Grandparenting, Social Relations, and Mortality in Old Age

Abstract

Guided by a convoy model of social relations, this study investigates the relationships between grandparenting status, social relations, and mortality among community-dwelling grandparents age 65 and older who are caring for their grandchildren. The data were drawn from the 2008 and 2016 waves of the Health and Retirement Study (N=564). Latent class analysis was used to identify the social network structure based on six indicators of interpersonal relationships and activities. A series of four hierarchical Weibull hazard models estimated the associations between grandparent caregiving, social relations, and mortality risk. Results of survival analyses indicate that co-parenting and custodial grandparents had higher all-cause mortality risk than grandparents who babysat occasionally; however, for custodial grandparents, the association was not significant once social relation variables were added to the model. This study suggests that community-based support may be beneficial to older grandparents and that improved relationship quality is integral to the well-being of older adults.

Introduction

Caring for grandchildren is considered one of the most challenging but significant events of later life. According to data from the U.S. Census Bureau, most Americans (83%) age 65 and older have grandchildren (Krogstad, 2015). In the United States, approximately 7 million grandparents lived with a grandchild in 2013, up from 5.8 million in 2000 (Krogstad, 2015). Among those who living with grandchildren, a significant share (37%) also serve as their grandchildren’s primary caregiver (Krogstad, 2015). A growing body of research on the impact of grandparenting examines differences that preceded the care work and co-residence (Baker & Silverstein, 2008). Some grandparents perceive providing childcare to grandchildren not as a burden or disruption to their daily lives, but rather as the most joyful phase of their lives, even more enjoyable than raising their own children (Baker & Silverstein, 2008; Harrington, 2014). However, grandparents who care for their grandchildren experience numerous challenges, which contribute to adverse physical and mental health (Sprang, Choi, Eslinger, & Whitt-Woosley, 2015). As a result of their children’s family crises (e.g., divorce, alcoholism, teenage pregnancy, parental abuse, or abandonment), they often have to accept a parental role and provide safeguards during times of strain (Hayslip & Kaminski, 2005). These situations are more prevalent among grandparents who live with grandchildren. Becoming a grandparent caregiver of a grandchild is also linked to weakened social networks, resulting in an increased sense of social isolation (Landry-Meyer & Newman, 2004). These challenges are related to high levels of stress and burden, particularly if childcare demands are intense and must be balanced with other roles, which can be overwhelming.

A higher level of social integration and more social support are associated with positive health-related outcomes, including fewer depressive symptoms, better self-rated health, improved life expectancy, and lower mortality (Antonucci, Fuher, & Dartigues, 1997; Tay, Tan, Diener, & Gonzalez, 2013); therefore, it is likely that positive social interactions and support from a spouse, family, and friends have a non-negligible favorable impact on grandparents’ health. However, the extent to which grandparenting benefits or harms health, especially mortality, remains unclear (Hilbrand, Coall, Gerstrof, & Hertwig, 2017). Further, despite the widespread provision of grandparental childcare, there is little knowledge of the degree to which grandparents’ social relations affect mortality. To fill this gap in the literature, this study investigates the effects of different aspects of grandparenting and social relations on mortality among community-dwelling older adults.

Grandparents’ Caregiving Role and Social Relations

Over the past few decades, the role of grandparents has identified somewhat controversial because young working families need help they cannot obtain via other resources. While many grandparents choose to care for their grandchildren, others must be pushed to take the role. As grandparents are aging, grandparent caregivers are at an elevated risk of financial strain, poor health, social isolation, role overload, role confusion, stress, and related issues (Whitley, Kelley, & Campos, 2013). Overly taking on primary parental role and intense caregiving demands are significantly associated with increased health risks (Whitley, Kelley, & Sipe, 2001). These health risks are related to, at least in part, adult children because the caregiving role usually involves helping their adult child with his or her problems as well, grandparent caregivers have higher levels of depression and anxiety than their non-caregiver peers (Kelly, Whitley, & Campos, 2013). Further, multigenerational families, which are brought together as a result of adversity of material resource restraints, such as poverty, mental health problems, or substance abuse, often experience conflicts between grandparents and adult children (Wagstaff & Cooper, 2019).

High rates of co-residence, thus, may reflect fewer living arrangement options due to income constraints and related factors. Researchers often assume that those who live with grandchildren share caregiving responsibility with their adult children, although the extent of the care they provide can vary widely (Chen & Liu, 2011). Providing the right amount of the right types of care for grandchildren may leave grandparents rejuvenated and fulfilled; in contrast, providing too much of the more intensive types of care may leave grandparents depleted and exhausted (Meyer & Kandic, 2017).

Accordingly recent studies revealed that grandparents may benefit from a moderate amount of time spent on caregiving. For example, Burn et al., (2014) found that grandmothers who spent one day a week caring for their grandchildren might have a lower risk of developing Alzheimer’s and other cognitive disorders compared to those who care for grandchildren for five days or more per week. Further, studies have found that providing a few hours of care per week was associated with a lower mortality risk (Burns et al., 2016; Hilbrand et al., 2017). Hilbrand et al. (2017) found that older adults who provided non-custodial care for grandchildren had a lower risk of death over a 20-year period than those who did not care for their grandchildren, based on data from the Berlin Aging Study among over 500 older adults (age 70 and over).

However, changes in grandparents’ social roles could be detrimental to their social relations. Although coping with familial difficulties may amplify a grandparent’s social isolation and further interfere with their relationships with others; however, there is very little research on late-life social relationships among grandparent caregivers (Burns, Stevens, & Lee, 2016). The convoy model of social relations postulates that relationships vary in structure (e.g., size, composition, contact frequency, geographic proximity), closeness and quality (e.g., positive, negative), and function (e.g., aid, affection, affirmation exchanges) (Antonucci et al, 2011). This is especially noteworthy given that older adults often have a lifetime of relationships (e.g., siblings, spouse, children), as well as newly emerging relations (e.g., grandchildren); thus, the convoy model can be used to assess new means of forging and maintain relationships and how these relationships influence well-being (Fuller, Ajrouch, & Antonucci, 2020).

Researchers have utilized a person-centered approach to create social network typologies that represent several attributes of an individual’s social network. Rather than identifying the influence of variables, the person-centered approach attempts to combine multiple social indicators into a meaningful typology in order to describe the characteristics of individuals’ networks. According to Fiori et al. (2006), these typologies reflect the complex, multidimensional, and aggregate nature of social life. For example, using data from a sample of 2,079 Jewish people, Litwin (2001) developed a typology of networks to examine social relationships in their complex and aggregate state, emphasizing multiple relationships and their functional specificities. The typologies were based on seven variables: current marital status, number of proximate children, frequency of contact with children, frequency of contact with friends, frequency of contact with neighbors, attendance at religious services, and attendance at social clubs. The analysis indicated five network types: divers, friend, neighbor, family, and restricted. Litwin and Shoivitz-Enra (2010) found that these network types were associated with mortality among adults age 70 or older. Compared to individuals with restricted and family-focused networks, those with diverse and friend-focused networks had a lower risk of mortality.

The convoy model further postulates that each individual is surrounded by a convoy, or a set of people with whom the individual maintains reciprocal emotional and instrumental support (Antonuicci et al., 2011). Hayslip et al. (2014) suggested that ensuring the provision of reliable emotional and instrumental support from friends and family is a key component in efforts to improve the health of grandparent caregivers. Traditionally, positive support is associated with increased longevity. Some studies have found that individuals who receive more positive support from a spouse, a child, or network members survive longer than those who receive less support (Lyyra & Heikkinen, 2006). Research has found that social support is important in the stress process because social relations prevent the development of secondary stressors, such as constriction of social life, poor self-esteem, and loss of self, ultimately improving the caregiving experience for custodial grandparents (Robitaille, Orpana, & Mclntosh, 2012). Penninx and colleagues (1997) found that persons who received a moderate or high level of positive emotional support had lower mortality risk than those who received a low level of emotional support. In contrast, recipients of a high level of instrumental support had a higher risk of death.

Factors Should Be Controlled on Mortality

Analyzing of the effect of social relations on mortality requires accounting for several factors associated with health outcomes that may lead to death. Most importantly, analyses must control for socio-demographic variables related to a higher propensity for death, including age, gender, education, income (Cooper, 2002), and nativity (Idler, Blevins, Kiser, & Hogue, 2017). Ahmad and Bath (2005) confirmed that age was the most important factor affecting mortality among older adults who resided in the community.

The effects of psychobiological pathways should be controlled in analyses of the effects of social relations on mortality (Berkman, Glass, Brissette, & Seeman, 2000). The first pathway is the health behavioral pathway, which includes health promotion and/or risk behaviors. Health- promoting behavior, such as regular physical activity, is positively related to health and inversely related to mortality (Landi et al., 2010). Risk behaviors, in contrast, are negatively associated with health and positively associated with mortality; for example, smoking and drinking alcohol increase mortality risk (Janssen & Kunst, 2005). The second pathway involves psychological factors such as emotional state (e.g., depression). However, the literature on the effect of depression on mortality is mixed. For example, Cuijpers and Smit (2002) found a positive relationship between depression and mortality, while Ben-Ezra and Shmotkin (2006) found no association between them after controlling for health. The third pathway is the physiologic pathway, which involves morbidity or major illness. Morbidity or major illness, namely cancer, diabetes, heart attack, and stroke, increases mortality risk in later life (Schupf et al., 2005).

In this study, the convoy model offers a broad perspective of the development of social relations, facilitating an analysis of how the structure of a social network, function, and quality of relationships influence mortality risk after controlling for psychological, behavioral, and physiological pathways. This perspective is helpful for examining the situation of grandparent caregivers. Thus, our model further posits that grandparenting status in mortality may be more directly explained by factors of social relations as additional mechanisms that may explain the association between grandparenting and risk of mortality. We hypothesized that grandparent caregivers who do not live with grandchildren, have a diverse/friend network, less instrumental support, and more positive support are less likely to have mortality risk.

Method

Data and Sample

The data were drawn from the Health and Retirement Study (HRS), a nationally representative sample of men and women over age 50 in the United States. The HRS consists of six birth cohorts who entered the study in different calendar years. Respondents were interviewed every two years once they had entered the study. According to the HRS study design, a new birth cohort is added every 6 years to maintain a representative sample of the population. Initial cohort response rates ranged from 70 percent to over 80 percent across waves; re-interview rates for all cohorts at each wave were between 92 percent and 95 percent (Health and Retirement Study, 2011). The HRS gathers in-depth economic, financial, and health information from respondents, and oversamples African Americans and Latinos. This study used data from 2008 when the psychosocial questionnaire was introduced as the baseline. The sample was limited to respondents who reported actual grandparenting hours and completed the psychosocial questionnaire. In addition, the study excluded persons younger than 65 at the time of data collection and those who lived in a nursing home. The final weighted sample included 564 respondents age 65 and older. The 2016 mortality reported that 102 respondents in the original analytical sample (18.09% of the sample) had died.

Measures

Grandparent caregiving

HRS asked respondents, “Did you spend 100 or more hours in the last two years taking care of grandchildren?” Respondents who answered “yes” were then asked how many hours they had spent on grandchild care and whether they lived with their grandchildren. The actual hours of grandparenting was only reported for those who do not live with grandchildren. The grandparent caregiving variable included four categories based on the caregiving hours reported and living arrangements with grandchildren (Fuller-Tomson & Minkler, 2001): 1) occasional babysitting (less than 10 hours per week), 2) intermediate/ extensive babysitting (10 or more hours per week), 3) custodial (living with a grandchild/grandchildren with no adult child present), and 4) co-parenting (living with at least one adult child and a grandchild/grandchildren).

Mortality

Morality was derived from the HRS linkages to the National Death Index (NDI). Since the 2000 wave, the HRS has created linkages to the NDI following each wave. The publicly released file is then updated with the match status, month of death, and year of death for verified matches (HRS, 2011). Surviving participants were right censored and survival time was measured by the number of months participants survived between 2008 and 2016.

Social network typology

Guided by previous literature (Fiori et al., 2006; Kim, Frediksen-Goldsen, Bryan, & Muraco, 2017), ten criterion variables were used: six principal social network indicators, marital status (1=married or living with a partner; 0=other), attendance at religious services (from 1= not at all to 5= more than once a week), and attendance at meetings such as political, community, or other interest groups (from 1=never to 6=more than once a week). The number of children in a close relationship was coded as 0 through 6 (with the highest category including those with 6 or more children). Average frequency of contact with children and friends was measured by three questions: How often do you meet up/ speak on the phone/ write or email with your children (friends)? For each question, scores ranged from 1 (less than once a year or never) to 6 (three or more times a week). The scores were summed and averaged (range: 1–6). Latent class analysis (LCA) with iterative updating was used to identify homogeneous groups within the study sample (Litwin & Shiovitz-Ezra, 2010). Missing data (e.g., 18% of friend contact frequency responses) were handled via full-information maximum likelihood (FIML). LCA was conducted using Stata plug-in.

Using LCA, we compared solutions ranging from a 2-cluster model to a 7-cluster model to find the optimal number of clusters. The optimal cluster selection will have a low BIC and high entropy values (Lanza, Flaherty, & Collins, 2003). Table 1 indicates that the 4-cluster solution was optimal and analyses of the six indicators yielded four latent network types at baseline (non-friend, family, friends, and diverse). The non-friend network type (26.97%) had low numbers of close children, low contact frequency with children/friends, and low attendance at religious meetings and organized group meetings. The family-focused network (29.98%) had high rates of being married and a high number of close children. Finally, both the diverse network type (18.37%) and the friends-focused network type (26.86%) were characterized by high numbers of and frequent contact with children and friends, while the diverse network type also featured the more frequent attendance at organized meetings (see Table 2).

Table 1.

Fit Statistics for Latent Class Analysis (LCA) Solutions with 2–7 Clusters

Model	AIC	BIC	Entropy
2 classes	30361.488	30489.939	.672
3 classes	30312.896	30487.594	.684
4 classes	29735.899	30003.079	.719
5 classes	30021.233	30242.171	.669
6 classes	29706.188	30019.610	.602
7 classes	29680.676	30040.341	.468

Note. AIC= Akaike’s Information Criterion (lower values indicate better fit); BIC= Bayesian Information Criterion (lower values indicate better fit).

Table 2.

Profiles of Social Network Types

Indicator	Total	Social Network Types				Statistics
		Cluster 1: Non-friends	Cluster 2: Family	Cluster 3: Friends	Cluster 4: Diverse
Structure Married (%)	62.5%	61.19	74.09	53.26	59.28	χ2= 28.60**
Number of close children	2.45	1.58	3.37	2.25	2.23	F= 179.79***
Contact with children	4.02	3.17	3.87	4.82	3.97	F= 171.05***
Contact with friends	3.64	2.66	3.19	4.41	4.04	F=158.17***
Religious services attendance	3.16	2.32	3.56	3.41	3.08	F= 118.69***
Meeting attendance	1.34	1.24	1.25	1.26	1.62	F= 541.18***
frequency (%)		20.76%	29.98%	26.86%	22.40%

Notes. The number of close children/friends (0=0; 6= 6 or more), Average of child contact and average of friend contact (1= less than once a year or never; 6=three or more times a week), Attendance of religious services (1=not at all; 5= more than once a week). Attendance of meetings (1=not at all; 6= more than once a week). Significance levels are denoted as

^*<.05,

^**<.01,

^***<.001.

Perceived positive support and negative strain

Participants rated the positive and negative aspects of their relationships with their spouse/partner, children, and friends on a 4-point scale from 1 (not at all) to 4 (a lot). Positive relationship quality included three items: How much do they understand the way you feel about things? How much can you rely on them if you have a serious problem? How much can you open up to them if you need to talk about your worries? Negative relationship quality included four items: How often do they make too many demands on you; criticize you; let you down when you are counting on them; get on your nerves. An index of positive social support (Cronbach’s α= .85) and an index of negative social strain (Cronbach’s α=.86) were created by averaging the summed scores.

Received instrumental support

This variable measured the number of IADL (Instrumental Activities of Daily Living) for which the respondent received instrumental support. The HRS asks about five IADLs: using the phone, preparing meals, grocery shopping, managing money, and managing medications (range 0–5).

Covariates

Psychobiological pathways included health promotion/risk behavior, emotional state, and physiologic pathways (Litwin, 2007). Health promotion and risk behavior were measured by physical exercise, current smoking, and alcohol use. Physical exercise (taking part in moderate activities and sports, such as gardening, walking, dancing, stretching, etc.) was re-coded on a 4-point scale (1=hardly ever or never, 2=one to three times a month, 3=once a week, 4=more than once a week). Current smoking was coded as yes (1) or no (0). Alcohol use was measured by the self-reported number of days per week the respondent usually drank. Emotional state was measured by a subset of eight items from the standard Center for Epidemiologic Studies Depression scale (CES-D). Respondents were asked about their depressive affect (e.g., I felt depressed), well-being (e.g., I was happy), and somatic symptoms (e.g., I could not get going). Participants reported whether they had experienced each of the symptoms much of the time in the past week (yes/no). A summary score was created by summing the number of symptoms (range: 0–8; alpha=.78). Finally, physiologic pathways were measured in terms of morbidity as diagnoses of diabetes, cancer, coronary heart disease, stroke, and psychiatric distress (except depression). Each disease was measured as a dichotomous variable (0=no, 1=yes). Socio-demographic variables included age (in years), gender (0=male, 1=female), household income (1=less than $25,000, 2= $25,000 or more), education (1= high school graduate or less, 2= some college or more), and nativity (1=US-born, 2=born outside the United States).

Statistical Procedures

Descriptive statistics were weighted using person-level sampling weights. Parametric hazard models with Weibull distribution were used to estimate the hazard ratios and 95% confidence intervals. The parametric models produced more efficient estimates than semiparametric (Cox) hazard models (Luo, Hawkley, Waite, & Cacioppo, 2011). The preliminary analyses using Cox models produced generally the same patterns (see appendix Table A).

A series of four hierarchical Weibull hazard models estimated the associations between the study variables and mortality risk. Model I included grandparent caregiving, and Models II-IV sequentially added variables related to socio-demographic characteristics, social relationships, and psychobiological pathways. The social relation variables, in particular, had a relatively high rate of missing values. The missing rates for the items measuring perceived positive support and negative strain were 17 percent (n=95) and 12 percent (n=65). Multiple imputation was employed to handle the missing data (Little & Rubin, 2002). Twenty datasets with no missing data were created and statistical analyses were conducted with each of these imputed datasets. Statistical analyses were conducted using Stata 16.

Results

Table 3 presents descriptive statistics by grandparent caregiving group for the sample at baseline. Intermediate/extensive babysitting grandparents were older (M=72.31, SD= 5.31) than other groups. Among custodial grandparents, the percentages of non-Hispanic Black (23%) and Hispanic (15%) respondents were significantly higher than in other groups. In addition, there was a significantly larger proportion of non US-born respondents among co-parenting (16%) and occasional babysitting (10%) grandparents. The proportion of participants with a high level of education was significantly lower in the custodial (13%) and co-parenting (26%) grandparent groups than in the occasional (30%) and intermediate/extensive (37%) babysitting groups. Custodial and co-parenting grandparents reported worse chronic conditions and more depressive symptoms than both occasional and intermediate/extensive babysitting grandparents. Further, the former two groups had high rates of current smoking and low rates of physical exercise. With regard to social network types, custodial grandparents were the group least likely to be in the friend-focused category (14%) and most likely to be in the non-friend category. Co-parenting grandparents were the group least likely to be in the diverse category (12%).

Table 3.

Descriptive Statistics: Group Difference at Baseline (N=564)

	Occasional babysitting grandparents (n=301, 55.34%)		Intermediate/ extensive babysitting grandparents (n=98, 17.34%)		Custodial grandparents (n=70, 12.49% )		Co-parenting grandparents (n=95, 16.82%)		p-value
	M(SD)/n(%)	Range	M(SD)/n(%)	Range	M(SD)/n(%)	Range	M(SD)/n(%)	Range
Time to death (Months) after T1	100.08 (22.90)	1–108	97.60 (26.24)	1–108	96.99 (24.85)	6–108	96.16 (23.64)	1–108	p =.055
Age at T1	71.24 (4.99)	65–89	72.31 (5.31)	65–86	71.57 (6.30)	65–90	69.38 (3.59)	65–82	p =.001
Female	170 (56.4)	__	55 (56.29)	__	40 (57.08)	__	51 (54.1)	__	p =.985
Race
White	249 (82.58)		79 (80.50)		42 (59.29)		65 (68.13)
Black	25 (8.43)		11 (11.02)		16 (22.77)		13 (13.88)
Hispanic	17 (5.77)		7 (7.30)		10 (14.74)		11 (11.68)
Other	10 (3.22)		1 (1.10)		2 (3.2)		6 (6.31)
Non US-born	29 (9.78)		4 (3.75)		2 (3.25)		16 (16.36)
Income ($25,000 or more)	229 (75.93)	__	71 (72.89)	__	43 (61.68)	__	64 (67.80)	__	p =.205
Education (Some college or more)	92 (30.48)	__	36 (36.55)	__	9 (13.29)	__	24 (25.66)	__	p =.043
Health
Diabetes	68 (22.49)	__	20 (20.4)	__	28 (39.83)	__	28 (29.74)	__	p = .036
Cancer	50 (16.73)	__	22 (22.8)	__	22 (31.62)	__	13 (13.82)	__	p = .029
Heart diseases	74 (24.58)	__	23 (23.62)	__	33 (46.6)	__	36 (37.97)	__	p = .004
Stroke	14 (4.81)	__	4 (4.04)	__	14 (19.63)	__	7 (10.06)	__	p = .002
Psychiatric Distress	48 (15.84)	__	6 (6.1)	__	11 (16.3)	__	17 (17.54)	__	p = .169
Depression	1.10 (1.62)	0-8	1.27 (1.89)	0-7	1.91 (2.48)	0-8	1.35 (1.55)	0-8	p = .010
Alcohol use per week	1.42 (2.29)	0-7	.95 (2.04)	0-7	.43 (1.25)	0-7	1.08 (1.96)	0-7	p = .008
Current smoking	27 (9)	__	8 (7.67)	__	25 (35.91)	__	16 (17.17)	__	p = .000
Exercise	3.24 (1.10)	1–4	3.11 (1.32)	1–4	2.89 (1.33)	1–4	2.61 (1.19)	1–4	p =.000
Social Network
Non-friends	47 (15.57)	__	12 (12.64)	__	28 (40.10)	__	30 (31.23)	__	p =.000
Family	83 (27.74)	__	31 (31.39)	__	19 (27.45)	__	36 (37.54)	__
Friends	99 (32.84)	__	24 (24.90)	__	10 (14.02)	__	18 (19.45)	__
Diverse	72 (23.85)	__	31 (31.08)	__	13 (18.43)	__	11 (11.77)	__
Instrumental support aid	.22 (.71)	0–5	.24 (.83)	0–5	.31 (.62)	0–4	.34 (.64)	0–3	p =.482
Support Quality
Positive social support	3.00 (.68)	1–4	2.91 (.70)	2–4	2.77 (.60)	2–4	2.93 (.59)	2–4	p =.121
Negative social strain	1.41 (.54)	1–4	1.39 (.52)	1–3	1.37 (.52)	1–3	1.55 (.50)	1–3	p =.148

Notes. Weighted results were reported.

Table 4 presents the results of the Weibull hazard models. Relative to occasional grandparents, co-parenting grandparents (HR=1.80, p< .05, 95% CI= 1.04, 3.14) had a significantly higher mortality risk and custodial grandparents (HR=1.71, p< .10, 95% CI= .97, 2.97) had a higher mortality hazard (although the association only reached a marginal level of statistical significance) over the focal period. When socio-demographic variables were added in Model II, being older and identifying as “other race” were significantly associated with higher mortality risk, while being female, having a high level of education, and being born outside of the United States significantly associated with lower mortality risk. Socio-demographic measures increased the hazard ratio for co-parenting grandparents (HR=2.02, p< .05, 95% CI= 1.15, 3.54). Model III added measures of social relations. Being in the diverse network group reduced the hazard ratio of mortality by 59% (HR=.41, p< .05, 95% CI= .20, .82) relative to being in the non-friend group. Perceived positive support reduced the hazard ratio of mortality by 28%, although the effect was only marginally significant (HR=.72, p< .10, 95% CI= .52, 1.01), while receiving instrumental support aid predicted a higher mortality risk (HR=1.38, p< .05, 95% CI= 1.14, 1.68). The hazard ratio for co-parenting grandparents remained significantly different than the ratio for occasional babysitting grandparents (HR=2.15, p< .05, 95% CI= 1.20, 3.85).

Table 4.

Hazard Ratios of Mortality from 2008 to 2016 (N=564)

	Model I		Model II		Model III		Model IV
Variables	Hazard ratio	95% CI	Hazard ratio	95% CI	Hazard ratio	95% CI	Hazard ratio	95% CI
Grandparent caregiving status (ref: Occasional babysitting grandparents)
Intermediate/ extensive babysitting grandparents	1.20	[.69, 2.08]	1.25	[.72, 2.18]	1.11	[.63, 1.95]	1.21	[.67, 2.20]
Custodial grandparents	1.71†	[.97, 2.97]	1.65†	[.93, 2.93]	1.55	[.84, 2.86]	1.18	[.60, 2.30]
Co-parenting grandparents	1.80*	[1.04, 3.14]	2.02*	[1.15, 3.54]	2.15*	[1.20, 3.85]	1.89*	[1.05, 3.42]
Socio-demographic
Age			1.10***	[1.06, 1.14]	1.10***	[1.04, 1.14]	1.10***	[1.05, 1.15]
Female			.49**	[.32, .76]	.47**	[.35, .87]	.49**	[.31, .78]
Race (ref: White)
Black			1.12	[.96, 1.14]	1.24	[.72, 2.15]	1.26	[.41, 1.33]
Hispanic			1.21	[.61, 2.40]	1.23	[.59, 2.54]	1.35	[.63, 2.73]
Other			3.21*	[1.09, 9.41]	3.06*	[1.32, 9.46]	3.26*	[1.65, 9.74]
Non-US born			.34*	[.12, .99]	.28*	[.10, .81]	.27*	[.09, .80]
Some college or more			.57*	[.33, .98]	.64	[.36, 1.12]	.67	[.38, 1.20]
$25,000 or more			.88	[.55, 1.39]	1.04	[.64, 1.72]	1.19	[.72, 1.98]
Social Relations
Social Network (ref: Non-friends)
Family					.69	[.39, 1.20]	.73	[.41, 1.34]
Friends					1.01	[.57, 1.79]	1.14	[.62, 2.12]
Diverse					.41*	[.20, .82]	.38*	[.18, .82]
Instrumental support aid					1.38**	[1.14, 1.68]	1.26*	[1.01, 1.55]
Perceived positive support					.72†	[.52, 1.01]	.70*	[.50, .98]
Perceived negative strain					1.10	[.77, 1.58]	1.07	[.75, 1.54]
Health Pathways
Current smoking							1.87*	[1.06, 3.32]
Alcohol use per week							1.09†	[.99, 1.21]
Exercise							.84†	[.69, 1.01]
Depression							1.01	[.89, 1.15]
Diabetes							1.56†	[.98, 2.49]
Cancer							1.64*	[1.03, 2.62]
Heart diseases							1.39	[.89, 2.17]
Stroke							1.50	[.77, 2.92]
Psychiatric distress							1.03	[.54, 1.98]
F (df), Prob>F	5.91 (3)		53.99 (11)***		80.18 (17)***		104.03 (26)***

Note. Significance levels are denoted as

^†< .10,

^*< .05,

^**<.01,

^***< .001.

After adding psychobiological pathway variables in Model IV, co-parenting grandparents remained at a significantly higher risk of mortality than occasional babysitting grandparents (HR=1.89, p< .05, 95% CI= 1.05, 3.42). In terms of health behavior, current smokers (at baseline) (HR=1.87, p< .05, 95% CI= 1.06, 3.32) had a higher risk of 8-year mortality. Both usual alcohol use per week (HR=1.09, p< .10, 95% CI= .99, 1.21) and exercise (HR=.84, p< .10, 95% CI= .69, 1.01) were marginally significant predictors of mortality. A cancer diagnosis (HR=1.64, p< .05, 95% CI= 1.03, 2.62) predicted a higher mortality risk. Importantly, the social relation variables retained their significant effect in the final model, despite the addition of the psychobiological pathway variables. Diverse network type (HR=.38, p< .05, 95% CI= .18, .82) and receiving instrumental support aid (HR=1.26, p< .05, 95% CI= 1.01, 1.55) remained significant predictors of death 8 years later. In the final model, perceived positive support emerged as an additional statistically significant predictor of mortality (HR=.70, p< .05, 95% CI= .50, .98).

Discussion

This study examined the association between grandparent caregiving and mortality hazards, accounting for the effects of social relations. Physiobiological pathways that may lead to death were controlled in the analyses. Researchers have long known that helping behavior is associated with reduced mortality risk. Recent research has examined the relationships between grandparent caregiving and mortality hazards and has assessed whether grandparents who provide babysitting for their grandchildren have a lower mortality risk than non-caregiving grandparents and non-grandparents (Hilbrand et al., 2017). However, because grandparent caregivers vary in their custodial status, living arrangement, and the amount of care provided to grandchild(ren), these prior analyses do not necessarily show that all grandparent caregiving is beneficial to survival. Further, there is little knowledge about grandparent caregivers’ social relations. Thus, this study extends previous research by examining the associations between various grandparenting status, social relations, and mortality risk.

Grandparenting and mortality

The results of this study reveal that grandparents who reside with their grandchildren experience increased mortality risk relative to grandparent caregivers who do not reside with grandchildren. In particular, co-parenting grandparents had an increased mortality risk even after controlling for social relationships and physiobiological pathways. Thus, the results suggest that co-residence may affect family relationships and intensify caregiving burden or stress. As Goodman and Silverstein (2006) reported, subtle aspect of family relationships, such as grandmother- grandchild closeness and grandmother-parent closeness, can become a crucial factor that influences caregivers’ health. Co-parenting grandparents were particularly likely to report stress related to living with their adult children (Musil & Standing, 2005). Although it is often assumed that custodial grandparents have the largest burden of care, the complex system of parental and grandparental involvement should be addressed in examinations of issues related to grandchild care provision. In addition, future research should examine various factors related to stress among grandparents who co-reside with grandchildren.

We found that custodial grandparents had a smaller increase in morbidity risk than co-parenting grandparents (both relative to occasional babysitting grandparents). Custodial grandparents are viewed as “part of a continuum of care that ebbs and flows with the needs and problems in the middle generation” (Baker, Silverstein, & Putney, 2008, p. 60), indicating they may become co-parenting grandparents if their adult children return to the home or become more involved in child care. During these complex familial transitions, many co-parenting grandparents take on a large share of parental responsibility despite parental presence in the household (Mutchler & Baker, 2004). In some cases, co-residing parents may be unable or unwilling to effectively contribute to parental responsibilities. For example, grandparents with a child who has AIDS share substantial responsibility for their grandchildren while the parents are in the advanced stages of the disease (Cowgill et al., 2007).

There is no clear boundary between custodial and co-parenting grandparent caregivers. Custodial grandparents may still share parental responsibility with the parent, even though the parent does not reside in the same house. Data from the U.S. Decennial Census reveal that a significant portion of custodial grandparents do not claim primary responsibility for their co-resident grandchildren (Mutchler & Baker, 2004). According to Baker et al. (2008), nearly two thirds of custodial grandparents reported having at least daily contact with the parent of their grandchildren. The presence of a parent in the household is not directly associated with the provision of childcare, whereas their absence may not completely impede support from the adult child. The living arrangements of parents in these families are often fluid, and grandparents are likely to move in and out of their caregiver role depending on the needs of their adult children and grandchildren (Lee, Ensminger, & LaVeist, 2005). Further, grandparents who face dwindling savings or diminishing health may co-reside with their adult children to gain assistance with finances or personal care (Ellis & Simmons, 2014). Grandparent caregiving may be more stressful for co-parenting grandparents with poorer health and fewer financial resources.

Social relations and mortality

The classification of network type allows a comprehensive consideration of the interpersonal environments of older grandparents in relation to health outcomes. Future research should include in-depth examinations of the non-friend network. The grandparents in this group were generally unmarried and lacked both contact with friends and organizational participation. Individuals with fewer social connections and less organizational participation may be less able than others to buffer the physiological and health impacts of the challenges of social life (Smith & Christakis, 2007). Therefore, a friendship network could be a significant asset in the reduction of all-cause mortality rates among older adults (Aida et al., 2012). Public investment in promoting the development of positive social networks would likely reduce negative health outcomes and mortality risk among older adults.

The results indicate that grandparents with a diverse social network may have a lower mortality risk than those with a non-friend network. A support network has a palliative effect on the emotional well-being of grandparent caregivers by buffering the deleterious effects of caregiving stress (Gerard, Landry-Meyer, & Roe, 2006). If caregivers have no one to help shoulder the caregiving burden, their stress may increase. Diverse exchanges of informal (from a friend or family member) and formal (via a contractual or paid arrangement) support are a component of intergenerational relations (Antonucci et al., 2011). Interventions designed to develop new social network linkages may be beneficial when the existing network is small, overburdened, or unable to provide effective support. Most often new ties are introduced in response to a major life transition or specific stressor. Heaney and Israel (2008) proposed self-help or mutual aid groups that provide a new set of network ties when members are facing a common stressor or want to bring about similar changes, either at the individual level or at a community level. Such groups could be particularly effective for participants who cannot mobilize social support from their social relationships.

In addition, this study found that positive social support may reduce mortality risk. Positive relationships with family members (adult children and grandchildren) may be particularly important for maintaining health and well-being for co-residing grandparents, and may do so in several ways, including increasing feelings of meaningfulness and feelings of belonging (Park, 2009). Positive interactions with others help grandparents feel emotionally supported and provide assistance during life changes via personal encouragement and enhancement (Hatfield, Hirsch, & Lyness, 2013). The provision of perceived emotional and instrumental support from friends and family is essential for the development of self-care skills (Hayslip et al., 2014) and resilience (Dolbin-MacNab, Roberto, & Finney, 2013).

Grandparent caregivers, especially co-residing grandparents, can obtain an array of support services, including home-based visitation services, case management, respite care, health services, support groups, parenting classes, legal assistance, and material aid (Whitley, Kelley, & Campos, 2013). These services are often packaged in the form of community-based interventions. A community-based health intervention is a good example. Drawing on the resilience model of family stress, adjustment, and adaptation (Cohon, Hines, Cooper, Packman, & Siggins, 2003), Kelley and colleagues (2013) examined the efficacy of a community-based intervention to improve the health of caregiving grandparents. After the intervention, the grandparents had greater knowledge about health behaviors, improved access to health resources, and improved self-care health practices (Kelley et al., 2013). This study indicated that community-based interventions tailored to grandparents’ needs may be effective in ameliorating the stresses that results from parenting demands and affects grandparents’ adaptation to the demands of raising grandchildren.

Study limitations

There are several limitations of the study that should be noted. First, the classification of custodial and co-parenting grandparents is somewhat inexact because the data do not identify the parent of the co-residing grandchild. Some households of co-parenting grandparents in the analysis may contain other relatives (i.e., an uncle or aunt) but not the parent. Second, the analysis excluded respondents with missing data on actual grandparenting hours and those under age 65. This study only included respondents who reported caregiving hours in order to capture their caregiving status. A number of respondents did not report the actual hours of care provided even though they responded that they did provide grandchild care. We did not group these grandparents in a separate category because that category would likely include different types of grandparent caregivers. Next, the HRS does not contain information on other factors that have a strong relationship with caregiving, such as the type of care provided, number of grandchildren, relationship between grandparents and grandchildren (emotional closeness and conflict), and grandchildren’s ages; future studies should include these measures . The age of grandchildren can serve as a proxy for the amount of care needed because caregiving activities change with the developmental state of a grandchild. For example, those who care for younger grandchildren (age 0–6) are engaged in most the physically demanding tasks while those who care for an adolescent or older grandchild provide minimal physical care. Although the current study design includes grandparent caregiving intensity based on reported caregiving hours and living arrangements, the factors described above (e.g., type of care provided) should be assessed and considered in future conceptualization of grandparent caregiving.

In addition, those who were in one type of grandparent caregiving status might have transferred to another caregiving status during the study’s observation period. Additional analyses indicate there were no significant differences in the sample over time (see appendix Table B), but future studies would benefit from considering this factor. Moreover, a significant proportion of grandparents who co-reside with grandchildren are younger than the respondents in the focal sample, and their experiences were not captured in the current study. Further, data on the extent of co-residing grandparents’ involvement in childcare are not available, although it is often assumed that these grandparents provide extensive care for grandchildren. While the literature has shown that co-parenting grandparents also have primary responsibility for the grandchildren, suggesting that both circumstances (co-residence and custodial care) exert considerable stress on grandparents (Szinovacz, DeViney, & Atknison, 1999), some studies have reported that custodial grandparents are at higher risk of social isolation and elevated emotional distress than co-parenting grandparents (Pruchno & McKenney, 2000). Comparisons between custodial and co-parenting grandparents would provide valuable information about these two types of caregiving families, especially differences in the extent of caregiving.

Conclusion

Despite these limitations, this study found that among older grandparent caregivers, custodial and co-parenting grandparents who live with their grandchildren had significantly higher rates of mortality than grandparents who babysat occasionally. Although altruistic behavior in general has been shown to help increase an individual’s well-being, extensive caregiving responsibility might be detrimental to older grandparent caregivers. Positive relationships with family members may help grandparents increase their well-being. Further, having frequent contact with family, friends, and neighbors may decrease the risk of mortality. Due to the complexity of social relationships, further research is needed to quantify and qualitatively describe grandparents’ relationships with friends, neighbors, and family.