Spouse Health Status, Depressed Affect, and Resilience in Mid and Late Life: A Longitudinal Study. Jamila Bookwala.

Transcription

1 Spouse Health Status, Depressed Affect, and Resilience in Mid and Late Life: A Longitudinal Study Jamila Bookwala Lafayette College Keywords: spousal illness, depressive symptoms, resilience, mastery, self-esteem, mid and late life Published in Developmental Psychology 50 (4): April DOI /a

2 Abstract This study used longitudinal data to examine the effects of spousal illness on depressive symptoms among middle-aged and older married individuals and the extent to which the adverse effects of illness in a spouse were mitigated by two psychological resources, mastery and self-esteem. Using 1,704 married participants who were 51 years of age on average, depressive symptoms were compared in four groups varying in their experience of spousal health transitions: those whose spouse remained ill at T1 and T2, those whose spouse declined in health from T1 to T2, those whose spouse s health improved from T1 to T2, and those whose spouse remained healthy at both time points. Mixed ANCOVA analyses showed that, as hypothesized, having a spouse who became or remained ill over time was linked to greater depressed affect by T2 whereas having a spouse improve in health was associated with a decline in depressive symptomatology. Moderated regression analyses indicated that while higher mastery and self-esteem were linked to lower depressed affect in general, these resources were especially protective against depressed affect for those whose spouse remained ill at both time points. These findings are at the intersection of life course theory and the stress process model highlighting the contextual forces in and the interconnectedness of individual development as well as the plasticity and resilience evident in adaptation to stress during mid and late life.

3 Spouse Health Status, Depressed Affect, and Resilience in Mid and Late Life: A Longitudinal Study Life course theory emphasizes that the meaning and relevance of life events and transitions such as those involving a spouse s illness are contingent upon their timing and sequencing within the developmental context and the extent to which they involve those connected to the individual s life (Baltes, 1987; Elder, 1998; Fuller-Iglesias, Sellars, and Antonucci, 2008). As such, the relevance for individual development of a life event such as illness in a spouse reflects two key principles of life course theory. First, the principle of timing and sequencing recognizes that the impact of a spouse s ill health will depend on one s own specific location in place and time in the life course when the illness occurs. Second, the principle of linked lives emphasizes that individual development results in and is the result of changes in the lives of those with whom one is interpersonally connected. Pearlin (2010) recently has highlighted the considerable overlap between these tenets of life course theory and the stress process model. He explains that the effects of stressful transitions depend on the age when they occur (the principle of timing and sequencing). Whereas non-normative events may be especially salient to well-being, it is plausible that poor health in a spouse in mid or late life may be detrimental despite the relatively normative nature of health-related changes as people age. These negative effects are likley because individuals also are typically experiencing their own aging then and because poor health in the spouse will likely change marital roles and responsibilities and the quality of the marital relationship. Pearlin further explains that while stressors in the lives of others are important sources of stress for individuals (the principle of linked lives), not all linked lives are equal. Stressors that occur within our closest role relationships have the potential to be the most detrimental to well-being. Inasmuch as the spousal relationship is perhaps the most significant adult relationship (Bookwala, 2012), poor health status of a spouse during the middle and late adulthood years is likely to be a significant life stressor and can pose serious challenges for individuals. Thus, we can expect that poor spousal health is likely to be associated with greater depressed affect among middle-aged and older individuals. Supporting evidence shows that

4 poor spousal health during the mid and late life years increases the risk of mental health or substance abuse disorders (Zivin and Christakis, 2007) and emotional loneliness (Korporaal, van Groenou, & Tilburg, 2008). Of course, both life course and stress process perspectives would predict that an improvement in spousal health would result in a corresponding improvement in psychological well-being. While this transition has not received much empirical attention to date, the current study includes a group of middle-aged and older individuals whose spouse s health improved over time in order to examine if there occurs a corresponding decline in their level of depressed affect. The stressful nature of poor spousal health notwithstanding, individuals can show considerable resilience in the event of life stressors (Pearlin, 2010; Ryff & Singer, 2003). Resilience, which refers to successful adaptation to adversity or stress (Fuller-Iglesias et al., 2008; Ryff & Singer, 2003; Ong, Bergeman, & Boker, 2009) can result from a variety of resource variables. Two common psychological or personal resources that can explain inter-individual variability in adaptation to life stressors are levels of personal mastery and self-esteem (Thoits, 2010; Windle, Markland, & Woods, 2008). Research specifically with middle-aged and older adults has confirmed that resilient outcomes in the face of diverse stressors (e.g., job loss, death of a loved one, a long-standing illness) are especially likely when these individuals rely on individual resources such as a sense of mastery (e.g., Harmell, Chattillion, Roepke, & Mausbach, 2011; Hildon, Smith, Netuveli, & Blane, 2008; Moen, Sweet, & Hill, 2010) and high self-esteem (e.g., Lee, Brown, Mitchell, & Schiraldi, 2008; Sargent-Cox, Anstey, & Luszcz, 2012). Employing life course theory and the stress process as guiding frameworks, the present study pursued two goals using longitudinal data from a probability-based sample of middle-aged and older adults who participated in the National Survey of Families and Households (Sweet & Bumpass, 2002). First, using two waves of data, it compared groups on depressed affect based on four different types of experience with spousal illness: having a spouse who was ill at both time points, having a spouse whose health declined (i.e., the spouse had an illness at follow-up but not at baseline), having one whose health improved (i.e., the

5 spouse had an illness at baseline but not at follow-up), and having a spouse who was healthy at both time points. Respondents whose spouse remained ill or became ill from T1 to T2 were hypothesized to experience greater depressive symptomatology than those whose spouse remained healthy. Moreover, those whose spouse s health improved from T1 to T2 were expected to show a rebound in psychological wellbeing, experiencing a decline in depressive symptoms. Second, the study evaluated the effectiveness of personal mastery and self-esteem in promoting resilience or positive adaptation (i.e., lower depressed affect) across the different study groups. Higher mastery and self-esteem were hypothesized to be especially salient to the psychological well-being of individuals with an ill spouse. Method Sample Participants included middle-aged and older married participants from the National Survey of Families and Households (Sweet & Bumpass, 2002). Eligible cases included those who were at least 40 years of age in the wave of data collection (T1) and had no missing data at T1 or the subsequent wave of data collection (T2; ). The final sample included 1,704 married elders averaging 51.3 years of age at T1 (SD=9.6; range=40-86 years), 46.7% of whom were male (n=795). The majority of the sample had a high school diploma (n=1,511, 88.7%), were employed at T1 (n=1,294, 75.9%) and T2 (n=1,055, 61.9%), and self-identified as White (n=1,509, 88.6%). Overall, respondents rated their health favorably at T1 (M=4.05, SD=0.8) with 81.9% (n=1,396) rating their own health as good or excellent. Although mean self-rated health showed no change from T1 to T2 on average (M=4.04, SD=0.9), slightly fewer respondents rated their health as good or excellent at T2 (77.9%, n=1,328). Spousal health status at T1 and T2 was determined based on participant responses to an item that asked if their spouse currently had any long-term physical or mental condition or disability that either limited or could limit her or his activities. Respondents who indicated that their spouse had such a long-term health condition at both baseline and follow-up constituted the ill-spouse group (n=315), those who answered no

6 at T1 and yes at T2 were categorized as the spouse-declined group (n=210); those who answered yes at T1 and no at T2 formed the spouse-improved group (n=309) and respondents who reported that their spouse was not ill at either wave formed the healthy-spouse group (n=870). The most common health conditions reported by the two groups whose spouse had an illness or disability at T1 (the ill-spouse and spouse-improved groups) included cardiovascular disease, arthritis and other musculoskeletal disorders, diabetes, cancer, asthma or other chronic respiratory tract diseases, and chronic hearing problems (81.8%). Co-morbid conditions when reported for the spouse at T1 most frequently included cardiovascular disease and arthritis and related diseases (53.6%). The most frequent health conditions reported for the spouse at T2 by the two relevant groups (the ill-spouse and spouse-declined groups) included arthritis and other musculoskeletal or connective tissue disorders, cardiovascular disease, diabetes, asthma or other chronic respiratory diseases, cancer, neurological or psychiatric disorders, sensory impairments, and some residual chronic health condition (79.7%). In addition, respondents who reported a second co-morbid condition for the spouse at T2 most commonly reported cardiovascular disease, arthritis and related diseases, and diabetes (59.2%). Measures Depressive symptomatology was assessed using the 12-item version (Kessler et al., 1992) of the original Center for Epidemiological Studies Depression scale (CES-D; Radloff, 1977). Sample items included feeling sad, lonely, and that everything you did was an effort. The NSFH used an 8-category set of response options to measure the number of days out of the past seven days (0-7) that each depressive symptom occurred. Item scores were summed with higher scores reflecting greater depressive symptomatology. Cronbach s alpha obtained for this 12-item CES-D was.93 and.89 at T1 and T2, respectively. Sample means for depressive symptomatology were (SD=12.5) at T1 and (SD=12.4) at T2.

7 Mastery was measured in the NSFH via four items from Pearlin and Schooler s (1978) personal mastery scale. Respondents rated the items using a 5-point scale ranging from 1=strongly agree to 5=strongly disagree; sample items included I can do just about anything I really set my mind to do and I have little control over the things that happen to me. Items were rescored as needed so that higher scores represented higher mastery and mean scores were computed. Because Cronbach s alpha is a function of scale length and only a shortened 4-item version of the original Pearlin et al. scale was included in the NSFH, Cronbach s alpha was computed in two ways for the short version and using a correction suggested by Nunnally (1978) to adjust for scale length. Cronbach s alpha for the 4-item scale was.65 at T1 and.56 at T2; this value increased to.79 and.72 if the scale was adjusted to be twice as long (Nunnally, 1978). Mean mastery scores were 3.67 at T1 and 3.68 at T2 (SD=0.7 and 0.6, respectively). Self-esteem at T1 and T2 was assessed via three items from the Rosenberg (1965) self-esteem scale (e.g., being a person of worth at least on an equal plane with others). Respondents rated items using a 5-point scale ranging from 1=strongly agree to 5=strongly disagree. Items were rescored so that higher scores represented higher self-esteem and mean scores were computed. Cronbach s alpha again was computed in two ways, in raw form as well as adjusting for the number of items included in the NSFH. For the brief 3- item scale, Cronbach s alpha was.66 at T1 and.64 at T2; if adjusted for length at twice as many items using Nunnally s correction (1978), the alpha would increase to.80 at T1 and.78 at T2. Mean scores on selfesteem were 4.12 and 4.05 at T1 and T2, respectively (SD=0.6 and 0.5). Other variables included in the analyses were respondents age, racial background (White vs. other), educational level (high school diploma vs. not), employment status, and self-rated health (assessed on a 5- point scale ranging from very poor to excellent); these variables were treated as statistical covariates. Because studies have shown that gender differences exist in the impact of spousal health on individuals psychological well-being (Walker & Luscz, 2009), group comparisons on depressed affect were further stratified by gender. Marital happiness, used in follow-up analyses to test the robustness of group

8 differences in depressed affect, was assessed with an 8-item measure; respondents indicated their level of happiness (1=very unhappy to 7=very happy) on different dimensions of their marital relationship with higher scores representing greater marital happiness. A Cronbach s alpha for marital happiness of.90 was obtained at both time points. Mean scores on marital happiness at T1 and T2 respectively were 5.56 (SD=1.2) and 5.83 (SD=1.1). Results Differences by Spousal Health Group A three-way mixed analysis of covariance (ANCOVA) was performed on depressive symptomatology using as factors spousal health (between-subjects factor; four levels: healthy-spouse group, spouse-improved group, spouse-declined group, ill-spouse group), gender (between-subjects factor), and time (repeated measures factor; two-levels: T1 and T2); covariates included age, racial background, education level, employment status, and self-rated health. Means and standard deviations obtained in the analysis are shown in Table 1. The results showed a significant main effect for spousal health, F(3,1689)=13.53, p<.001, η 2 =.02. Overall, the ill-spouse and spouse-declined groups scored significantly higher on depressive symptomatology (M=12.82 and 11.11, respectively) than the healthy-spouse group (M=8.95). The spouseimproved group (M=9.60) also scored significantly lower on depressive symptoms than the ill-spouse group. Notably, the spouse-improved and healthy-spouse groups did not vary significantly from each other, nor did the ill-spouse and spouse-declined groups. More central to the study hypothesis, the mixed ANCOVA showed a significant time x spousal health interaction, F(3,1689)=3.03, p<.05, η 2 =.01. Two types of simple effects tests assessed between-group and within-group differences in follow-up analyses (see Figure 1). Between-group comparisons showed that, at T1, depressive symptoms varied significantly by spousal health group, F[3,1692]=8.20, p<.001, η 2 =.01 with the ill-spouse group scoring significantly higher on depressive symptomatology (M=12.76) than the healthyspouse group (M=8.97). At T2, more differences emerged across spousal health groups, F[3,1692]=11.36,

9 p<.001, η 2 =.02. As expected, respondents in the spouse-improved group scored significantly lower on depressive symptoms at T2 (M=8.50) than those in the ill-spouse and spouse-declined groups (M=12.98 and 11.66, respectively); the latter two groups did not differ significantly. The spouse-declined and ill-spouse groups, also as hypothesized, reported significantly more depressive symptoms at T2 than the healthyspouse group (M=9.15). Within-group analyses showed that, as expected, the spouse-improved group reported a significant decline in depressive symptomatology from T1 to T2, F(1,300)=3.96, p<.05, η 2 =.01 (see Figure 1). In comparison, no significant changes were seen over time in the remaining three groups: the ill-spouse group and the spouse-declined group trended toward an increase in depressive symptoms from T1 to T2 but these increases were not statistically significant; the healthy-spouse group remained stable, scoring at the low end at T1 and T2. In addition to the significant main effect for spousal health and interaction effect of spousal health x time, the overall three-way mixed ANCOVA showed the expected main effect for participant gender on depressive symptoms (F[1,1689]=34.45, p<.001, η 2 =.02), with women scoring significantly higher on depressive symptomatology than men (M=12.15 and 9.09, respectively). No significant effects were obtained for gender x time, gender x spousal health, and the three-way interaction of spousal health x gender x time; the effect for time was significant but small (F[1,1689]=6.00, p<.05,η 2 <.01). To ensure that the obtained differences based on spousal health and time x spousal health that are described above were not accounted for by differences in marital happiness, the three-way mixed ANCOVA on depressive symptoms was repeated with the addition of T1 and T2 marital happiness scores as statistical covariates. The pattern of findings remained unchanged; the main effect for spousal health and its interaction with time on depressive symptoms remained significant even after adjusting for differences in respondents marital happiness. Psychological Resources as Moderating Variables

10 The second goal of this study was to examine the extent to which mastery and self-esteem moderated the relationship between poor spousal health status and depressive symptoms. In a preliminary step, mixed ANCOVAs parallel to those above were conducted to determine if women and men in the four spousal health groups varied in their mean levels of mastery and self-esteem over time (see Table 1 for means and standard deviations). These analyses revealed no significant spousal health x time interaction for mastery (F[3,1689]=1.93, p>.12) or self-esteem (F[3,1689]=1.68, p>.17). A significant main effect of spousal health was found for mastery (F[3,1689]=8.73, p<.001, η 2 =.02) with the spouse-improved (M=3.71) and the healthy-spouse (M=3.74) groups scoring higher on mastery than the ill-spouse group (M=3.55). A significant main effect of time was seen for self-esteem (F[1,1689]=9.33, p<.01, η 2 =.01) with self-esteem decreasing slightly over time (4.11 vs. 4.06). Finally, a significant main effect of gender was found for mastery (F[1,1689]=18.17, p<.001, η 2 =.01) and self-esteem (F[1,1689]=24.79, p<.001, η 2 =.01). Women scored lower than men on both mastery (M=3.59 vs. 3.72, respectively) and self-esteem (M=4.02 vs. 4.15, respectively). Next, moderated regression analyses were performed to assess the extent to which self-esteem and mastery mitigated the impact of spousal illness on depressive symptoms. This was accomplished by a series of moderated regression models in which a) depressive symptomatology at T1 was regressed separately on T1 mastery and self-esteem and b) depressive symptomatology at T2 was regressed on T2 mastery and selfesteem while adjusting for T1 depressive symptomatology. For all moderated regression models, a) mastery and self-esteem scores at each time point were centered around their mean; b) three dummy variables were created for the spousal health variable with the ill-spouse group treated as the reference group; c) interaction terms were computed between wave-specific centered scores on mastery and self-esteem and each of the three dummy variables for spousal health groups; and d) control variables were entered on the first step (age, gender, education, ethnicity, employment status, and self-rated health).

11 Mastery as a Moderator: As shown in Table 2, T1 mastery did not moderate the link between spousal health transitions and T1 depressive symptoms on step 3. A similar model for T2 depressive symptomatology, however, showed support for the moderating effects of mastery (see Table 2). After accounting for the covariates on step 1 (R 2 =.21) and spousal health transitions and T2 mastery on step 2 (ΔR 2 =.07), the introduction of the interaction terms between the spousal health groups and T2 mastery on step 3 significantly improved the model ( F[3,1687]=3.80, p=.01, ΔR 2 =.01). The interaction terms involving the healthy-spouse group (β=.10, t[1687]=2.70, p<.01) and spouse-improved group (β=.09, t[1687]=3.14, p<.01) emerged as statistically significant indicating that the salience of mastery in reducing depressive symptomatology was significantly different for these two groups than for the ill-spouse group. Follow-up regression analyses conducted separately for the ill-spouse, healthy-spouse, and spouse-improved groups indicated that while higher mastery was significantly associated with lower depressive symptomatology for all three groups, its protective effect was greatest in the ill-spouse group. A comparison across models showed that higher mastery was more strongly linked to lower depressive symptomatology in the ill-spouse group (ΔR 2 =.091; β=-.34, t[304]=-6.22, p<.001) than in the healthy-spouse group (ΔR 2 =.057; β=-.25, t[869]=-7.96, p<.001) and spouse-improved group (ΔR 2 =.03; β=-.19, t[298]=- 3.55, p<.001). Self-esteem as a Moderator: The moderated regression analysis using T1 self-esteem also showed similar patterns as seen with mastery (see Table 3). Whereas T1 self-esteem did not moderate the link between spousal health transitions and T1 depressive symptoms on step 3, the moderated regression analysis with T2 self-esteem showed support for the moderation model. In this model, the introduction on step 2 of spousal health transitions and T2 self-esteem explained 4% of the variance in depressive symptoms. Then, adding the interaction terms between spousal health transitions and T2 self-esteem on step 3 significantly improved the overall model F[3,1687]=3.65, p<.05, ΔR 2 =.01) with the interaction terms involving the healthy-spouse and spouse-improved groups again achieving statistical significance (β=.12, t[1687]=3.18, p=.001 and β=.06,

12 t[1687]=2.20, p<.05, respectively). As seen with mastery, follow-up regression models demonstrated that the link between self-esteem and depressive symptoms was significant in all three groups; however, it predicted lower depressed affect more strongly in the ill-spouse group ( R 2 =.07, β=-.30, t[304]=-5.23, p<.001) compared to the healthy-spouse ( R 2 =.02, β=-.06, t[859]=-4.64, p<.001) and spouse-improved ( R 2 =.02, β=-.17, t[298]=-3.16, p<.01) groups. Discussion The current study examined the effects of a common life stressor during the mid and late life years illness in a spouse on depressed affect and the role of two psychological resources namely, personal mastery and self-esteem in mitigating the effects of a spouse s illness on depressed affect. First, as predicted, respondents whose spouse became ill or remained ill with a long-term illness or physical disability reported significantly more depressive symptoms at T2 than those whose spouse remained healthy at both time points. These findings supplement the empirical literature showing that a spouse s disability or hospitalization during the middle or later years is associated with poorer psychological well-being (Korporaal et al., 2008; Zivin & Christakis, 2007). Theoretically, the findings are consistent with life course theory (Elder, 1998; Fuller-Iglesias et al., 2008) and its intersection with the stress process model (Pearlin, 2010). First, as explained by the principle of timing and sequencing of life events or transitions, during middle and late adulthood when individuals are advancing in age themselves, having a spouse who becomes ill or remains ill can result in higher depressed affect. Second, the principle of linked lives explains that individuals live interconnected lives and their health and well-being are linked to those around them (Elder, 1998; Fuller-Iglesias et al., 2008) especially those with whom they share their closest role relationships such as a spouse (Pearlin, 2010). Thus, change in the health status of one spouse affects the well-being of the other. The current findings make other important contributions. They cohere with the literature on crossspouse concordance of well-being (Bookwala, in press), supporting findings from studies that show that one

13 spouse s well-being is significantly linked to the other s as people age (Bookwala & Schulz, 1996; Hoppman, Gerstorf, & Hibbert, 2011). The data also support the prevalence of plasticity and resilience as individuals age (Baltes, 1987): respondents whose spouse had improved in health from T1 to T2 experienced a decline in depressive symptoms by T2, showing a significant recovery of psychological well-being when the stressor of an ill spouse was no longer present. These findings demonstrate that psychological well-being in mid and late life is influenced in important ways by context just as a spouse s illness can portend an increase in one s depressive symptomatology, a spouse s health recovery or effective health management can result in a rebound in one s psychological wellbeing. Consistent with the second hypothesis, the current findings showed that the adverse effects on depressed affect of a spouse remaining ill can be mitigated by higher levels of both mastery and self-esteem. Both higher mastery and self-esteem predicted lower depressed affect most strongly among respondents whose spouse was ill at both waves; in comparison, the benefits of these resources for psychological wellbeing in those whose spouse had remained healthy or whose health had improved between T1 and T2, while present, were significantly smaller in magnitude. Thus, these results confirm that mastery and selfesteem are significantly more salient to and vital for diminishing depressive symptoms in the face of stress that may be prolonged or recurrent. The protective role of psychological resources has been seen among individuals coping with their own poor health (e.g., van Jaarsveld, Ranchor, Sanderman, Ormel, & Kempen, 2005). The current findings demonstrate that these psychological resources are also valuable in promoting well-being among middle-aged and older adults within the context of a spouse s illness. The results also are consistent with life span models of resilience (Fuller-Iglesias et al., 2008; Ong et al., 2009; Ryff & Singer, 2003) showing that as individuals age they are well-positioned to adapt to the demands of an ill spouse if they have key psychological resources at their disposal. More importantly, Moen et al. (2010) point out that personal resources can be further built when life demands are duly dealt with and life-course fit is restored. Thus, the current findings have important clinical

14 implications. Given that higher levels of psychological resources such as mastery and self-esteem are effective in improving adaptation to the stress of an ill spouse, an important goal for clinicians is to make available to their middle-aged and older clients coping with an ill spouse effective interventions that focus on building psychological resources and strengths. Community-based interventions for individuals in these age groups can be successful in building their resources for stress adaptation (e.g., Collins & Benedict, 2006). For example, Reich and Zautra (1989) found that older adults experiencing conjugal bereavement or physical disability who were randomly assigned to a personal mastery-enhancing intervention in which they were encouraged to review their lifestyle, choices, and decisions and to recognize life domains in which they enjoyed personal control experienced increased levels of perceived mastery. McAuley, Katula, Duncan, and Mihalko (2000) found that an intervention of physical activity used with older adults improved global and domain-specific aspects of self-esteem. Likewise, von Humboldt and Leal (2012) found person-centered therapy to significantly enhance self-esteem levels in a sample of older adults. Similar interventions targeting individuals coping with an ill spouse may be effective in enhancing mastery and self-esteem among both middle-aged and older adults, promoting better adaptation to the occurrence of illness in a spouse and to the subsequent quality of care provided to the ill spouse as the need arises; analyses (not discussed earlier) showed that at T1 9.1% (57 of 624) and at T2 17.7% (93 of 525) of participants with an ill spouse provided assistance to the spouse with daily activities in the preceding year, rates that can be expected to increase with advancing age and illness in the spouse. Furthermore, Aldwin s deviation amplification model of stress and coping (Aldwin & Brustrom, 1997; Aldwin et al., 1996) explains that as individuals age, their past experiences in successfully adapting to life stressors can result in increases in psychological resources which, in turn, augment their coping repertoire available during future times of stress. Thus, to the extent that resources such as mastery and self-esteem are maintained or heightened following successful adaptation to the challenges of having an ill spouse, individuals are likely to adapt successfully to other stressors they may encounter subsequently in mid and late life.

15 Despite the significant findings seen in the current study, it is important to note its limitations. The two time points of data collection were separated by approximately a decade and data are unavailable on spousal health or other variables during the intervening period. The analyses also do not account for specific characteristics of the spouse s illness (e.g., its nature or duration). Thus, only general conclusions can be drawn regarding the effects of spousal health transitions on the depressed affect of middle-aged and older individuals; more fine-tuned analyses taking into account illness-specific characteristics are not feasible. Moreover, because no direct measures of respondents stress appraisals or changes therein regarding the spouse s illness were available in the dataset, a more comprehensive test of the stress process model could not be undertaken. Finally, the moderator model tested the effects of two psychological resources, mastery and self-esteem. Other psychological resources such as agency as well as social resources such as social support and social engagement also merit attention in the context of spousal health transitions. Future research using shorter intervals between data points, richer detail about the nature and duration of the spousal health transitions, and a more diverse set of psychosocial resources would supplement the current findings in important ways. Nevertheless, the present study makes important scientific contributions on the effects on depressed affect of spousal health transitions that occur during the middle and late adulthood years and factors that promote resilience in the face of spousal illness.

16 References Aldwin, C. A., & Brustrom, J. (1997). Theories of coping with chronic stress: Illustrations from the health psychology and aging literatures. In B. Gottlieb (Ed.), Coping with chronic stress (pp ). New York: Plenum Press. Aldwin, C. A., Sutton, K. J., & Lachman, M. (1996). The development of coping resources in adulthood. Journal of Personality, 64, Baltes, P. B. (1987). Theoretical propositions of life-span developmental psychology: On the dynamics between growth and decline. American Psychologist, 23, Bookwala, J. (in press). Affective contagion. Encyclopedia of Quality of Life Research. Heidelberg, Germany: Sage. Bookwala, J. (2012). Marriage and other partnered relationships in middle and late adulthood. In R. Blieszner & V. Bedford (Eds.), Handbook of Families and Aging, 2nd edition (pp ). Greenwood/ABC-CLIO: Westport, CT. Bookwala, J., & Schulz, R. (1996). Spousal similarity in well-being. The Cardiovascular Health Study, Psychology and Aging, 11, Collins, C. C., & Benedict, J. (2006). Evaluation of a community-based health program for the elderly: Lessons from Seniors CAN. American Journal of Health Promotion, 21, Elder, G. (1998). The life course as developmental theory. Child Development, 69, Fuller-Iglesias, H., Sellars, B., & Antonucci, T. (2008). Resilience in old age: Social relations as a protective factor. Research in Human Development, 5, Harmell, A. L., Chattillion, E. A., Roepke, S. K., & Mausbach, B. T. (2011). A review of psychobiology of dementia caregiving: A focus on resilience factors. Current Psychiatry Reports, 13,

17 Hoppmann, C. A., Gerstorf, D., & Hibbert, A. (2011). Spousal associations between functional limitation and depressive symptom trajectories: Longitudinal findings from the Study of Asset and Health Dynamics among the Oldest Old (AHEAD). Health Psychology, 30, Hildon, Z., Smith, G., Netuveli, G., & Blane, D. (2008). Understanding adversity and resilience at older ages. Sociology of Health and Illness, 30, Kessler, R. C., Andrews, G., Mroczek, D., Ustun, B., Wittchen, H-U. (1998). The World Health Organization Composite International Diagnostic Interview Short Form (CIDI-SF). International Journal of Methods in Psychiatric Research, 7, Korporaal, M., van Groenou, M. I. B., van Tilburg, T. G. (2008). Effects of own and spousal disability on loneliness among older adults. Journal of Aging and Health, 20, Lee, H-S., Brown, S. L., Mitchell, M. M., & Schiraldi, G. R. (2008). Correlates of resilience in the face of adversity for Korean Women immigrating to the US. Journal of Immigrant Minority Health, 10, McAuley, E., Blissmer, B., Katula, J., Duncan, T. E., & Mihalko, S. L. (2000). Physical activity, self-esteem, and self-efficacy relationships in older adults: A randomized controlled trial. Annals of Behavioral Medicine, 22, Moen, P., Sweet, S., & Hill, R. (2010). Risk, resilience, and life course fit: Older couples encores following job loss. In P. S. Fry & Keyes, C. (Eds.), New frontiers in resilient aging: Life-strengths and well-being in late life (pp ). New York, NY: Cambridge University Press. Nunnally, J. C. (1978). Psychometric theory. New York, NY: McGraw-Hill. Ong, A. D., Bergeman, C. S., & Boker, S. M. (2009). Resilience comes of age: Defining features in later adulthood. Journal of Personality, 77, Pearlin, L. I. (2010). The life course and the stress process: Some conceptual comparisons. Journal of Gerontology: Social Sciences, 65B, Pearlin, L. I., & Schooler, C. (1978). The structure of coping. Journal of Health and Social Behavior, 19, 2-22.

18 Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, Reich, J. W., & Zautra, A. J. (1989). A perceived control intervention for at-risk older adults. Psychology and Aging, 4, Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. Ryff, C. D., & Singer, B. (2003). Flourishing under fire: Resilience as a prototype of challenged thriving. In C. L. M. Keyes & Haidt, J. (Eds.), Flourishing: Positive psychology and the life well-lived (pp ). Washington, DC: American Psychological Association. Sargent-Cox, K. A., Anstey, K. J., & Luszcz, M. A. (2012). Change in health and self-perceptions of aging over 16 years: The role of psychological resources. Health Psychology, 31, Sweet, J. A., & Bumpass, L. L. (2002). The National Survey of Families and Households Waves 1, 2, and 3: Data description and documentation. Center for Demography and Ecology, University of Wisconsin-Madison Thoits, P.A. (2010). Stress and health: Major findings and policy implications. Journal of Health and Social Behavior, 51(S), S41-S53. Von Humboldt, S., & Leal, I. (2012). Person-centered therapy and older adults self-esteem: A pilot study with follow-up. Studies in Sociology of Science, 3, Van Jaarsveld, C. H. M., Ranchor, A. V., Sanderman, R., Ormel, J., & Kempen, G. I. J. M. (2005). The role of premorbid psychological attributes in short- and long-term adjustment after cardiac disease: A prospective study in the elderly in the Netherlands. Social Science and Medicine, 60, Walker, R. B., & Luszcz, M. A. (2009). The health and relationship dynamics of late-life couples: A systematic review of the literature. Ageing & Society, 29, Windle, G., Markland, D. A., & Woods, R. T. (2008). Examination of a theoretical model of psychological resilience in older age. Aging and Mental Health, 12,

19 Zivin, K., & Christakis, N. A. (2007). The emotional toll of spousal morbidity and mortality. The American Journal of Geriatric Psychiatry, 15,

20 Table 1 Group Means and Standard Deviations on Depressive Symptoms, Marital Happiness, Psychological Resources as a function of Gender and Time Healthy-spouse T1 T2 (n=870) Spouse health improved T1 T2 (n=309) Spouse health declined T1 T2 (n=210) Ill-spouse T1 T2 (n=315) T1 T2 T1 T2 T1 T2 T1 T1 Depressive Symptoms abcd Male Female Personal Mastery ab Male Female Self-esteem bc Male Female Higher scores=greater depressive symptomatology, higher mastery, and higher self-esteem. a Significant main effect for spousal health transitions group at p<.05; b Significant main effect for gender at p<.05 c Significant main effect for time at p<.05 d Significant time x spousal health interaction effect at p<.05

21 Table 2. Regression Analyses Testing Mastery as a Moderator in the link between Spousal Health and Depressive Symptomatology (N=1,704). T1 Depressive Symptoms T2 Depressive Symptoms B β s.e. t B β s.e. t Step 1 (Control variables) F(6,1697)=54.09***; R 2 =.16 F(9,1694)=49.96***; R 2 =.21 Gender *** ** Age *** Education *** *** Race *** * T1 employment status T1 self-rated health *** T2 employment status T2 self-rated health *** T1 depressive symptoms *** Step 2 (Tests of direct relationships) F(4,1693)=43.50***; R 2 =.08 F(4,1690)=39.01***; R 2 =.07 T1 mastery *** Healthy-spouse group ** ** Spouse-improved group *** Spouse-declined group T2 mastery *** Step 3 (Test of moderating effects) 1 F(3,1690)=1.18; R 2 =.00 F(3,1687)=3.80; R 2 =.01 T1 mastery x healthy-spouse group T1 mastery x spouse-improved group T1 mastery x spouse-declined group T2 mastery x healthy-spouse group ** T2 mastery x spouse-improved group ** T2 mastery x spouse-declined group Spousal health status groups were dummy coded (0=no, 1=yes) using the ill-spouse group as the reference group. Note: * p<.05; ** p <.01; p <.001 Higher scores represented being female, older, non-white, no high school diploma, being employed, having better health, higher mastery, and greater depressive symptomatology.

22 Table 3. Regression Analyses Testing Self-esteem as a Moderator in the link between Spousal Health and Depressive Symptomatology (N=1,704). T1 Depressive Symptoms T2 Depressive Symptoms B β s.e. t B β s.e. t Step 1 (Control variables) F(6,1697)=54.09***; R 2 =.16 F(9,1694)=49.96***; R 2 =.21 Gender *** ** Age *** Education *** *** Race *** * T1 employment status T1 self-rated health T2 employment status T2 self-rated health *** T1 depressive symptoms *** Step 2 (Tests of direct relationships) F(4,1693)=51.20***; R 2 =.09 F(4,1690)=22.36***; R 2 =.04 T1 self-esteem *** Healthy-spouse group *** *** Spouse-improved group * *** Spouse-declined group T2 self-esteem *** Step 3 (Test of moderating effects) 1 F(3,1690)=0.53; R 2 =.00 F(3,1687)=3.65; R 2 =.01 T1 self-esteem x healthy-spouse group T1 self-esteem x spouse-improved group T1 self-esteem x spouse-declined group T2 self-esteem x healthy-spouse group *** T2 self-esteem x spouse-improved group * T2 self-esteem x spouse-declined group Spousal health status groups were dummy coded (0=no, 1=yes) using the ill-spouse group as the reference group. Note: * p<.05; ** p <.01; p <.001 Higher scores represented being female, older, non-white, no high school diploma, being employed, having better health, higher self-esteem, and greater depressive symptomatology.

23 Depressive Symptoms healthy spouse spouse improved spouse declined ill spouse Time 1 Time 2 Figure 1. A significant time x spousal health interaction was found. Mean depressive symptomatology scores for T1 and T2 are displayed for each spousal health transitions group; means are adjusted for age, ethnicity, employment status, and self-rated health. At T1, the ill-spouse group scored significantly higher than the healthy-spouse group. At T2, the spouse-improved group scored lower than the ill-spouse and spouse-declined groups and the latter two groups also scored higher than the healthy-spouse group. Within groups, only the spouse-improved group showed a significant change from T1 to T2 showing a decline in depressive symptoms when the spouse was no longer ill.