Caregiving and risk of mortality and functional decline in white and black elderly adults: findings from the Health ABC study

Abstract

Context

Although caregivers report more stress than non-caregivers, few studies have found greater health decline in older caregivers. We hypothesized that caregivers may be more physically active than non-caregivers, which may protect them from health decline.

Objective

To evaluate total, and race- and gender-specific risk of mortality and functional decline in elderly caregivers versus non-caregivers, and whether these associations were mediated by total physical activity (including daily routine, leisure-time exercise, and caregiving activity).

Design and setting

The Health, Aging and Body Composition (Health ABC) study, a cohort study of 3075 healthy adults, aged 70–79 years in 1997–1998 who resided in Memphis, Tennessee or Pittsburgh, Pennsylvania and were followed through their eighth year of participation.

Participants

Participants were classified as caregivers (n=680, 22%) or non-caregivers (n=2369) if they reported providing “regular care or assistance for a child or a disabled or sick adult”.

Main Outcome Measure

All-cause mortality and incident mobility limitation, defined as reported difficulty walking ¼ mile or climbing 10 steps on two consecutive semi-annual contacts.

Results

Overall, 20.6% of caregivers died and 50.9% developed mobility limitation, versus 22.0% and 48.9% of non-caregivers, respectively. Associations with health outcomes differed by race and gender. Mortality and mobility limitation rates were 1.5 times higher in white caregivers compared to non-caregivers (e.g., among white females, adjusted hazards ratio for mortality, HR = 1.6, 1.0–2.5), but were lower in black female caregivers versus non-caregivers (e.g., HR for mortality = 0.9, 0.5–1.4). Physical activity mediated these associations in most race-gender groups. High-intensity caregivers (i.e., spending ≥ 24 hours/week caregiving) had elevated rates of decline when adjusted for physical activity, but lower rates when not adjusted for it.

Conclusion

Older white caregivers have poorer health outcomes than black female caregivers. Physical activity appears to mask the adverse effects of high-intensity caregiving in most race-gender groups.

INTRODUCTION

Most studies of informal caregiving and health outcomes have been guided by psychosocial stress theories^{1, 2}, but this framework may not sufficiently explain the association between caregiving and health decline in older adults. Although many cross-sectional and short-term prospective studies have found poorer immune status in caregivers compared to non-caregivers^3–6, evidence that caregiving increases disease incidence or mortality has been inconsistent^7–9. Three prospective studies found that caregivers to an ill or disabled spouse had elevated rates of all-cause mortality^{10, 11} and fatal- and non-fatal- coronary heart disease (CHD)¹². These associations, however, were limited to spousal caregivers who were stressed by caregiving activities¹¹, and women who performed caregiving tasks for nine or more hours per week¹². In contrast, among women caregivers, stress did not increase the rate of CHD, and caregivers to a friend or non-spouse relative had lower rates of CHD than non-caregivers¹². These inconsistent results suggest the need to examine factors which may influence the association between caregiving and health decline in older adults, particularly race, gender, and the overall level of physical activity of caregivers and non-caregivers.

Incorporating physical activity into studies of caregiving outcomes may provide a more accurate description of the effect of caregiving on physical health. Moderate physical activity protects against the major health outcomes on which caregivers and non-caregivers have been compared - heart disease^13–15 and mortality¹⁶ – as well as incident mobility disability¹⁷. Physical activity is also associated with lower stress¹⁸ and depressive symptoms^{18, 19}, which are uniformly higher in caregivers than non-caregivers⁷. Moreover, elderly caregivers report more overall physical activity than same-aged non-caregivers²⁰, and elderly persons who become caregivers and who continued caregiving had better physical functioning than noncaregivers²¹. These studies suggest that if physical activity is correlated with better health, and healthier elderly persons are more likely to become caregivers and to perform more caregiving activities, then ignoring physical activity would likely underestimate the association between caregiving and health decline. To date, no study has considered the potential mediating effects of overall physical activity on the association between caregiving and health decline.

There also is evidence that the association between caregiving and health decline may differ by race and gender. Caregiving involvement and the familial relationship of caregivers to the care recipient vary by race and gender. Specifically, caregivers who are black^22–24 and women^25–27 spend more hours per week in caregiving activities. Also, black caregivers are more likely to take care of grandchildren²⁸, non-spouse relatives, and friends^{23, 24}, while whites and women are more likely to be the main caregiver for a relative who needs assistance^{26, 27, 29}. Furthermore, caregivers who are white^{22, 24} and women⁷ report higher levels of stress and depressive symptoms than other caregivers. Finally, physiological responses to caregiving differ in these subgroups in that male caregivers have poorer immune response^{30, 31} and are more likely to develop CHD indicators³² than female caregivers, and black women caregivers show more cardiovascular and cortisol reactivity than their white counterparts, despite reporting less perceived stress³³. Thus, comparing results across race-gender categories could identify groups of vulnerable caregivers and suggest mechanisms by which caregiving may affect health decline.

The current study used data from the Health, Aging and Body Composition (Health ABC) study to evaluate the association between caregiving and health decline, measured by all-cause mortality, and incident mobility limitations. Mortality was chosen as an outcome to permit comparisons with previous studies on caregiving and mortality^{10, 11}. Incident mobility limitation was selected because it represents a critical component of quality of life in older adults and is likely a predictor of whether caregivers can continue to participate in caregiving activity. Analyses were performed within race-gender subgroups.

This study tested the following hypothesis: Overall, caregivers and non-caregivers will not differ in rates of all-cause mortality or incident mobility limitations. However, when adjusting for total amount of physical activity, caregivers will have higher rates of these health outcomes than non-caregivers.

METHODS

Study population

The study population consisted of 3075 well-functioning white and black men and women aged 70–79 years participating in the Health, Aging and Body Composition (Health ABC) study. Potential white participants were identified from a random sample of Medicare beneficiaries who lived in designated zip codes areas surrounding Pittsburgh, Pennsylvania and Memphis, Tennessee. Black participants were recruited from all age-eligible residents in these zip codes. Eligibility criteria included self-report of no difficulty walking one-quarter of a mile, climbing 10 steps, or performing basic activities of daily living (ADLs), no reported need of walking aids, no known life-threatening cancers, and no plans to move from the area for three years. The baseline home interview and clinic-based examination took place between April 1997 and June 1998. All respondents provided written informed consent, and all protocols were approved by the institutional review boards at the study sites. Followup clinical examinations or home-based interviews were performed annually, and short telephone interviews were performed at six-months between each annual interview. For the current study, caregiver status and information on co-variates came from the baseline interview.

Measures

Caregiver status and level of caregiving intensity were ascertained from responses to the question “Do you currently provide any regular care or assistance for a child or a disabled or sick adult?” Participants who answered affirmatively were asked how many hours per week they provided that care. Caregivers were categorized according to tertiles of caregiving intensity, based on the sample-based distribution of the number of hours per week they provided care: low (1–7 hours/week), medium (8–23 hours/week), and high (24–168 hours/week).

Mortality

All-cause mortality was documented through August 9, 2005. Survival time was calculated as the number of days from the baseline interview until mortality or the last known date of contact with the participant. All deaths were confirmed by death certificate.

Incident mobility limitation was based on reported difficulty walking ¼ mile or climbing up 10 steps without resting at 2 consecutive semi-annual followup interviews ¹⁷. Because the eligibility criteria for the Health ABC study required respondents to be free of these limitations at baseline, time to incident mobility limitation was calculated as the number of days from the baseline interview to the first interview at which difficulty performing the activity was reported by the respondent or his/her proxy.

Co-variates

Demographic variables included the respondent's age, race (respondents self-identified as white or black), gender, highest education level (post-secondary school versus less education), marital status (married versus other), and Health ABC site.

Living with non-spouse others in household

We classified respondents according to whether they lived with relatives or friends, but not a spouse. Respondents who lived alone, or who lived with a spouse only were categorized as not living with non-spouse others. We created this variable to capture participants who had an opportunity to be caregivers to grandchildren and relatives other than a spouse because they lived in the same household.

Presence of prevalent medical conditions was a dichotomous variable based on the presence of one or more of the following conditions: respondent's self-report of CHD, cerebrovascular disease, congestive heart failure, pulmonary disease, peripheral arterial disease; self-report or taking medications for diabetes, ECG evidence of a myocardial infarction, clinical osteoporosis, or diagnosis of knee or hip osteoarthritis.

Depressive symptoms

The 10-item version of the Center for Epidemiologic Studies-Depressive Symptoms Scale (CES-D) was used³⁴. Response options for each item included four categories, ranging from experiencing the symptom “rarely/none of the time” to “most or all of the time.” Respondents were categorized according to whether their scores were 10 or greater, indicating being at-risk for depression.

Body mass index (BMI) (kg/m2) was determined from measured height and weight obtained at the baseline clinical examination.

Total physical activity (kcal/kg) was measured by a self-report questionnaire designed for Health ABC based on validated physical activity questionnaires³⁵ and described previously^{17, 36}. Participants indicated whether they had performed activities during the past seven days as part of their daily routine and for leisure-time exercise, along with the amount of time spent in each activity. These activities included gardening, heavy chores, light house work, grocery shopping, laundry, climbing stairs, walking for exercise and for other purposes, and moderate- and high-intensity exercise. In addition, time spent in paid work, volunteer work and caregiving activities was documented. A metabolic equivalent value in kilocalories (kcal) per week per kilogram (kg) of body weight was determined for each activity. A total physical activity score was calculated as the sum of the metabolic equivalent values for each activity – those activities performed as part of caregiving as well as those performed during the respondent's daily routine and leisure activities -- and multiplied by the participant's body weight in kilograms.

Physical activity spent in family care activities (kcal/kg) was calculated for respondents who reported that they were a caregiver in the past week, using the methods described above for calculating total physical activity. This measure was calculated in order to separate the amount of physical activity due to caregiving versus non-caregiving activities.

Statistical Analyses

The demographic and health characteristics of caregivers and non-caregivers were compared in the total sample, and within each race-gender group using t-tests for continuous variables and chi-square tests for categorical variables. Cox proportional hazards models were performed to compare rates of mortality and incident mobility limitation between caregivers and non-caregivers. For these analyses, person-time of followup was measured as the number of days from the baseline interview until date of death, date of the first interview of two consecutive reports of a mobility limitation, or date of the last contact for respondents who did not develop these outcomes and who remained in the study throughout the followup period or were lost to followup.

Three separate Cox proportional models were performed for each outcome. The first (Model A) adjusted for Health ABC site and age at baseline only. The second (Model B) adjusted for Health ABC site and all confounders and significant co-variates except for total physical activity. The third (Model C) added the total physical activity variable to Model B. Models B and C were derived by backward elimination procedures³⁷. First, all potential confounders were added to a model that included the caregiver variable and Health ABC site. Then, variables that were not statistically significant and contributed the least were excluded one at a time, as long as their exclusion did not meaningfully change the association between caregiving and the outcome of interest. Hazards ratios (HR) and 95% confidence intervals (CI) were calculated. Non-caregivers were the reference group in each model. The analyses of caregiving intensity included an indicator variable for each level (i.e., low, medium, high) to allow for comparisons with the non-caregivers while simultaneously adjusting for the other levels of caregiving intensity. For example, the indicator variable for “low intensity” caregivers equaled “1” for caregivers who provided 1–7 hours of care per week, and “0” for respondents who were medium – or high- intensity caregivers or non-caregivers.

SAS version 9.1 was used for all analyses³⁸.

RESULTS

Sample characteristics

The baseline sample included 3075 participants, 52% were women, 58% were white, and their mean age was 73.6 years. The sample characteristics are described in more detail elsewhere ³⁹. Approximately 22% of the respondents reported being a caregiver to a child or disabled adult (see Table 1). Black women were most likely to be caregivers (28.8%), and to spend the most time caregiving: Nineteen percent spent eight or more hours a week performing caregiving activities, whereas white men were least likely to be caregivers (18.2%) and less involved in caregiving activities than other race-gender groups.

TABLE 1.

Distribution of caregiver status, mortality and incident mobility limitation according to caregiving status among Health ABC respondents: total sample, and by gender/race group

Sample group	Caregiver status	Sample distribution n %		% died	% incident mobility limitation
Total sample	Caregiver: No Yes	2369 680	77.7 22.3	22.0 20.6	48.9 50.9
	Caregiving intensity:
	Non-caregiver	2369	79.1	22.5	48.9
	Low (1–7 hrs/wk)	218	7.3	24.3	49.1
	Medium (8–23 hours/wk)	191	6.4	23.6	53.4
	High (24–168 hours/wk)	218	7.3	15.1	50.9
White Males	Caregiver: No Yes	762 170	81.8 18.2	21.4 28.2	36.9 43.5
	Caregiving intensity:
	Non-caregiver	762	82.4	21.4	36.9
	Low (1–7 hrs/wk)	60	6.5	30.0	38.3
	Medium (8–23 hours/wk)	52	5.6	34.6	55.8
	High (24–168 hours/wk)	51	5.5	21.6	39.2
White Females	Caregiver: No Yes	662 186	78.1 21.9	14.5 16.1	46.0 53.2
	Caregiving intensity:
	Non-caregiver	662	79.3	14.5	46.0
	Low (1–7 hrs/wk)	70	8.4	17.1	58.6
	Medium (8–23 hours/wk)	50	6.0	20.0	54.0
	High (24–168 hours/wk)	53	6.4	11.3	49.1
Black Males	Caregiver: No Yes	432 116	78.8 21.2	35.4 30.2	53.4 49.1
	Caregiving intensity:
	Non-caregiver	432	80.3	35.4	53.4
	Low (1–7 hrs/wk)	37	6.9	37.8	54.1
	Medium (8–23 hours/wk)	30	5.6	33.3	46.7
	High (24–168 hours/wk)	39	7.2	20.5	48.7
Black Females	Caregiver: No Yes	513 208	71.2 28.8	23.6 13.0	66.9 55.8
	Caregiving intensity:
	Non-caregiver	513	73.5	23.6	66.9
	Low (1–7 hrs/wk)	51	7.3	17.7	45.1
	Medium (8–23 hours/wk)	59	8.5	11.9	54.2
	High (24–168 hours/wk)	75	10.7	10.7	61.3

Overall, caregivers were slightly younger than non-caregivers, and were more likely to be married and/or live with others, and have completed fewer years of education, as shown in Table 2. Caregivers and non-caregivers did not differ in physical or mental health characteristics, including prevalent medical conditions, mean BMI, or high depressive symptoms. However, caregivers in each race/gender group were significantly more physically active than non-caregivers. When kilocalories expended in caregiving activities were excluded from total physical activity, these differences were diminished slightly, but in all subgroups except black men, caregivers remained more physically active than non-caregivers.

TABLE 2.

Baseline characteristics of caregivers and non-caregivers, total sample and by race-gender, Health ABC

VARIABLE	Total sample		White males		White females		Black males		Black Females
	NCG	CG	NCG	CG	NCG	CG	NCG	CG	NCG	CG
	(n=2369)	(n=680)	(n=762)	(n=170)	(n=662)	(n= 186)	(n=432)	(n=116)	(n=513)	(n=208)
Age: mean (std)	73.7 (2.9)	73.4 (2.9) a	74.0 (3.0)	73.6 (2.7)	73.6 (2.8)	73.5 (2.8)	73.5 (2.8)	73.5 (2.8)	73.5 (2.9)	73.1 (3.0)
Education level: % college+	44.0	36.0 c	60.7	55.3	49.0	40.3 a	26.5	23.3	27.5	23.3
% Married	53.8	58.2 a	78.2	81.7	47.1	55.3	59.5	76.5 c	21.5	31.1 b
% live with non-spouse others	20.8	35.1 d	14.7	22.4 b	13.5	23.8 c	30.7	49.1 c	30.9	47.6 d
% prevalent conditions	41.8	41.6	42.0	50.0	38.1	35.0	43.3	44.0	45.2	39.4
% High depressive symptoms	5.5	7.0	4.6	3.5	6.8	9.2	4.3	6.1	6.1	8.3
BMI: mean (std)	27.4 (4.8)	27.5 (5.0)	27.0 (3.7)	26.9 (3.7)	26.0 (4.6)	26.1 (4.4)	27.3 (4.5)	26.8 (4.3)	29.7 (5.8)	29.6 (6.0)
Total physical activity (kcal/kg/week): mean (std)	67.8 (51.4)	136.7 (93.3)d	72.9 (56.9)	134.6 (79.4)d	72.4 (46.8)	132.2 (85.2)d	66.1 (59.2)	141.0(108.3)d	56.0(37.4)	140.3(101.6)d
Total physical activity-family caregiving activity (kcal/kg/wk): mean (std)	67.8 (51.4)	80.0 (58.4)d	72.9 (56.9)	88.4 (61.6)c	72.4 (46.8)	84.5 (58.5) b	66.1 (59.2)	71.6 (58.1)	56.0(37.4)	73.6 (54.9)d

^ap<0.05

^bp<0.01

^cp<0.001

^dp<0.0001

Mortality

As shown in Table 1, 20.6% of the caregivers and 22.0% of the non-caregivers died over the follow-up period. The incidence of mortality in caregivers and non-caregivers differed substantially across race/gender groups. For example, caregivers had a higher unadjusted risk of mortality compared to non-caregivers among white men (28.2% versus 21.4%), but a slightly lower risk among black men (30.2% versus 35.4%), and less than half the risk among black women (13.0% versus 23.6%). Caregiving intensity did not appear to be associated with mortality; in most race/gender groups, high intensity caregivers had a lower unadjusted risk of mortality than other caregivers.

Similar associations were observed in site- and age-adjusted Cox proportional hazards models (see Table 3, Model A). Notably, white male caregivers had significantly higher mortality rates than non-caregivers, whereas black female caregivers had a lower rate. Moreover, in all race/gender groups except for white men, high intensity caregivers exhibited the lowest mortality rates. These associations were not substantially changed when adjusted for site, age and other significant co-variates except for physical activity (Model B). However, adjusting for total physical activity in addition to all significant co-variates resulted in a higher mortality rate among caregivers overall, and for most levels of caregiving intensity for white male and female participants (Model C). Moreover, high intensity caregivers had elevated mortality rates compared to non-caregivers in all race-gender groups except for black men.

TABLE 3.

Caregiver status and mortality, Cox proportional hazards models with and without adjustment total physical activity; total sample and by race-gender group, Health ABC^t

Sample group	VARIABLE	Mortality rate per 1000 p-years	MODEL A: Age-adjusted HR (95% CI)	MODEL B: Significant covariables* HR (95% CI)	MODEL C: Significant covariables + total physical activity HR (95% CI)
Total sample	Caregiver: No Yes	34.0 31.0	1.00 0.92 (0.77–1.11)	1.00 0.91 (0.75–1.10)	1.00 1.15 (0.94–1.40)
	Caregiving intensity:
	Non-caregiver	34.0	1.00	1.00	1.00
	Low (1–7 hrs/wk)	37.1	1.07 (0.81–1.42)	1.10 (0.83–1.47)	1.20 (0.90–1.60)
	Medium (8–22 hours/wk)	35.2	1.05 (0.78–1.43)	0.98 (0.72–1.32)	1.17 (0.86–1.60)
	High (24–168 hours/wk)	22.7	0.70 (0.49–0.99)	0.68 (0.48–0.97)	1.12 (0.75–1.67)
White Males	Caregiver: No Yes	32.2 43.3	1.00 1.39 (1.01–1.93)	1.00 1.32 (0.96–1.84)	1.00 1.58 (1.11–2.24)
	Caregiving intensity:
	Non-caregiver	32.2	1.00	1.00	1.00
	Low (1–7 hrs/wk)	46.5	1.48 (0.91–2.42)	1.43 (0.88–2.33)	1.54 (0.94–2.52)
	Medium (8–22 hours/wk)	52.9	1.61 (0.98–2.62)	1.49 (0.91–2.44)	1.72 (1.04–2.86)
	High (24–168 hours/wk)	33.2	1.19 (0.64–2.19)	1.12 (0.60–2.07)	1.61 (0.81–3.21)
White Females	Caregiver: No Yes	20.9 23.7	1.00 1.15 (0.76–1.73)	1.00 1.19 (0.79–1.80)	1.00 1.61 (1.04–2.51)
	Caregiving intensity:
	Non-caregiver	20.9	1.00	1.00	1.00
	Low (1–7 hrs/wk)	24.6	1.18 (0.64–2.15)	1.21 (0.66–2.22)	1.36 (0.74 – 2.49)
	Medium (8–22 hours/wk)	30.4	1.60 (0.83–3.09)	1.60 (0.83–3.10)	2.17 (1.10– 4.72)
	High (24–168 hours/wk)	16.6	0.81 (0.35–1.84)	0.86 (0.33–1.96)	1.90 (0.75 – 4.86)
Black Males	Caregiver: No Yes	57.4 48.2	1.00 0.83 (0.58–1.20)	1.00 0.81 (0.56–1.17)	1.00 0.92 (0.62 – 1.36)
	Caregiving intensity:
	Non-caregiver	57.4	1.00	1.00	1.00
	Low (1–7 hrs/wk)	63.8	1.13 (0.65–1.95)	1.11 (0.64–1.92)	1.13 (0.65 – 1.96)
	Medium (8–22 hours/wk)	49.6	0.84 (0.44–1.59)	0.77 (0.41–1.46)	0.81 (0.42 – 1.55)
	High (24–168 hours/wk)	32.6	0.56 (0.27–1.14)	0.55 (0.27–1.11)	0.67 (0.30 – 1.48)
Black Females	Caregiver: No Yes	35.8 19.1	1.00 0.54 (0.35–0.81)	1.00 0.54 (0.36–0.83)	1.00 0.87(0.54 – 1.40)
	Caregiving intensity:
	Non-caregiver	35.8	1.00	1.00	1.00
	Low (1–7 hrs/wk)	26.9	0.72 (0.37–1.42)	0.82 (0.42–1.63)	0.99 (0.50 – 1.97)
	Medium (8–22 hours/wk)	17.1	0.48 (0.22–1.02)	0.47 (0.22–1.00)	0.73 (0.33 – 1.61)
	High (24–168 hours/wk)	15.4	0.44 (0.22–0.91)	0.44 (0.22–0.90)	1.23 (0.51 – 2.94)

^tAll models adjusted for Health ABC site

^*Covariables that were included after backward elimination in Models B and C included: Total sample: age, race, gender, prevalent diseases, BMI, education (in Model B but not Model C) White males: age, prevalent diseases, BMI (in Model C but not Model B White females: age, prevalent diseases Black males: education, prevalent diseases, BMI (in Model C but not Model B) Black females: prevalent diseases

Functional decline

The pattern of associations between caregiving and incident mobility limitation was similar to that of caregiving and mortality (see Tables 1 and 4). In site- and age-adjusted analyses, white male and female caregivers had higher rates of incident mobility limitation than their non-caregiver counterparts. In contrast, caregiving appeared to be protective in black female caregivers and was not associated with decline in black male caregivers. In addition, caregiving intensity was not associated with developing mobility limitations. Adjusting for significant confounders except for physical activity had minimal effects on these results (see Model B). However, when adjusted for total physical activity and other confounders (Model C), white male and female caregivers had significantly higher rates of incident mobility limitation than non-caregivers, but the apparently protective effect of caregiving among black females was attenuated. As with the associations with mortality, high-intensity caregivers had elevated rates of incident mobility limitation compared to non-caregivers in all race-gender groups: the hazards ratios ranged from 1.32 (95% CI = 0.89–1.97) among black females to almost 1.5 in each of the other race-gender groups. For the total sample, when adjusting for physical activity and other co-variates, there was a non-statistically significant trend toward a dose-response association between intensity of caregiving and incident mobility limitation, ranging from HR = 1.09 (95% CI = 0.90−1.33) among low-intensity caregivers to 1.45 (95% CI = 1.14−1.83) among high intensity caregivers.

TABLE 4.

Caregiver status and incident mobility limitation, Cox proportional hazards models with and without adjustment for total physical activity: total sample and by race-gender group, Health ABC^t

Sample group	VARIABLE	Incidence rate per 1000 p-years	MODEL A: Age-adjusted HR (95% CI)	MODEL B: Significant covariables HR (95% CI)*	MODEL C: Significant covariables + total physical activity HR (95% CI)
Total sample	Caregiver: No Yes	116.8 123.4	1.00 1.07 (0.95–1.20)	1.00 0.97 (0.86–1.10)	1.00 1.18 (1.03–1.34)
	Caregiving intensity:
	Non-caregiver	116.8	1.00	1.00	1.00
	Low (1–7 hrs/wk)	117.9	1.00 (0.82–1.22)	1.02 (0.84–1.25)	1.09 (0.90–1.33)
	Medium (8–22 hours/wk)	131.8	1.15 (0.94–1.41)	1.00 (0.82–1.23)	1.19 (0.97–1.47)
	High (24–168 hours/wk)	122.5	1.08 (0.89–1.31)	0.91 (0.75–1.12)	1.45 (1.14–1.83)
White Males	Caregiver: No Yes	77.0 94.0	1.00 1.25 (0.96–1.62)	1.00 1.16 (0.89–1.50)	1.00 1.53 (1.15–2.03)
	Caregiving intensity:
	Non-caregiver	77.0	1.00	1.00	1.00
	Low (1–7 hrs/wk)	81.5	1.09 (0.71–1.66)	1.14 (0.74–1.74)	1.22 (0.79 – 1.88)
	Medium (8–22 hours/wk)	135.8	1.75 (1.19–2.58)	1.68 (1.14– 2.47)	2.11 (1.41 – 3.14)
	High (24–168 hours/wk)	77.7	1.07 (0.68–1.69)	0.85 (0.53– 1.34)	1.46 (0.86 – 2.47)
White Females	Caregiver: No Yes	104.3 127.8	1.00 1.24 (0.99–1.56)	1.00 1.25 (0.99–1.57)	1.00 1.52 (1.18–1.96)
	Caregiving intensity:
	Non-caregiver	104.3	1.00	1.00	1.00
	Low (1–7 hrs/wk)	145.9	1.42 (1.02–1.98)	1.55 (1.12–2.16)	1.64 (1.18–2.29)
	Medium (8–22 hours/wk)	134.8	1.38 (0.93–2.05)	1.29 (0.86–1.94)	1.50 (0.99–2.28)
	High (24–168 hours/wk)	112.0	1.07 (0.72–1.60)	0.98 (0.65–1.47)	1.46 (0.91–2.34)
Black Males	Caregiver: No Yes	134.7 121.4	1.00 0.90 (0.67–1.20)	1.00 0.92 (0.69–1.24)	1.00 1.08 (0.79–1.47)
	Caregiving intensity:
	Non-caregiver	134.7	1.00	1.00	1.00
	Low (1–7 hrs/wk)	145.9	1.06 (0.67–1.67)	1.11 (0.70–1.75)	1.15 (0.72–1.81)
	Medium (8–22 hours/wk)	105.9	0.77 (0.45–1.32)	0.73 (0.42–1.25)	0.78 (0.45–1.39)
	High (24–168 hours/wk)	123.1	0.95 (0.59–1.51)	0.93 (0.57–1.50)	1.49 (0.84–2.65)
Black Females	Caregiver: No Yes	208.5 150.3	1.00 0.76 (0.62–0.95)	1.00 0.78 (0.63–0.96)	1.00 0.85 (0.67–1.09)
	Caregiving intensity:
	Non-caregiver	208.5	1.00	1.00	1.00
	Low (1–7 hrs/wk)	111.0	0.54 (0.36–0.83)	0.59 (0.39–0.91)	0.63 (0.41–0.97)
	Medium (8–22 hours/wk)	140.6	0.72 (0.50–1.03)	0.76 (0.52–1.09)	0.90 (0.61–1.32)
	High (24–168 hours/wk)	175.6	0.92 (0.68–1.26)	0.90 (0.66–1.24)	1.32 (0.89–1.97)

^tAll models adjusted for Health ABC site

^*Covariables that were included after backward elimination in Models B and C included: Total sample: age, race, gender, education, prevalent diseases, BMI, depressive symptoms White males: age, prevalent diseases, BMI, depressive symptoms, lives with others (latter in Model C but not Model B) White females: age, education, prevalent diseases, BMI, depressive symptoms Black males: education, prevalent diseases, BMI, depressive symptoms Black females: age, education, prevalent diseases, BMI, depressive symptoms (latter in Model C but not Model B)

DISCUSSION

This study found that among healthy community-dwelling elderly adults, caregivers had modestly elevated rates of mortality and incident mobility limitation. In addition, performing caregiving activities at least 24 hours/week was associated with a significantly increased rate of mobility limitation, but not mortality. The relationship between caregiving and health decline differed by race-gender group. White caregivers had higher rates of health decline than non-caregivers, with adjusted hazards ratios ranging from 1.52 for functional decline in white males to 1.61 for mortality in white females. These results are consistent with previous research by Schulz and Beach (1999) which found that older stressed caregivers to a spouse were 1.63 times more likely to die than married non-caregivers¹¹. We found no association between caregiving and health decline among black males, whereas black female caregivers had lower rates of health decline than non-caregivers, though these associations were attenuated when adjusted for total physical activity. Furthermore, in most race-gender groups, adjusting for total physical activity in addition to other confounders resulted in higher rates of health decline in high-intensity caregivers compared to non-caregivers, which was not observed in analyses that did not adjust for physical activity. These results complement research by Lee et al (2003), who found higher rates of CHD in women caregivers who performed caregiving tasks for a spouse at least 9 hours/week, but not in those who performed caregiving tasks less frequently¹².

To our knowledge, this is the first study on caregiving and physical health decline in elderly adults within different race-gender groups. Regardless of race-gender group, elderly caregivers who were most intensely involved in caregiving activities had the highest rates of incident mobility limitation compared to non-caregivers, when adjusting for physical activity. These results indicate that caregiving can have deleterious effects on physical health and quality of life even if it is not a strong risk factor for mortality: mobility limitations are costly,⁴⁰ increase the risk of mortality,⁴¹ and impair a person's ability to provide optimal care for the care recipient. These findings extend results of previous studies of caregiving effects on mortality^{10, 11} and CHD incidence¹².

The differences observed across race-gender groups support studies that found variations in caregiving involvement^22–27, psychological^{7, 22, 24} and physiological^30–33 effects of caregiving by these demographic categories. These results also suggest hypotheses to explain the relationship between caregiving and health outcomes. Specifically, the consistently higher rates of health decline among white male caregivers supports studies that found poorer immune response in male versus female caregivers^{30, 31}. Although white men may have less involvement in caregiving activities and report less stress than women, caregiving may have a stronger impact on their physical health because they may have fewer informal resources (such as social support networks⁴², interaction with other caregivers, or participation in caregiver support groups) to buffer its effects. The protective associations between minimal-moderate caregiving and health outcomes among black female caregivers are consistent with studies that found healthier persons are more likely to become caregivers²³ and that the physical effort of caregiving activities may benefit health²⁰. Moreover, these caregivers are more likely to care for relatives other than a spouse. If they are raising a grandchild or great-grandchild, they may get greater psychological satisfaction from the caregiving role or may be motivated to stay healthy, which could explain their lower risk of health decline.

This study was limited by its measure of caregiver status, and lack of measures to describe the caregiving situation. Because of the wording of the caregiver status variable, respondents may have interpreted it as providing regular child care, rather than taking care of a child who was sick or disabled. This possibility is supported by the fact that the highest percentage of caregivers were among black women, who are most likely to take care of grandchildren²⁸. In addition, the number of hours per week in caregiving activities served as a proxy measure of caregiving stress and may not reflect time spent in particularly stressful caregiving activities, nor the stressfulness of caregiving overall. For example, caregivers who are very stressed because they care for a person with advanced dementia or with severe limitations may perform fewer hours of caregiving per week because they have assistance with activities from other relatives or formal caregivers. This study also could not analyze the relationship between caregiving-related risk factors and health decline due to the absence of measures describing the caregiving situation that have been associated with stress in other studies. Such measures include the caregiver's relationship to the care recipient (e.g., spouse, grandparent), the care recipient's disability (e.g., dementia, general frailty, or developmental disabilities) and the number of instrumental and basic activities of daily living performed by the caregiver. Nonetheless, few previous studies of caregiving and health decline included these measures.

Another concern was that the results adjusted for total physical activity may have reflected better health status in general, rather than the effects of physical activity. Therefore, we re-ran the multivariable models (Model C), replacing the total physical activity variable with a measure of health status or physical performance (ie, self-rated health status, ease in walking one mile or up two flights of stairs, and walking speed over 400-meters or over 20-meters): these measures had weak-to no impact on the associations between caregiving and health decline, supporting the conclusion that caregivers' overall physical activity mediates their health outcomes. A final limitation was the lack of an independent measure of general stress that would have allowed us to distinguish stressed from non-stressed caregivers and non-caregivers, and to investigate how much the associations with health outcomes were due to caregiving, stress, or their combination.

Nonetheless, this study had several strengths. The Health ABC sample included healthy respondents who had no mobility limitation at baseline. Thus, unlike some other community-based studies²³, caregivers were not necessarily healthier than non-caregivers at baseline, and we were able to evaluate incident functional decline. The large number of black and male respondents allowed us to compare associations across race-gender subgroups. Moreover, in contrast to most studies of caregivers which are based on registries or convenience samples⁴³, the Health ABC sample constitutes a population-based sample of community-dwelling adults. Thus, the results are not likely to be biased by the types of caregivers who would volunteer for studies, and are generalizable to community-dwelling adults in their seventies. Finally, the study included measures of physical activity, health and demographic characteristics that allowed us to adjust for overall physical activity as well as important confounders.

In conclusion, this study tested hypotheses about caregiving and health decline on two outcome measures. Contrary to our hypothesis, caregiving was associated with higher rates of health decline on both measures in certain race-gender groups. The associations with incident mobility limitation corroborated the results observed with mortality. Adjustment for physical activity revealed adverse effects of caregiving that were not observed in analyses that excluded physical activity. Because this is the first study to evaluate the association between caregiving and health decline by race-gender group, as well as the role of physical activity in this association, additional studies are needed to replicate these findings. Nonetheless, these results indicate the importance of including physical activity when assessing caregivers' health status and when conducting studies of caregiving outcomes. Given the increasing number of elderly caregivers in the United States²⁵, these results underscore the potential toll of caregiving on caregivers' physical health. This study adds to the evidence from previous studies that support the need for more services and interventions to assist caregivers, reduce their stress, and maintain their physical health and ability to provide optimal care for their family members.