education, 25.9% attended some college classes, 6.9% had a college degree, and 5.2% had a graduate, medical or law degree. Income levels were difficult to assess immediately following the loss and were therefore based off of the second time point (i.e., 4 months). Widows reported a range of income levels, with 1.7% earning less than $7,500, 20.7% earning $7,500 to $14,999, 39.7% earning $15,000 to $24,999, 8.6% earning $25,000 to $40,000, and 17.2% earning over $40,000. All women included in this study were Caucasian and 79.3% were widowed from their first marriage. Additionally, 37 of the women reported that they expected the loss (e.g., the husband had been ill or hospitalized prior to death) and 20 reported that the loss was unexpected.

Table 1.

Demographic information at Time 1

Respondent Characteristic	Mean	SD
Age (range = 57–83 years)	71.3	6.3
Length of marriage (range = 14–63 years)	46.5	4.8
	%
Education
Finished middle school	3.4
Completed high school	39.7
Vocational education	13.8
Some college classes	25.9
College degree	6.9
Post college professional degree	1.7
Graduate, medical, law degree	5.2
Income1
Less than $7,500	1.7
$7,500 to $14,999	20.7
$15,000 to $24,999	39.7
$25,000 to $40,000	8.6
Over $40,000	17.2
Caucasian	100.0
Marriage number
First	79.3
Second	19.0
Nature of the loss
Expected	63.8
Unexpected	34.5

Note. N_Time1= 44 to 58.

₁Income-levels were difficult to assess immediately following the loss; therefore the income levels are based off of data collected at 4-months postloss.

Data collection took place between 2000 and 2003 after the approval by the Institutional Review Board at the University of Notre Dame.¹ At the onset of the project, 361 recent widows of men 60 years of age or older were identified based on information obtained through newspaper death notices from a midsize northern Indiana city and surrounding areas (see Bisconti et al., 2004). Approximately 7 days after the death, a letter describing the purpose of the study was sent to the 266 widows when full address information was available; this letter was followed up with a phone call. There was correspondence with 211 of the 266 women (79%), including 19 widows and/or family members who declined before the follow-up phone call could be made; 121 individuals who declined or had a family member decline for them during the phone call (e.g., with the vast majority stating that they “just weren’t interested”); and 71 women who expressed interest. Of the 71 widows who initially accepted, 9 cancelled before the initial interview took place, resulting in a sample of 62 widows (29%) who participated in the initial interview. The current study utilized data obtained from widows that have substantial data at each of the seven collection points across the first two years (N = 57). Time 1 was collected at 1-month postloss, Time 2 was collected at 4-months postloss, and every subsequent time point there after was collected in 4-month intervals (i.e., T₃ = 8-months, T₄ = 12-months, T₅ = 16-months, T₆ = 20-months, T₇ = 24-months).

Materials

The measures included in this study were standard measurements of social support, perceived social control, problem-focused and emotion-focused social coping, depression, and life satisfaction.

Quantity and frequency of social support from family and friends

Eight items from a modified version of the Interview Schedule for Social Interaction (ISSI; Henderson, Duncan-Jones, Byrne, & Scott, 1980) were used to assess structural support (i.e., combined quantity and frequency) from family and friends. For example, quantity of support included, “How many people do you meet or talk to on the phone in a typical week?” Frequency of support included, “In a typical week, how many times do you meet or talk on the telephone to people with whom you do no live?” The response format for quantity of support was a five-part checklist ranging from nobody to 11 or more, which was assessed separately for family and for friends. For frequency of contact, the response format was a four-part checklist from never to once a day, which was also assessed separately for family and for friends. Higher scores on friend or family support indicated more support. The factor structure for this measure has been established using an older population (Bergeman, Plomin, Pedersen, McClearn, & Nesselroade, 1990; Bisconti & Bergeman, 1999). Cronbach’s alphas in the current sample include .75 for friend support and .77 for family support.

Perceived adequacy

Perceived adequacy questions immediately followed the quantity of support questions by stating after each “Are you satisfied with this number?” The perceived satisfaction response format consisted of “Would like more,” “Satisfied,” and “Would like fewer.” Because few people reported that they would like fewer support network members, these responses were rescored to be equal to “Would like more” and were labeled “Not satisfied.” Higher scores indicated more satisfaction with one’s support network. Cronbach’s alpha in the current sample was .92.

Perceived control for social support

Perceived control for social support was a subscale of the Desire for Control measure created by Reid and Ziegler (1981), designed specifically for older adults. The scale is a 13-item assessment based on a four-point scale ranging from 1 (strongly agree) to 4 (strongly disagree), which is used to rate the endorsement of with control-related statements. Examples of statements included, “I find that if I ask my family (or friends) to visit me, they come” and “I can rarely find people who will listen closely to me.” A higher score indicated more perceived social control for social support. In the current sample, Cronbach’s alpha was .86.

Coping

Coping was assessed using two subscales of the COPE (Carver, Scheier, & Weintraub, 1989), including seeking emotional support and seeking instrumental support. Seeking emotional support is a four-item assessment that is designed to measure socially oriented emotion-focused coping efforts. Sample items included, “I talk to someone about how I feel,” and “I get sympathy and understanding from someone.” Seeking instrumental support is also a four-item assessment designed to measure social aspects of problem-focused coping efforts. Sample items included, “I ask people who have had similar experiences what they did,” and “I try to get advice from someone about what to do.” The overall scale is based on a four-point response format (1 = I have not done this at all, 2 = I have done this a little bit, 3 = I have done this a medium amount, and 4 = I have done this a lot). A higher score indicated a greater amount of coping by way of seeking social support for either emotional or instrumental purposes. It should be noted that participants did not evaluate their coping efforts specifically in regard to their loss, but instead were instructed to, “Think about what you do and how you feel when you are under stress. For each question, check the box that best describes how you have responded to stress over the past two weeks.” In the present sample, Cronbach’s alpha was .66 for emotional support seeking and .85 for instrumental support seeking.

Depression

The CES-D scale was designed to assess depression, emotional distress, or both (Radloff, 1977). This scale consists of 20 items that represent a depressive symptom for which participants rate the frequency of occurrence during the past week; 16 items measure cognitive, affective, behavioral, and somatic symptoms associated with depression, whereas 4 items assess positive affect. Both reliability and validity have been well established (Devins & Orme, 1984). Responses are made on a 4-point scale that ranges from 1 (rarely or none of the time) to 4 (all or most of the time), with a higher score on the scale indicating an increased amount of distress. Sample statements include “I felt lonely,” “I had crying spells,” and “I thought my life had been a failure.” In a study (N = 232) from the same geographical region, Cronbach’s alpha was .88 (Bisconti & Bergeman, 1999). For the current study the Cronbach’s alpha was .82.

Life satisfaction

Life satisfaction was assessed by the Life Satisfaction Index-Form Z (Wood, Wylie, & Sheafor, 1969), which is a modified version of the Life Satisfaction Index-Form A (Neugarten, Havinghurst, & Tobin, 1961) consisting of 13-items assessing life satisfaction and morale. A 4-point Likert-type format was used to assess ratings of the participant’s life satisfaction on a scale of 1 (completely disagree) to 4 (completely agree). Items included statements such as “As I look back on my life, I am fairly well satisfied,” “These are the best years of my life.” And “I have made plans for things I will be doing a month or year from now.” Items are summed and a higher score indicates more satisfaction. In the current sample of widows, the Cronbach’s alpha was .84.

Analysis Plan

In order to assess longer-term change in regards to well-being and social support trajectories, multilinear modeling (MLM), and more specifically, growth curve modeling, was utilized to assess well-being across the first two years of widowhood. In the present study, SAS PROC MIXED V9.2 was used to analyze the data and run the growth curve analyses. Over the past couple decades MLM statistical procedures have advanced our knowledge about change variables within developmental research fields (Bryk & Raudenbush, 1987; Singer & Willett, 2003). When studying variables that develop and have the potential to change across time, the use of ordinary least squares regression models are insufficient (Singer & Willett, 2003). Assessing a critical population at one time point is an inaccurate portrayal of recovery after that life event, given that health and well-being varies after conjugal loss. Utilizing MLM allowed for a more flexible estimation procedure, as well as a dynamical assessment of well-being and multiple social support components.

Results

Descriptive Statistics

The following descriptive analyses were performed on the data collected at one month. For a complete list of means, standard deviations, and ranges for each measure, please refer to Table 2. The widows exhibited high levels of both family support (M = 24.66; SD = 4.30) and friend support (M = 23.26; SD = 4.13). Collectively, widows had a high combined score for perceived levels of support from both family and friends (M = 30.18; SD = 2.87) showing high levels of satisfaction with their social networks. Perceived control for social support levels were relatively high for the current sample (M = 38.29; SD = 4.40), supporting the widows perceived control beliefs about the readiness and availability of their social networks. Widows expressed a greater amount of emotional support seeking behaviors (M = 11.45; SD = 2.77) compared to instrumental support seeking behaviors (M = 8.39; SD = 3.09). At one month, widows expressed moderate levels of depression (M = 36.73; SD = 9.77) and on average high levels of life satisfaction (M = 36.50; SD = 5.54).

Table 2.

Properties of the Major Study Variables at One Month Postloss.

Variable	n	M	SD	Range
Family support	57	24.66	4.30	16.0 – 34.0
Friend support	57	23.26	4.12	16.0 – 32.0
Perceived amount of support from friends and family	57	30.18	2.87	18.1 – 48.0
Perceived control over support	57	38.29	4.40	26.0 – 48.0
Emotional support	55	11.45	2.77	7.0 – 16.0
Instrumental support	54	8.39	3.09	4.0 – 16.0
Depression	57	36.73	9.77	22.0 – 64.0
Life satisfaction	57	36.50	5.54	19.5 – 50.9

MLM Analyses

In order to test our hypotheses, we examined growth models to determine whether or not there was evidence of change in the trajectories of life satisfaction, depression, functional support components, and structural support components. Tables 3–5 show the fixed and random effects for the variable trajectories. Overall, a basic linear model was used for depression and all of the social support components, whereas a quadratic model was a better fit for the life satisfaction trajectories.

Table 3.

Unconditional Linear and Quadratic Functions of Time with Depression and Life Satisfaction

	Depression		Life Satisfaction
Variable	Linear Model		Quadratic Model
Fixed Effects	B	SE(B)	B	SE(B)
Intercept	35.66	1.39**	42.08	1.02**
Linear effect	−0.78	0.19**	−3.89	0.44**
Quadratic effect			0.56	0.05**
Random Effects	B	SE(B)	B	SE(B)
Intercept	70.70	19.51**	15.05	11.37
Linear effect	0.68	0.37**	−0.76	2.31
Quadratic effect			−0.004	0.03
Residual	37.37	3.36**	17.05	1.78**

Note.

^**p ≤ .01

Table 5.

Unconditional Linear Functions of Time with Functional Support Components.

	Perceived Control		Perceived Support		Emotional		Instrumental
Variable	Linear Model		Linear Model		Linear Model		Linear Model
Fixed Effects	B	SE(B)	B	SE(B)	B	SE(B)	B	SE(B)
Intercept	38.01	0.56**	30.12	0.34**	11.16	0.38**	8.57	0.38**
Linear effect	−0.07	0.08**	0.04	0.05	−0.26	0.06**	−0.06	0.06
Random Effects	B	SE(B)	B	SE(B)	B	SE(B)	B	SE(B)
Intercept	14.25	3.47**	4.07	1.27**	4.54	1.61**	4.23	1.62**
Linear effect	0.11	0.07	−0.01	0.03	0.02	0.04	0.00	0.04
Residual	4.93	0.47**	3.54	0.33**	4.36	0.42**	4.95	0.49**

Note.

^**p ≤ .01

Well-being variables

In the simple linear model for depression (See Table 3), there were significant fixed effects for both the intercept (β_0j = 35.7, p = .001), as well as a linear effect (β_1j = −0.78, p = .001), meaning average depression levels at baseline and over time varied (i.e., widows significantly varied in downward slopes over time). There were also significant random effects for the intercept (γ₀₀ = 70.70, p = .001) and marginally significant linear trends (γ₁₀ = 0.68, p = 0.06), meaning widows varied on their baseline starting points but exhibited the same rate of decline over time. In other words, we found that depression did in fact decrease across our 2-year time-span, which is consistent with our first hypothesis.

In the quadratic model for life satisfaction (See Table 3), there were significant fixed effects for the intercept (β_0j = 42.08, p =.001), linear effect (β_1j = −3.89, p = .001), and curvilinear effect (β_2j = 0.56, p = .001). Unlike the depression results, life satisfaction exhibited a significant curvilinear effect that best fit the data, meaning there was evidence of a curvature in the overall means for our sample of widows. Instead of exhibiting stability or a gradual increase in life satisfaction like we originally predicted, widows started high on life satisfaction and reached an ultimate low at about one year after the loss and then eventually increased in satisfaction during the second year of widowhood. There were no significant random effects, meaning one overall curvature model best described all of the widows’ patterns of life satisfaction across the two years. In other words, life satisfaction actually decreased across the first year of widowhood, hitting an all-time low at the 1-year anniversary, before increasing in the second year after death. Clearly, this did not support our resilience hypothesis.

Functional Support Variables

In the simple linear model (See Table 4), we found significant fixed effects for the intercept for perceived support (β_0j = 30.02, p = .001), perceived control over support (β_0j = 38.01, p = .001), and instrumental support (β_0j = 8.57, p = .001), meaning average levels at baseline varied (i.e., widows significantly varied on start points for levels of functional support). However, there were no significant fixed effects for the linear effect, meaning there was evidence of stability across time for 3 out of the 4 functional support measures. We found that even though there were significantly different start points for the widows on measures of perceived support, perceived control over support, and instrumental support, the widows exhibited the same slope across time (i.e., stability) and no variance in their trajectories. Moreover, emotional support actually exhibited significant fixed effects for both the intercept (β_0j = 11.16, p = .001) and linear effect (β_1j = −0.26, p = .001), meaning widows reported a slight linear decline in emotional support across the first 2 years of bereavement. Although the widows exhibited a slight decline in emotional support across time, the widows showed evidence of the same slope pattern and did not display variance in their trajectories over time. Therefore, our findings on functional support partially support our second hypothesis, as many of our components exhibited stability over time. However, we did not find an all-time high in the beginning of bereavement as we originally predicted. Instead we found stability from the beginning for perceived support, perceived control over support, and instrumental support, and an overall decline in emotional support provided to the widows over time.

Table 4.

Unconditional Linear Functions of Time with Structural Support Components.

	Support From Family		Support from Friends
Variable	Linear Model		Linear Model
Fixed Effects	B	SE(B)	B	SE(B)
Intercept	24.71	0.60**	23.23	0.56**
Linear effect	−0.31	0.09**	−0.12	0.08
Random Effects	B	SE(B)	B	SE(B)
Intercept	17.44	3.95**	15.16	3.50**
Linear effect	0.28	0.09**	0.16	0.07*
Residual	4.37	0.42**	4.46	0.42**

Note.

^**p ≤ .01 and

^*p ≤ .05

Structural Support Variables

In the simple linear model for structural support components (See Table 5), we found a significant fixed effect for the intercept for support provided by friends (β_0j = 23.23, p = .001), whereas support provided by family exhibited significant fixed effects for the intercept (β_0j = 24.71, p = .001) and linear effect (β_1j = −0.31, p = .001). Therefore, support provided by friends remained stable across the first 2 years of widowhood, whereas support provided by family actually declined over time. This partially supports our second hypothesis, as we believed structural components of support would show a gradual decline across time, as social networks potentially grow smaller. However, like functional components, there was not an apparent elevation in support around the beginning months postloss. Instead, support from friends was stable across time and did not fluctuate, whereas support from family members slightly declined across time.

Discussion

We examined the role of social support, life satisfaction, and depression in a sample of recently bereaved widows over a two-year period, putting forth two separate hypotheses. First, we predicted that our outcomes would yield a resilient pattern across our 4-month interval data. Our findings partially supported this prediction as depression showed a downward trend, where widows decreased in depression levels across time. However, life satisfaction did not exhibit a resilient trajectory (i.e., stable across time) and instead showed a gradual decline across time, resulting in an all time low around the one year mark and gradually increasing across the second year of bereavement. Second, we predicted that both functional and structural components would show an elevation around the beginning of bereavement and eventually become stable with the passage of time. Typically widows are inundated with multiple types of support surrounding the first months of the loss, but this support may dissipate as supportive individuals move on and subjectively perceive the widow to be adjusted with her new single lifestyle. However our results show that widows may maintain stable levels of support across the first two years of widowhood both structurally and functionally. The only two support components that exhibited a decline across time were emotional support and the amount of support provided by family members. Therefore, instead of supporting a compensatory model of support during the beginning months of bereavement, our sample displayed stability in many social support components as soon as one-month postloss.

Based on previous resiliency literature, widows may typically demonstrate signs of resilience, as the loss of a spouse becomes a somewhat predictable normative life event for older adult women (Bonanno et al., 2002). However, it is interesting to note that our well-being components differed in terms of the best-fit model (e.g., linear versus quadratic trend). Depression simply had an overall linear downward trend over time (e.g., levels of depression decreased over time); whereas life satisfaction had curvature over time (e.g., levels of life satisfaction decreased, leveled off, and eventually increased over time). One explanation for this finding is the clear distinction between measurements of depression and life satisfaction (Kemp, Adams, & Campbell, 1997; O’Rourke, 2004). Typically when measuring depression, items usually tap into emotional states (e.g., “I felt lonely”), whereas life satisfaction items evaluate global traits of one’s life that may not directly tap into emotion (e.g., “These are the best years of my life”). The overall linear decline of depression levels reveal that widows’ emotional states at the time of measurement were clearly improving in terms of suffering from less depressive symptoms. However, life satisfaction levels may not have been affected by the loss immediately, but were rather negatively affected as time progressed and neared the anniversary of the death. As coping with the loss and adjusting to a new lifestyle became easier, levels of life satisfaction eventually increased back to baseline levels. It is also important to point out that life satisfaction still declined despite stability in many of the social support components. However, the slight declines in emotional support and support from family members may have also influenced the trajectories of life satisfaction in our sample. Future studies may look into these bereavement components further as examining predictive relationships may provide a better understanding of the widowhood process.

Another explanation for the life satisfaction trajectories includes evaluating anniversary effects. Anniversary effects are typically related to traumatic events and may be triggered by environmental cues (e.g., sight, sounds, smells), specific memories, emotional reactions or even subsequent temporary upsurges of grief (Rando, 1993). In an attempt to develop a model to examine anniversary effects, Chow (2010) suggested that there may be both declines in well-being around the time of the death as well as before, creating anticipatory anniversary effects. This may help explain the decrease in life satisfaction prior to the anniversary of the death in the current study, as widows may have been reminded of the death more often and may have exhibited lower levels of life satisfaction prior to the first death anniversary. In regards to longitudinal effects, Carnelley, Wortman, Bolger, and Burke (2006) even found that in a national sample of widows, reactions to the anniversary of the death were common for up to 8 years postloss. Therefore, by developing programs and services that fit the needs of widows subject to anniversary effects, helping professionals may target key points within the bereavement period (Elwert & Christakis, 2008). Formal services may choose to target critical time frames during a widow’s bereavement and offer services such as interventions or remembrance programs around the anniversary of the death, birthdays, holidays, or anniversaries to guide the widow back to positive levels of well-being (Chow, 2010).

In terms of our social support variables, our findings support previous literature in that the stability and change in social support trajectories truly demonstrate dynamic outcomes that may vary by the type of support provided (e.g., instrumental or emotional) as well as the source of that support (e.g., friend or family member). Based on the resiliency literature, we postulated that in order to achieve resilient outcomes, social support may serve as a critical component during the beginning months to maintain stability. In other words, widows may maintain a stable level of well-being due to an increase in supportive behaviors from friends and family members. However, based on our findings, the loss of a spouse did not lead to a total disruption in supportive relationships or an actual superficial increase in the beginning of widowhood, and instead supported a model of continuity (Atchley, 1971). Majority of our support components exhibited stability across the first two years of widowhood, although it is also important to point out that there was evidence of decline in emotional support as well as the amount of support provided by family members. Dean and colleagues (1992) posit that although trajectories of support tend to be stable, there may still be variations in the type of support (e.g., functional versus structural) and source of that support over time. For example, in our study, widows may have utilized support from friends more frequently than relying on family members, which could potentially explain the decline in family support across time. The need for emotional support may have also subsided as time progressed from the beginning months after the loss. Therefore, both the stability and instability of various social support components need to be further examined as support during bereavement may have beneficial or adverse effects on a widow’s well-being.

Even though the results yielded interesting findings, there are several limitations to this study. First, we studied a very homogeneous sample, having all Caucasian women from the same mid-west region of the United States. This therefore limits the study’s generalizability to widows in other areas of the country, as well as widows belonging to different racial and ethnic backgrounds. Previous researchers have found that there are many cultural differences, including a variety of ethnic practices and beliefs that are unique to different cultural backgrounds (Eisenbruch, 1984). A second limitation to this study is the absence of widowers. Researchers have found widowhood to differentially affect men and women, with women reporting higher levels of instrumental support and less strain in relationships with their children, whereas men report more informal social integration and higher levels of strain with household concerns (Umberson et al., 1992). Interestingly, researchers have also found that in general older adult women report more depressive symptoms than older adult men, but have found these gender effects to be reversed in widowhood, that is, depression levels are highest amongst widowed men (Bennett, Smith, & Hughes, 2005). A third limitation is that the study did not have any preloss data to properly assess depression and social support before the death. Not knowing depression and social support levels before the loss greatly limits our understanding of the psychological, emotional, and physical ups and downs that appears after the death. Having prospective data is crucial in truly understanding the trajectories of well-being during widowhood. Previous researchers have found that utilizing pre-bereavement data leads to a clearer picture of the widowhood process (Carnelley et al., 1999).

Even with the aforementioned limitations, one of the strongest contributions of the present study is through its use of longitudinal data. Although we know that understanding how individuals change after the loss of a spouse is critically important, it is still difficult to conduct studies across time. This longitudinal approach allowed us to help elucidate two phenomena. One, we were able to examine the true trajectory of change using growth curve modeling in two different traditional outcome measures, examining perhaps the difference between more state- like measures of well-being and more global trait-like measures. This allowed us to empirically capture the anniversary effect at 1-year postloss. And two, we were able to examine the trajectories of both structural and functional components of social support across time. It is rare that patterns of support are examined singularly and modeled across time, and has instead been used as important predictors or mediators of well-being. However, trajectories of social support enable researchers to evaluate the stability or fluctuation of support provided to a widowed individual, along with their perceptions and satisfactions with that support. This may provide significant implications for how we conceptualize treatment for this population, as many interventions and support groups target decline in a widow’s social network instead of stability.

In a systematic review of bereavement care interventions, Forte, Hill, Pazder, and Feudtner (2004) found that there is still no consistent pattern of treatment used to combat emotions, behaviors, or levels of grief pertaining to bereavement except for the pharmacologic treatment of depression. The development of evidence-based bereavement interventions is still in its infancy stage, needing replicated published studies, sound study designs, adequate reporting of intervention procedures, and consistent theoretical backgrounds (Forte et al., 2004). By no means are the current researchers stating that this study is generalizable to other populations, but rather is a step in the right direction when evaluating samples that would be prone to attending support groups or bereavement interventions.

Although many stigmas and stereotypes exist about how one should showcase their grief or express their emotions and behaviors, many individuals display resilient reactions after an aversive event like conjugal loss (Bonanno et al., 2002; Bonanno, Wortman, & Nesse, 2004). Throughout the widowhood literature, we know that social support is only one focal point after the loss of a spouse. However, it is necessary to understand, as an individual’s ability to adapt to such an aversive life event may be dependent upon social interactions and social dependability, ultimately affecting a potentially resilient outcome. With this in mind, the resiliency framework may be of future use when assessing widows longitudinally after the death of a spouse, as it may be the stability of various facets of bereavement that aid in the maintenance of resilience after loss.