The role of social engagement and support in the association between combined caregiver–cancer survivor status and depressive symptoms

Abstract

Background and Objectives

This study examined the overall impact of dual role status (cancer survivor and caregiver) compared with caregivers who were not cancer survivors, cancer survivors who were not caregivers, and noncaregivers and not cancer survivors, on depressive symptoms. We also explored the potential mediating role of social engagement and support in the association between combined cancer-caregiver status and depressive symptoms.

Research Design and Methods

We analyzed cross-sectional data from the National Social Life, Health, and Aging Project. We used causal mediation analysis methods to examine social engagement and social support as mediators of the association between intersections of cancer survivor and caregiver status and depressive symptoms. Models were adjusted for age, sex, ethnicity, race, education, household size, and partnership status.

Results

Cancer survivors with caregiving responsibilities had higher levels of depressive symptoms compared to caregivers only (Estimate  = 1.440, SE  = 0.423, p < .001), cancer survivors only (Estimate  = 1.286, SE  = 0.390, p < .001), and noncaregivers who were not cancer survivors (Estimate  = 1.590, SE  = 0.372, p < .001). A significant indirect effect of combined cancer-caregiver status on depressive symptoms was found through family support (Estimate  = −0.099, 95% CI: −0.189, −0.020) and through reliance on partner (Estimate  = 0.124, 95% CI: 0.021, 0.260).

Discussion

Cancer survivors with caregiving responsibilities reported higher depressive symptoms but also reported more family support and less reliance on their partners, which were associated with fewer depressive symptoms.

Implications

Cancer survivors face unique challenges as family caregivers. Attention to the unique needs of this group, including higher risk of depressive symptoms and complex social relationships, should be a priority for intervention efforts in public health.

As of 2025, approximately one in 18 Americans is a cancer survivor with 91% of survivors aged 50 years or older (Wagle et al., 2025). Individuals are considered cancer survivors from the time of cancer diagnosis onward for the rest of their life. Currently, there are 18.6 million cancer survivors in the United States, and this number is projected to rise to 22 million by 2035 (Wagle et al., 2025). Simultaneously, the number of family caregivers (individuals who provide unpaid assistance with personal, household, and health-related care to individuals with disabilities, chronic diseases, or functional limitations (National Cancer Institute [Internet], 2025; Roth et al., 2015) in the United States reached 63 million in 2025, with 41% being adults age 65 and older (National Alliance for Caregiving, 2025).

Cancer survivors who also assume caregiving responsibilities have been referred to in the literature as “surviving caregivers” (Mahmood et al., 2025). Among older cancer survivors, approximately one in five (nearly four million Americans), assume the potentially demanding role of family caregiving (Wang et al., 2024). Cancer survivorship is associated with heightened comorbidity challenges (Jiang et al., 2022), and caregiving is associated with higher odds of depressive symptoms and reduced quality of life (Liu et al., 2020). Undertaking this “dual role” of caregiving while being a cancer survivor has been found to be associated with an increased risk of physical decline, higher odds of depressive symptoms and reduced quality of life compared to caregivers only, cancer survivors only, and individuals who are neither caregivers nor cancer survivors (Mahmood et al., 2025; Wang et al., 2024).

Previous studies have shown that both cancer survivors and caregivers are at higher risk of depressive symptoms. Clinically significant depressive symptoms affect approximately 12%–27% of older cancer survivors (Mitchell et al., 2013; Nakamura et al., 2025; Parajuli et al., 2021). Furthermore, even subclinical depressive symptoms can negatively impact quality of life in this population (Robb et al., 2007). In older cancer survivors, depressive symptoms can compromise mental and physical well-being by hindering physical and social activity, decreasing self-sufficiency, and elevating the risk of disability, comorbidity, and premature death (Battat et al., 2024; Miller et al., 2024; Pandey & Anand, 2025). Research has shown that the transition to providing extensive and sustained caregiving assistance substantially worsens psychological health, nearly tripling the risk of clinically significant depressive symptoms in older caregivers compared to non-caregivers (Haley et al., 2020). Finally, the dual role of being both a cancer survivor and a caregiver can present unique challenges that intensify existing social and health inequities among the aging population, placing these individuals at heightened risk of adverse mental health outcomes (Mahmood et al., 2025; Wang et al., 2024).

To our knowledge, only two studies have examined the psychological well-being of cancer survivors who are caregivers, comparing them to controls using nationally representative data (Mahmood et al., 2025; Wang et al., 2024). Both studies had an epidemiological focus, estimating prevalence and odds ratios rather than evaluating the direct effect estimate of dual cancer-caregiver status on depressive symptoms. Therefore, we have not identified any studies that compare the more complex interplay of cancer-caregiving status across four groups: individuals who never had cancer and are not caregivers, caregivers only, cancer survivors only, and cancer survivors who have taken on caregiving roles.

Theoretical framework

To examine the complex relations between cancer survivorship, caregiving, and depressive symptoms, this study is conceptualized within the frameworks of the Stress and Coping Model (Lazarus, 1984) and Social Buffering Theory (Cohen & Wills, 1985). The Stress and Coping Model posits that an individual’s perception of stress is determined by their appraisal of a situation as threatening or challenging to their well-being, as well as by their assessment of the adequacy of internal and external resources available for coping. Coping is conceptualized as the set of strategies utilized to manage these stressors (Lazarus, 1984). Social Buffering Theory suggests that social support can mitigate the negative effects of stress by buffering against depressive symptoms following illness, including cancer (Li et al., 2024; Nørskov et al., 2021; Park & Kim, 2021; Pasek et al., 2017). Numerous studies have demonstrated that this protective effect is particularly pronounced among individuals experiencing high stress (buffering effect), a trend observed in both cancer survivor cohorts and the broader community (Kim et al., 2010; Pasek et al., 2017; Vasher et al., 2024). Caregivers aged 60 and above with higher levels of social support experience fewer depressive symptoms and tend to exhibit greater resilience to stress (Hayward & Wallace, 2014; National Archive of Computerized Data on Aging, 2024). In addition, satisfaction with social support and spirituality were found to be significant psychological factors associated with lower levels of depressive symptoms in cancer survivors (Payne et al., 2014). For cancer survivors who become caregivers, these two frameworks show promise in understanding potential dual burden of managing the long-term consequences of their cancer experience while concurrently providing caregiving assistance.

In addition, both the Stress and Coping and Social Buffering frameworks suggest that social support plays a crucial role in helping cancer survivors develop effective coping mechanisms for managing their disease and its consequences, while also promoting a healthy approach to psychological recovery (Pasek et al., 2017). Social support can act as a moderator of the relation between comorbidities and psychological distress (stronger benefits with high levels of stress), providing a lasting positive psychological impact, even long after exposure to the disease (stressor) (Nørskov et al., 2021). In addition, both informal connections and formal social activities have been found to moderate the relationship between cancer survivorship and depressive symptoms, meaning they can weaken or strengthen the impact of cancer on depressive symptoms depending on their level or presence (Park & Kim, 2021). Social connectedness has been shown to mediate the link between social isolation and depression, meaning that social isolation leads to fewer social connections, which in turn increases depressive symptoms (Li et al., 2024). This may explain why more socially isolated cancer survivors tend to have higher depressive symptoms due to having fewer social connections. Together, these frameworks provide a comprehensive approach to explore the impact of social support and social engagement in reducing depressive symptoms among dual-role individuals. Guided by these frameworks, we developed a conceptual framework to explain the complex relations between mental health in individuals with dual caregiving-cancer survivor roles, as well as the potential buffering effects of social support and engagement (Figure 1).

Figure 1.

Conceptual framework: Relationship between cancer survivor-caregiver status and depressive symptoms.

While the mediation effects of social support and social engagement on depressive symptoms have been previously examined in the individuals who were neither caregivers nor cancer survivors, caregivers (Vasher et al., 2024), and cancer survivors (Kim et al., 2010; Li et al., 2024), previous studies have not comprehensively explored the role of social support and engagement in older dual-role individuals or identified mechanisms that explain how social relationships may help protect against depressive symptoms. Thus, examining the role of social engagement and support in addressing these mental health challenges among dual-role individuals is essential for advancing scientific understanding and identifying specific mechanisms that can be leveraged through targeted public health programs.

The aims of this study are to (1) examine overall impact of combined cancer-caregiver status on depressive symptoms compared to individuals who are neither cancer survivors nor caregivers, cancer survivors who are not caregivers, and caregivers who are not cancer survivors; (2) study the potential role of social support and social engagement as mediators of the association between combined cancer survivor and caregiver status and depressive symptoms. We examined a total of six social support variables as mediators (i.e., support from partners, family, and friends, and reliance on partners, family, and friends) and a total of five social engagement variables (i.e., religious attendance, volunteer work, participation in meetings or groups, socializing with friends or relatives, and visiting neighbors).

Research design and methods

Data source and sample design

We used data from Round 2 (2010–2011) of the National Social Life, Health, and Aging Project (NSHAP) (National Archive of Computerized Data on Aging, 2024). NSHAP is a population-based longitudinal study in the United States that explores the health and social determinants influencing the well-being of older adults living in the community (National Archive of Computerized Data on Aging, 2024). It employed a multistage, area-probability sampling design to select community-dwelling adults aged 57–85 years in the United States. The Round 2 sample included respondents from Round 1 (2005–2006) who were reinterviewed, along with their spouses or cohabiting partners, regardless of age (National Archive of Computerized Data on Aging, 2024). Eligibility criteria included being an English- or Spanish-speaking adult residing in the community and capable of participating in an in-person interview. Data were collected through face-to-face interviews conducted in respondents’ homes, supplemented by self-administered questionnaires that assessed health, social networks, and psychological well-being (National Archive of Computerized Data on Aging, 2024). We used Round 2 data, because it included the largest number of participants with a history of cancer and caregiving (as compared to round 1 and 3). We did not use Round 4 data, because data collection occurred during COVID-19, which may have affected caregiving, social relationships, and mental health patterns (National Alliance for Caregiving, 2025).

In total, 3,377 older adults aged 57 and above participated in Round 2 of NSHAP (Hayward & Wallace, 2014). Participants were included in the current cross-sectional study if they had data on caregiving status (providing unpaid care), cancer history, depressive symptoms, social engagement, and social support (Figure 2). Participants were excluded from the study if they had missing or incomplete data on the key study variables. Therefore, the analytical sample comprised 2,435 participants.

Figure 2.

Flowchart of study participants’ selection process.

To study support from partner and reliance on partner as mediators, we created a subsample consisting of participants who had a partner at the time of the interview and had complete data on support from partner and reliance on partner. This subsample comprised 1,932 participants.

Measures

Independent variable: Combined caregiving and cancer status

Cancer status was determined based on participants’ responses to two questions (1) “Has a doctor ever told you that you have cancer (other than skin cancer)?” and (2) “Has a doctor ever told you that you have skin cancer (including Melanoma or other)?” Participants answering “Yes” to either question were classified as having a history of cancer. Caregiving status was determined based on participants’ responses to the question “Are you currently assisting an adult who needs help with day-to-day activities because of age or disability?” (Yes/No). Combined caregiving and cancer status was coded as a categorical variable. The reference group (0) included participants who did not have cancer and caregiving responsibilities. Those who had cancer but were not caregivers were coded as (1), those who did not have cancer but were caregivers were coded as (2). Participants who had both cancer and caregiving responsibilities were coded as (3).

Dependent variable: Depressive symptoms

Depressive symptoms were assessed using the 11-item short form of the CES-D (Center for Epidemiologic Studies Depression Scale) (Payne et al., 2014). Scores were summed across all items to create a continuous variable ranging from 0 to 22, with higher values indicating more depressive symptoms (Payne et al., 2014).

Mediators

We studied social support and social engagement as mediators of the association between combined cancer survivor-caregiver status and depressive symptoms. Social support and social engagement were conceptualized according to the definitions proposed by Cohen (2004). Social support was differentiated into instrumental support (reliance on partner, friend, and family) and emotional support (ability to open up to a partner, family member, and friends about caregivers' worries). Social engagement was conceptualized as a multidimensional construct reflecting active participation in a wide range of social activities and relationships, including attending religious services, volunteering, participating in organized group meetings, socializing with friends or relatives, and visiting neighbors. Social support was measured with four variables in the primary analytical sample: (1) family support, (2) reliance on family, (3) friends support, and (4) reliance on friends. Two additional mediator variables were studied in the subsample of participants who were partnered at the time of the interview: (1) partner support and (2) reliance on a partner. Social engagement was measured with five variables: (1) attending religious services, (2) volunteering, (3) attending meetings of any organized groups, (4) socializing with friends or relatives, and (5) social visits to neighbors. Therefore, our analyses included eleven potential mediators of the association between combined cancer survivor-caregiver status and depressive symptoms.

Partner, family, and friend support was assessed based on the participants’ responses to the question: “How often can you open up to your partner/member of your family/friend if you need to talk about your worries?” Response options included: never (0), hardly ever or rarely (1), some of the time (2), and often (3). For the current analysis, we merged the never and hardly ever or rarely categories due to sparse data. Therefore, each of these three variables was coded as categorical (0 = low [from never to hardly ever or rarely]; 1 = moderate [some of the time]; 2 = high [often]).

Reliance on partner, family, and friends was assessed based on participants’ responses to the question: “How often can you rely on your partner/member of your family/friend for help if you have a problem?” Response options included: never (0), hardly ever or rarely (1), some of the time (2), and often (3). For the current analysis, we merged the never and hardly ever or rarely categories due to sparse data. Therefore, each of these three variables was coded as 0 = low (from never to hardly ever or rarely); 1 = moderate (some of the time); 2 = high (often).

Frequency of attending religious services was assessed based on participants’ responses to the question: “Thinking about the past 12 months, about how often have you attended religious services?” Response options included: never (0), about once or twice a year (1), several times a year (2), about once a month (3), every week (4), and several times a week (5). For the current analysis, we merged categories due to sparse data. Therefore, in our analyses the variable was coded as 0 = never (up to several times a year); 1 = occasional (about once a month); 2 = frequent (between every week to several times a week).

Frequency of attending meetings of any organized groups was assessed based on participants’ responses to the question: “In the past 12 months, how often did you attend meetings of any organized group?” Response options included: never (0), less than once a year (1), about once or twice a year (2), several times a year (3), about once a month (4), every week (5), and several times a week (6). For the current analysis, we merged some of these categories due to sparse data. Therefore, in our analyses each of these variables was coded as 0 = rare (up to twice a year); 1 = occasional (from several times a year to about once a month); 2 = frequent (from every week to several times a week).

Frequency of volunteering was assessed based on participants’ responses to the question: “In the past 12 months, how often did you do volunteer work for religious, charitable, political, health-related, or other organizations?” Response options included: never (0), less than once a year (1), about once or twice a year (2), several times a year (3), about once a month (4), every week (5), and several times a week (6). For the current analysis, we merged some of these categories due to sparse data. Therefore, in our analyses each of these variables was coded as 0 = rare (up to twice a year); 1 = occasional (from several times a year to about once a month); 2 = frequent (from every week to several times a week).

Frequency of socializing with friends or relatives was assessed based on participants’ responses to the question: “In the past 12 months, how often did you get together socially with friends or relatives?” Response options included: never (0), less than once a year (1), about once or twice a year (2), several times a year (3), about once a month (4), every week (5), and several times a week (6). For the current analysis, we merged some of these categories due to sparse data. Therefore, in our analyses each of these variables was coded as 0 = rare (up to twice a year); 1 = occasional (from several times a year to about once a month); 2 = frequent (from every week to several times a week).

Frequency of social visits to neighbors was assessed based on participants’ responses to the question: “How often do you and people in this area visit each other’s homes or when you meet on the street?” Response options included: never (0), rarely (1), sometimes (2), and often (3). For the current analysis, we merged some of these categories due to sparse data. Therefore, in our analysis, the variable was coded as 0 = never; 1 = occasional (from rarely to sometimes); 2 = frequent (often).

Confounder variables

Confounders included demographic and caregiving characteristics selected a priori based on previous research of their association with cancer status, caregiving, social support, social engagement, and depressive symptoms (Gutiérrez-Sánchez et al., 2023; Ho et al., 2025; Mahmood et al., 2025; Wang et al., 2024). Demographic and caregiving characteristics included: age (in years), sex (0 = male; 1 = female), education level (1 = less than high school; 2 = high school or equivalent; 3 = vocational certificate/some college degree/associate degree; 4 = bachelor’s degree or higher), race (1 = White/Caucasian; 2 = Black/African-American; 3 = Other (Asian, Pacific Islander, American Indian or Alaskan Native)), ethnicity (1  = Latino or Hispanic; 0 = Non-Latino or Hispanic), household size (range: 0–60), and partnership status (0 = not partnered; 1 = partnered).

Statistical analysis

Sample characteristics were reported using medians with interquartile ranges for continuous variables and percentages for categorical variables. We reported the characteristics for the complete sample and for subsamples of dual-role individuals, caregivers who are not cancer survivors, cancer survivors who are not caregivers, and noncaregivers and not cancer survivors. To examine differences in sample characteristics across these four groups, we used Kruskal–Wallis tests for non-normally distributed continuous variables and Extended Fisher’s Exact Tests for categorical variables. We also reported the caregiving characteristics for caregivers who were not cancer survivors and dual-role individuals using medians with interquartile ranges for continuous variables and percentages for categorical variables. Descriptive statistics for social support and social engagement variables treated as continuous variables were reported using medians with interquartile ranges.

We used R software (version 4.4.1) to clean the dataset and conduct regression and mediation analyses. First, we used a multiple linear regression analysis to estimate the overall association between the four groups based on cancer survivor and caregiver status and depressive symptoms. Second, mediation analyses were conducted to examine the indirect effects of dual cancer survivor and caregiver status on depressive symptoms through social support and social engagement (Figure 3). We estimated single mediator models, in which the indirect effect of combined cancer survivor and caregiver status (X) on depressive symptoms (Y) was assessed through one mediator at a time. In our main analyses, we compared dual-role individuals to noncaregivers who did not have cancer. We performed four secondary analyses, where we compared (1) dual-role individuals to caregivers who never had cancer, (2) dual-role individuals to cancer survivors without caregiving responsibilities, (3) cancer survivors without caregiving responsibilities to noncaregivers who did not have cancer, and (4) caregivers who did not have cancer to noncaregivers who did not have cancer.

Figure 3.

Theoretical model illustrating the mediating role of social support and social engagement in the relation between combined caregiver and cancer status and depressive symptoms.

To conduct mediation analysis, we used ordinal logistic regression to estimate the association between combined cancer survivor and caregiver status and the social engagement and social support mediator variables. Linear regression was used to estimate the associations between combined cancer survivor and caregiver status, each mediator, and depressive symptoms. In the latter model, we also assessed potential exposure-mediator interaction by adding an interaction term between combined cancer survivor and caregiver status and the respective mediator. If exposure-mediator interaction was present based on p < .05, we included it in the analyses to estimate the following effects: pure natural direct effect (PNDE), total natural direct effect (TNDE), pure natural indirect effect (PNIE), and total natural indirect effect (TNIE). In the absence of exposure-mediator interaction, we estimated a natural direct effect (NDE) and natural indirect effect (NIE). The mediation models were adjusted for age, sex, race, ethnicity, education, partnership status, and household size. We estimated 95% percentile bootstrap confidence intervals using 1,000 bootstrap resamples for the indirect effect estimates. An indirect effect estimate was considered significant if the 95% percentile bootstrap confidence interval did not include zero. A two-tailed p-value of .05 was considered significant for all other effect estimates. The “mediation” package was used to conduct the mediation analysis (Tingley et al., 2014).

The hypothesized temporal ordering was guided by the conceptual framework (see Figure 1) and the reference periods of the study measures. Cancer history reflects a diagnosis that occurred before the survey assessment, while caregiving status represents a current caregiving role that was conceptualized as preceding the assessment of depressive symptoms. Social support and social engagement were conceptualized as psychosocial resources through which cancer history and caregiving experiences may influence depressive symptoms. Most of these potential mediators were assessed with reference to the 12 months preceding the survey, supporting their temporal ordering before the outcome. Depressive symptoms were assessed for the week preceding the interview and were therefore treated as the most proximal outcome.

In sensitivity analyses, we repeated the mediation analyses where we treated all mediators as continuous variables using their original scale from the questionnaire. In these sensitivity analyses, we used linear regression to estimate the associations between combined cancer and caregiver status and each of the mediators.

Results

Descriptive statistics for the sample

Descriptive statistics for the total sample and four groups (individuals who were neither cancer survivors nor caregivers, cancer survivors who were not caregivers, caregivers who were not cancer survivors, and cancer survivors with a history of cancer) are shown in Table 1. Several sociodemographic characteristics differed significantly between the groups, including age, sex, race, ethnicity, and household size. Cancer survivors with caregiving responsibilities were older (median age: 73 (67–79.8) compared with caregivers who did not have cancer 71 (66–77) and those who were not caregivers and not cancer survivors 70 (66–77), while cancer survivors without caregiving responsibilities had a median age of 75 (69–80). Cancer survivors with caregiving responsibilities were more likely male (56.9%) compared to caregivers without a history of cancer (40.5%), and individuals who were not caregivers and not cancer survivors (44.1%), while cancer survivors without caregiving responsibilities were predominantly male (57.1%). Cancer survivors with caregiving responsibilities were predominantly White (87.3%) compared to caregivers without a history of cancer (71.0%), and individuals who were not caregivers and not cancer survivors (79.4%). Similarly to individuals in dual roles, cancer survivors without caregiving responsibilities were predominantly White (91.5%). Household size differed statistically across groups (p < .001), but the median household size was 1 in all groups. Social engagement and social support characteristics did not differ meaningfully across groups, except for reliance on a partner. Reliance on a partner differed significantly across groups (p < .01): high reliance was reported among 41.3% of caregivers who did not have cancer, 46.1% of dual-role individuals, 44.0% of cancer survivors who were not caregivers and 42.6% of individuals who were neither cancer survivors nor caregivers.

Table 1.

Descriptive statistics for study sample.

Characteristics	Total sample		Not cancer survivor & not caregiversa		Caregiversb		Cancer survivorsc		Dual role: survivor-caregiversd		p e
	N	M (IQR) or N (%)	N	M (IQR) or N (%)	N	M (IQR) or N (%)	N	M (IQR) or N (%)	N	M (IQR) or N (%)
Age	2,435	72 (67–78)	1,509	70 (66–77)	269	71 (66–77)	555	75 (69–80)	102	73 (67–79.8)	<.001
Sex											<.001
Male	1,150	47.2	666	44.1	109	40.5	317	57.1	58	56.9
Female	1,285	52.8	843	55.9	160	59.5	238	42.9	44	43.1
Race											<.001
White/Caucasian	1,986	81.6	1198	79.4	191	71.0	508	91.5	89	87.3
Black/African-American	288	11.8	194	12.9	55	20.4	30	5.4	9	8.8
Othere	161	6.6	117	7.8	23	8.6	17	3.1	4	3.9
Ethnicity											<.001
Non-Hispanic/Latino	2,204	90.5	1344	89.1	234	87.0	529	95.3	97	95.1
Hispanic/Latino	231	9.5	165	10.9	35	13.0	26	4.7	5	4.9
Education											.09
Lower than high school	393	16.1	259	17.2	59	21.9	64	11.5	11	10.8
High school or equivalent	619	25.4	390	25.8	61	22.7	137	24.7	31	30.4
Vocational certificate/some college degree/associate degree	787	32.3	470	31.1	91	33.8	193	34.8	33	32.4
Bachelor degree or above	636	26.1	390	25.8	58	21.6	161	29.0	27	29.5
Household size	2,435	1 (1–1)	1,509	1 (1–1)	269	1 (1–1)	555	1 (1–1)	102	1 (1–1)	<.001
Partnership											.14
Not partnered	700	28.7	452	30.0	69	25.7	166	29.9	13	12.7
Partnered	1,735	71.3	1057	70.0	200	74.3	389	70.1	89	87.3
Depressive symptoms	2,435	9 (7–11.5)	1,509	8 (7–11)	269	8 (7–12)	555	9 (7–12)	102	10 (8–13)	<.001
Social engagement
Religious attendance											.06
Rare/never	1043	42.8	665	44.1	92	34.2	236	42.5	50	49.0
Occasional	217	8.9	132	8.7	25	9.3	55	9.9	5	4.9
Frequent	1175	48.3	712	47.2	152	56.5	264	47.6	47	46.1
Volunteer work											.53
Rare/never	1402	57.6	880	58.3	141	52.4	322	58.0	59	57.8
Occasional	506	20.8	311	20.6	55	20.4	119	21.4	21	20.6
Frequent	527	21.6	318	21.1	73	27.1	114	20.5	22	21.6
Attending meetings/groups											.45
Rare/never	1115	45.8	714	47.3	119	44.2	234	42.2	48	47.1
Occasional	655	26.9	390	25.8	72	26.8	164	29.5	29	28.4
Frequent	665	27.3	405	26.8	78	29.0	157	28.3	25	24.5
Socializing with friends or relatives											.24
Rare/never	205	8.4	143	9.5	19	7.1	36	6.5	7	6.9
Occasional	939	38.6	565	37.4	109	40.5	221	39.8	44	43.1
Frequent	1,291	53.0	801	53.1	141	52.4	298	53.7	51	50.0
Social visits to neighbors											.90
Rare/never	244	10.0	153	10.1	30	11.2	51	9.2	10	9.8
Occasional	1,635	67.1	1006	66.7	176	65.4	381	68.6	72	70.6
Frequent	556	22.8	350	23.2	63	23.4	123	22.2	20	19.6
Social support from social ties
Family support											.41
Low	407	16.7	255	16.9	40	14.9	101	18.2	11	10.8
Moderate	859	35.3	538	35.7	87	32.3	198	35.7	36	35.5
High	1,169	48.0	716	47.4	142	52.8	256	46.1	55	53.9
Friends support											.32
Low	747	30.7	458	30.4	75	27.9	180	32.4	33	32.4
Moderate	1,033	42.4	648	42.9	120	44.6	231	41.6	34	33.3
High	655	26.9	402	26.6	74	27.5	144	25.9	35	34.3
Reliance on family											.75
Low	205	8.4	127	8.4	20	7.4	48	8.6	10	9.8
Moderate	545	22.4	344	22.8	65	24.2	116	20.9	20	219.6
High	1685	69.2	1038	68.8	184	68.4	391	70.5	72	70.6
Reliance on friends											.91
Low	474	19.5	299	19.8	53	19.7	106	19.1	16	15.7
Moderate	916	37.6	567	37.6	105	39.0	205	36.9	39	38.2
High	1045	42.9	643	42.6	111	41.3	244	44.0	47	46.1
Partner supportf											.13
Low	72	3.7	47	4.0	8	3.5	12	2.7	5	5.2
Moderate	441	22.8	263	22.5	61	26.5	88	20.0	29	30.2
High	1419	73.4	857	73.4	161	70.0	339	77.2	62	64.6
Reliance on partnerf											<.01
Low	46	2.4	24	2.2	8	5.0	7	1.6	7	7.3
Moderate	219	11.3	122	10.5	41	16.4	42	9.6	14	14.6
High	1667	86.3	1021	87.5	181	78.6	390	88.8	75	78.1

Note. The variables: “depressive symptoms,” “age,” and “household size,” did not follow normal distribution; therefore, descriptive statistics are presented as median and interquartile range.

IQR = interquartile range; M = median; N = number of participants.

^a Not cancer survivor & Not caregivers refers to individuals who were neither caregivers nor cancer survivors.

^b Caregivers refers to individuals who provided care to others but were not cancer survivors.

^c Cancer survivors refers to individuals who were cancer survivors but did not have caregiving responsibilities.

^d Dual Role: Survivor-Caregivers refers to individuals who were both cancer survivors and caregivers.

^eThe race category “Other” includes identifying as Asian, Pacific Islander, American Indian or Alaskan Native.

^fDescriptive statistics for the variables “partner support” and “reliance on partner” were assessed only among participants who were partnered at the time of the interview.

Descriptive statistics for the subsample of participants with partners are presented in Supplementary Table 1. Descriptive statistics for social support and social engagement variables treated as continuous variables are presented in Supplementary Table 2. The caregiving characteristics of the study sample are presented in Supplementary Table 3. In terms of caregiving characteristics, individuals who were both cancer survivors and caregivers and those who provided care but were not cancer survivors did not differ significantly in age (p  = .05), caregiving type (p  = .54), primary caregiving status (p  = .15), providing most of the care (p  = .09), hours (p  = .53) and days spent providing care (p  = .97), or emotional strain due to caregiving (p  = .19), with the exception of financial strain related to caregiving (p  = .04). Dual-role individuals were more likely to report higher levels of financial strain (63.7%) compared with caregivers who were not cancer survivors (53.9%).

Overall association between combined cancer and caregiver status and depressive symptoms

Table 2 includes estimates of the overall association between combined cancer and caregiver status and depressive symptoms, using three reference groups: (1) individuals who neither had cancer nor caregiving responsibilities, (2) caregivers who never had cancer, and (3) cancer survivors who did not have caregiving responsibilities. Dual-role individuals reported significantly higher depressive symptoms than each reference group: (1) individuals who neither had cancer nor caregiving responsibilities, (2) caregivers who never had cancer, and (3) cancer survivors who did not have caregiving responsibilities. Caregivers who never had cancer did not significantly differ in CES-D scores from those who neither had cancer nor caregiving responsibilities or from cancer survivors who did not have caregiving responsibilities. Cancer survivors who did not have caregiving responsibilities did not differ in CES-D scores from those who neither had cancer nor caregiving responsibilities or from caregivers who never had cancer.

Table 2.

Multiple linear regression analysis of the association between cancer-caregiver status and depressive symptoms.

Exposure group	Reference group 1	Reference group 2	Reference group 3
	Never had cancer, nor had caregiving responsibilities	Caregivers who never had cancer	Cancer survivors who did not have caregiving responsibilities
	Estimate β (SE), p	Estimate β (SE), p	Estimate β (SE), p
Cancer survivors with caregiving responsibilities	1.590 (0.372), p < .001	1.440 (0.423), p < .001	1.286 (0.390), p < .001
Caregivers who never had cancer	0.150 (0.240), p = .53	—	−0.155 (0.275), p = .57
Cancer survivors who did not have caregiving responsibilities	0.304 (0.184), p = .10	0.155 (0.275), p = .57	—

Note. Effect estimates were controlled for confounding by age, sex, race, ethnicity, partnership status, household size, and education.

Mediating effect of social support and social engagement

Table 3 includes the effect estimates from the mediation analyses. Here we presented the mediation effect estimates for dual-role individuals as cancer survivors and caregivers, compared to those who never had cancer, nor caregiving responsibilities. The only statistically significant exposure-mediator interaction term was observed for combined caregiver-cancer status and reliance on family (β(moderate) = −3.271, SE  = 1.438, p  = .02). For this mediator, we reported the PNDE, TNDE, PNIE, and TNIE estimates. For all other mediators, we reported the NDE and NIE estimates. Detailed results from the exposure-mediator interaction analyses are presented in Supplementary Tables 4 and 5. Mediation effect estimates based on the comparison between other groups (e.g., dual roles compared to caregivers who never had cancer, and dual roles compared to cancer survivors without caregiver responsibilities) are presented in Supplementary Tables 6 and 7.

Table 3.

Analysis of mediators of the association between combined cancer-caregiver history and depressive symptoms: Single-mediator models (reference group: Individuals who never had cancer, nor had caregiving responsibilities).

Mediator	a path (95% CI)	b path (95% CI)	h path (95% CI)	Direct effect (95% CI)	Indirect effect (95% CI)
Mediator 1: Religious attendance	−0.108 (−0.512, 0.294)	Occasional: 0.187 (−0.346, 0.719) Frequent: −0.159 (−0.467, 0.148)	—	NDE: 1.594 (0.765, 2.430)	NIE: 0.005 (−0.022, 0.030)
Mediator 2: Volunteer work	−0.087 (−0.497, 0.311)	Occasional: −0.365 (−0.738, 0.008) Frequent: −0.407 (−0.779, −0.034)	—	NDE: 1.586 (0.778, 2.420)	NIE: 0.001 (−0.031, 0.050)
Mediator 3: Socializing with friends or relatives	−0.047 (−0.436, 0.349)	Occasional: −0.965 (−1.519, −0.411) Frequent: −1.627 (−2.172, −1.081)	—	NDE: 1.589 (0.737, 2.430)	NIE: −0.013 (−0.085, 0.110)
Mediator 4: Social visits to neighbors	−0.140 (−0.567, 0.280)	Occasional: −0.137 (−0.625, 0.352) Frequent: −0.627 (−1.173, −0.081)	—	NDE: 1.574 (0.767, 2.400)	NIE: 0.024 (−0.027, 0.060)
Mediator 5: Attending meetings of any organized groups	−0.067 (−0.455, 0.316)	Occasional: −0.317 (−0.674, 0.040) Frequent: −0.435 (−0.798, −0.072)	—	NDE: 1.586 (0.757, 2.420)	NIE: 0.006 (−0.036, 0.050)
Mediator 6: Family support	0.499 (0.109, 0.900)	Moderate: −0.341 (−0.769, 0.087) High: −1.000 (−1.420, −0.580)	—	NDE: 1.694 (0.880, 2.520)	NIE: −0.099 (−0.189, −0.020)
Mediator 7: Friends support	0.276 (−0.116, 0.669)	Moderate: −0.123 (−0.468, 0.222) High: −0.570 (−0.964, −0.176)	—	NDE: 1.638 (0.827, 2.490)	NIE: −0.034 (−0.098, 0.020)
Mediator 8: Reliance on family	0.015 (−0.418, 0.472)	Moderate: 0.244 (−0.487, 0.977) High: −0.873 (−1.539, −0.208)	Moderate: −3.271 (−6.090, −0.451) High: −2.286 (−4.749, 0.177)	Moderate: TNDE: 1.571 (0.806, 2.340) PNDE: 1.568 (0.796, 2.340) High: TNDE: 1.571 (0.806, 2.340) PNDE: 1.568 (0.796, 2.340)	Moderate: TNIE: 0.002 (−0.155, 0.170) PNIE: −0.001 (−0.092, 0.100) High: TNIE: 0.002 (−0.155, 0.170) PNIE: −0.001 (−0.092, 0.100)
Mediator 9: Reliance on friends	0.205 (−0.175, 0.591)	Moderate: −0.273 (−0.681, 0.135) High: −0.600 (−1.007, −0.192)	—	NDE: 1.615 (0.809, 2.460)	NIE: −0.024 (−0.086, 0.020)
Mediator 10: Partner supporta	−0.458 (−0.890, −0.008)	Moderate: 0.024 (−0.830, 0.878) High: −0.954 (−1.767, −0.140)	—	NDE: 1.791 (0.981, 2.590)	NIE: 0.110 (−0.012, 0.220)
Mediator 11: Reliance on partnera	−0.833 (−1.346, −0.282)	Moderate: 0.379 (−0.713, 1.471) High: −0.743(−1.754, 0.267)	—	NDE: 1.786 (0.894, 2.580)	NIE: 0.124 (0.021, 0.260)

Note. CI = confidence interval; NDE = Natural Direct Effect; NIE = Natural Indirect Effect; PNDE = Pure Natural Direct Effect; PNIE = Pure Natural Indirect Effect; TNDE = Total Natural Direct Effect; TNIE = Total Natural Indirect Effect.

All models were adjusted for age, sex, ethnicity, race, education, household size, and partnership status.

^aThe mediation effect of partner support and reliance on partner were assessed only among participants who were partnered at the time of the interview. All models were adjusted for age, sex, ethnicity, race, education, household size, and partnership status.

Both family support and reliance on a partner mediated the association between being a cancer survivor with caregiving responsibilities and depressive symptoms. Dual-role individuals were more likely to experience family support compared to individuals who never had cancer nor were caregivers (path “a”) (β = 0.499, 95% CI: 0.109, 0.900, p  = .01). Individuals with high family support, on average, had fewer depressive symptoms compared to individuals with low family support (path “b”) (β = −1.000, 95% CI: −1.420, −0.580, p < .001). Therefore, dual-role individuals, on average, had lower depressive symptoms compared to individuals who never had cancer nor were caregivers, because of higher family support (NIE = −0.099; 95% CI: −0.189, −0.020). Dual-role individuals were less likely to rely on their partners compared to individuals who never had cancer nor were caregivers (path “a”) (β = −0.833, 95% CI: −1.346, −0.282, p  = .003). Individuals with moderate reliance on their partners, on average, had more depressive symptoms compared to individuals with low reliance on their partners (path “b”) (β (moderate) = 0.379, 95% CI: −0.713, 1.471, p  = .50), while individuals with high reliance on their partners, on average, had fewer depressive symptoms compared to individuals with low reliance on partners (β (high) = −0.743, 95% CI: −1.754, 0.267, p  = .15). Therefore, dual-role individuals, on average, had higher depressive symptoms compared to individuals who never had cancer nor were caregivers, because of lower reliance on their partner (NIE = 0.124; 95% CI: 0.021, 0.260).

Statistically significant mediator-outcome associations (i.e., “b” paths) were observed for the following mediators: volunteer work, socializing with friends or relatives, social visits to neighbors, attending meetings of any organized groups, friends support, reliance on family, reliance on friends, and partner support. The effect estimates all indicate that individuals who received social support and engaged in social events, on average, had fewer depressive symptoms compared to those who did not receive social support nor engaged in social events. However, cancer-caregiver status was not associated with any of these mediators, so there were no statistically significant indirect effects through these mediators.

Results from our secondary analyses were largely in line with the results presented above. First, we found no statistically significant indirect effect estimates for any of the social support and social engagement variables when comparing dual-role individuals to caregivers who never had cancer (see Supplementary Table 6). Second, partner support (NIE  = 0.128, 95% CI: 0.004, 0.240) and reliance on partner (NIE  = 0.115, 95% CI: 0.018, 0.260) mediated the association between combined cancer-caregiver status and depressive symptoms when comparing dual-role individuals to cancer survivors who were not caregivers (see Supplementary Table 7). No statistically significant indirect effect estimates were found for any of the social support or social engagement variables when comparing individuals who had cancer and were not caregivers to individuals who did not have cancer and were not caregivers (see Supplementary Tables 8 and 10). Both volunteer work and reliance on a partner mediated the association between being a caregiver without a history of cancer and depressive symptoms, compared to individuals who did not have cancer and were not caregivers (see Supplementary Tables 9 and 11). Caregivers who did not have cancer had, on average, lower depressive symptoms than individuals who never had cancer nor caregiving responsibilities, because they frequently engaged in volunteer work (NIE = −0.020; 95% CI: −0.075, 0.000). Caregivers who did not have cancer, on average, had higher depressive symptoms compared to individuals who never had cancer nor were caregivers, because they were less likely to rely on their partners (NIE  = 0.061; 95% CI: 0.010, 0.150).

Sensitivity analyses

Results from the sensitivity analyses, in which mediators were treated as continuous variables, were consistent with the models in which mediators were treated as categorical variables (see Supplementary Table 12). Both approaches yielded significant indirect effect estimates for family support (NIE (categorical)  = −0.099; 95% CI: −0.189, −0.020); NIE (continuous)  = −0.097; 95% CI: −0.181, −0.020) and reliance on partner (NIE (categorical) = 0.124; 95% CI: 0.021, 0.260); NIE (continuous) = 0.118; 95% CI: 0.021, 0.250). These results indicate that the main findings were robust to different operationalizations of the mediators.

Discussion

To our knowledge, this is the first empirical study to examine the overall impact of combined cancer survivor and caregiver status on depressive symptoms, as well as the mediating role of social connection and engagement in this association. The aims of this study were to (1) examine overall impact of combined cancer-caregiver status on depressive symptoms compared to individuals who were neither cancer survivors nor caregivers, cancer survivors who were not caregivers, and caregivers who were not cancer survivors; (2) study the potential role of social support and social engagement as mediators of the association between combined cancer survivor and -caregiver status and depressive symptoms. Dual-role individuals experienced higher depressive symptoms than noncaregivers without a history of cancer. The results from the mediation analyses demonstrated that only family support and reliance on partner mediated the relation between cancer survivor and caregiver status and depressive symptoms.

Our estimate of the overall effect of combined caregiver and cancer status on depressive symptoms shows important results not found in previous research. Our results demonstrated that dual-role individuals have higher levels of depressive symptoms than noncaregivers without a history of cancer, caregivers without a history of cancer, and cancer survivors who did not have caregiving responsibilities. Of note, none of the previously published projects (Mahmood et al., 2025; Wang et al., 2024) made these comparisons using validated measures of depressive symptoms. Our results are consistent with Mahmood et al. (Mahmood et al., 2025), who reported over double the odds of psychological distress in dual-role individuals when compared to cancer survivors with no caregiving responsibilities, and with Wang et al. (2024), who found poor mental health twice as common among dual-role individuals compared to individuals who neither had a cancer diagnosis nor were caregivers. However, caution is needed when interpreting the significantly higher levels of depressive symptoms in dual role caregivers versus caregivers without a history of cancer, because the descriptive data on caregiving characteristics shown in Supplementary Table 3 suggests non-significant but somewhat higher levels of some indicators of how much care is being provided by the dual role group. Examples include providing most help or care (64.7% dual role, 55.4% caregiver only), providing care all of the time (22.5% dual role, 15.6% caregiver only), and some or a lot of emotional strain due to providing care (63.7% dual role, 53.9% caregiver only). It is important that future studies on this topic comprehensively evaluate descriptive data for all aspects of provided care. Overall, our results indicate that the combination of cancer survivorship and caregiving responsibilities is associated with higher levels of depressive symptoms than either condition alone or neither condition. This finding is consistent with research showing that multiple stress exposures across the life course additively increased the risk of depressive symptoms and clinical depression (Kok et al., 2021; Vinkers et al., 2014).

Dual-role individuals were more likely to receive high family support, and high family support was significantly associated with reduced depressive symptoms. This indicates that dual-role individuals tend to seek and receive greater family support, which in turn reduces depressive symptoms. This finding aligns with stress-buffering hypothesis, which suggests that family support, as a form of social support, can protect individuals from adverse psychological effects, including potential caregiving stress and the challenges associated with cancer survivorship (Carpenter et al., 2010; Cohen & Wills, 1985; Ho et al., 2025; Manne & Badr, 2010; Yoo et al., 2017). Our findings also indicate that dual-role individuals were likely to receive support from their families. However, unlike prior research suggesting that individuals facing multiple stressors, such as chronic illness, are generally more motivated to seek social support from a broader range of sources, including family, friends, community or healthcare provider; our results showed that only family support played a significant role for dual-role individuals (Cohen & Wills, 1985; Gutiérrez-Sánchez et al., 2023; Reeves et al., 2014; Yoo et al., 2017).

We found that dual-role individuals were less likely to rely on their partners. This dynamic can be particularly relevant for older dual-role individuals, as many provide care for their spouses. High reliance on the care recipient can evoke guilt or anxiety about becoming a burden, especially when the care recipient (partner) simultaneously requires care (Lilleheie et al., 2021; Turner et al., 2024). In addition, caregiver’s own health limitations may increase their reliance on partner (care recipient), diminishing the caregiver’s ability to provide effective support and further intensifying the caregiver’s psychological burden (Lilleheie et al., 2021; Turner et al., 2024). Understanding these issues will likely require greater attention to measurement of the complex circumstances of cancer, including types of cancer, severity of previous cancer symptoms, and whether cancer was life threating. Future studies including such variables will likely further advance this area of research.

Although the associations between cancer-caregiver status and friends’ support and reliance on family were not significant, receiving these forms of support were associated with fewer depressive symptoms. Previously, Yoo et al. (2017) also found that higher perceived social support was linked to lower depressive symptom scores, with these associations being stronger in cancer survivors than among non-caregivers who did not have cancer. In our study, only the association between reliance on family and depressive symptoms was moderated by combined cancer and caregiver status, with the association also being stronger among those in dual roles. To conclude, the findings from our study highlight the importance of maintaining and encouraging supportive social connections, particularly strengthening reliance on partner and family support, for cancer survivors who also take on caregiving responsibilities.

Strengths, limitations, and future directions

The findings from our study provide novel insights into the role of social connection and engagement in depressive symptoms among vulnerable populations of older adults, using nationally representative data from the United States. To our knowledge, this is the first study that examined the detailed aspects of social support and engagement among older dual-role individuals and explained how social relationships may help protect against depressive symptoms in this population. However, several limitations should be noted. First, the NSHAP included a relatively small sample of participants who reported a cancer diagnosis and lacked complete information on the timing and duration of cancer survivorship. Second, although in our study caregivers without a history of cancer did not have significantly different depressive symptoms compared with individuals who had neither cancer nor caregiving responsibilities, previous research has often documented adverse mental health effects associated with caregiving, including increased caregiver burden, anxiety, and depressive symptoms (Haley et al., 1995, 2020; Karimi Moghaddam et al., 2023). However, previous research shows that elevated depressive symptoms in caregivers versus noncaregiving controls were found only among caregivers who reported high strain (Roth et al., 2009). Moderate strain caregivers did not differ in depressive symptoms from noncaregivers, and caregivers reporting no strain reported lower depressive symptoms than noncaregivers. In this study, Roth et al. (2009), 33% of caregivers reported no strain, 49% reported moderate strain, and 18% reported high strain. The present dataset did not set minimum amounts of caregiving or strain to be considered a caregiver. Also, NSHAP did not establish a minimum threshold for caregiving duration or hours provided per week, whereas studies with stricter caregiving criteria have reported stronger associations with adverse mental health outcomes (Haley et al., 2020). Third, the dataset did not include a nonbinary gender category, which may limit inclusivity of gender measurement and introduce potential misclassification. Fourth, the dataset did not include eligibility criteria for caregivers, measures of caregiving intensity, such as assistance with activities of daily living (ADLs) or instrumental activities of daily living (IADLs), nor information about the caregivers’ relationship with the care recipient, both of which may moderate the observed associations. Fifth, NSHAP caregiving status was assessed using a single, broadly worded question, which may have resulted in misclassification by excluding some caregivers (e.g., cancer, other serious illness, and pediatric caregivers) while including individuals who may not meet stricter definitions of informal caregiving. Future studies should collect information on caregiving intensity, so that the association between cancer survivorship and mental health can be determined across levels of caregiving intensity. Of note, many of these limitations, including lack of information about timing and duration of cancer survivorship, and lack of a minimum threshold for caregiving duration or hours of care, were also present in the two previously published studies on this topic (Mahmood et al., 2025; Wang et al., 2024).

We focused on the dual role of caregiving and cancer survivorship because cancer survivorship in caregivers is a topic that has received considerable research attention, and cancer survivorship has been recognized as a prominent public health issue, even led to a National Cancer Institute Office of Cancer Survivorship (Nekhlyudov et al., 2017). However, other highly prevalent health conditions, such as heart disease, also affect many Americans and can lead to chronic disability, distress, and posttraumatic growth (Bluvstein et al., 2013; Martin et al., 2025) that might affect reactions to caregiving, and would also be worthy of study.

Public health practice and policy implications

Our findings underscore the importance of promoting and enhancing social support and social engagement within the population of cancer survivors who are caregivers, as higher levels of all social support and social engagement constructs (except for religious attendance) were associated with lower depressive symptoms. This approach would allow public health interventions to integrate structured support programs that are customized to each population’s specific needs to mitigate the impact of depressive symptoms. These programs could range from support group activities to family-based interventions delivered in clinical and community health settings (Gabriel et al., 2020; Kiemen et al., 2023). Specifically, future intervention studies should focus on increasing social support and social engagement among individuals with dual roles, as this may translate into lower depressive symptoms through activities such as volunteering, socializing with friends or relatives, visiting neighbors, and ensuring support from both family and friends. This study is especially important for advocating policy changes, such as incorporating social prescribing into healthcare systems to benefit cancer survivors, caregivers, and individuals facing a dual burden (Aggar et al., 2025; Connolly et al., 2025; Mulligan et al., 2024). Ensuring customized and targeted interventions for older adult vulnerable populations through community-based social support services can enhance resilience and mental well-being among high-risk groups, ultimately reducing depressive symptoms. (American Psychological Association Help Center, 2004; Lin et al., 2025; Lun, 2024; Ong et al., 2018; Palacio et al., 2020; Resnick, 2014; Rosenberg & Eckstrom, 2023).

Supplementary material

Supplementary material is available at The Gerontologist online.

Funding

This work was supported by the Cancer Epidemiology Education in Special Populations (CEESP) Program through funding from the National Cancer Institute (R25 CA112383 to N.B.).

Conflicts of interest

None declared.

Data availability

The datasets utilized in this study were drawn from the National Social Life, Health, and Aging Project, Round 2 (2010–2011), and are available in the National Archive of Computerized Data on Aging (NACDA-ICPSR) repository at https://harmonize.icpsr.umich.edu/.

Author contributions

Natalia Babenko (Conceptualization [equal], Formal analysis [lead], Funding acquisition [lead], Methodology [equal], Resources [equal], Software [lead], Writing—original draft [equal]), Judith J.M. Rijnhart (Conceptualization [equal], Formal analysis [supporting], Funding acquisition [supporting], Methodology [equal], Resources [equal], Supervision [equal], Writing—original draft [supporting], Writing—review & editing [equal]), Martine Extermann (Conceptualization [equal], Funding acquisition [supporting], Methodology [equal], Resources [equal], Supervision [equal], Writing—review & editing [equal]), and William E. Haley (Conceptualization [equal], Funding acquisition [supporting], Methodology [equal], Resources [equal], Supervision [equal], Writing—review & editing [lead])