Appendix D

Number of Years and Percentage of Adult Life Spent Caring for an Older Adult

Commissioned Analysis by Vicki A. Freedman, Ph.D.

INTRODUCTION

Adults may be called on to provide care to an older adult one or more times during their lifetime. Young adults may participate in the care of their grandparents; adults in their 50s and 60s may need to care for an aging parent or parent-in-law; and older adults may provide care to spouses or siblings. The number of years that adults can be expected to spend on average in a caregiving role in the United States has not been previously quantified.

This memo provides estimates for the United States of the average number of years expected and percentage of remaining life to be spent providing care to an adult age 65 or older with an activity limitation. Findings are presented for informal (family or unpaid non-relative) adult caregivers to older adults with one or more activity limitations and for an alternative (narrower) definition of caregiving to older adults who meet criteria for severe limitations.

GENERAL APPROACH

The estimates presented here draw on a widely used life table methodology developed for generating active life expectancy estimates.¹ Instead of generating years and percentage of life spent without disability, we use the methodology to calculate years and percentage of life spent caregiving.

___________________

¹ Details of the method are available in Sullivan (1971) and the statistical underpinnings developed in Imai and Soneji (2007). Step-by-step calculations are available in Jagger et al. (2006).

The method involves three steps. First, the proportion of adults providing care is calculated for 10-year age groups. Numerators are drawn from the 2011 National Survey of Caregiving (NSOC) linked to the National Health and Aging Trends Study (NHATS) and denominators are from the 2011 Current Population Survey (CPS). Then, life tables provided by the National Center for Health Statistics are used to generate person-years lived and life expectancy for each age group. Finally, caregiving rates are combined with the life table estimates to apportion life expectancy into the average number of years and percentage of remaining life expected to be providing care. Additional methodological details are provided in the technical appendix.

CAREGIVING DEFINITIONS

We include care provided to adults ages 65 and older who live in community or residential care settings (other than nursing homes) and received assistance in the prior month with self-care or mobility activities (eating, bathing, dressing, or toileting; getting out of bed; getting around inside; getting outside) or household activities (doing laundry, shopping for groceries or personal items, making hot meals, handling bills and banking, and keeping track of medications), the latter for health or functioning reasons. For the alternative definition, we include only care to older adults who live in community or residential care settings (other than nursing homes) and either have probable dementia or received assistance in the past month with two or more self-care activities (eating, bathing, dressing, toileting, or getting in or out of bed).

For both definitions, caregivers are family members or unpaid non-relatives ages 20 and older who provided assistance in the past month with mobility, self-care, or household tasks; transportation; money matters other than bills or banking; or medical activities (sitting in with the sample person at physician visits; helping with insurance decisions).²

LIMITATIONS

The analysis has several limitations. First, estimates are sensitive to the definition of caregiving. Although we have demonstrated sensitivity to narrower definitions, using a broader definition that does not require the older adult to have a limitation or that includes a broader (or undefined) set of care tasks would yield higher estimates. Second, estimates of lifetime caregiving do not provide insights into the distribution of years spent caring

___________________

² We also generated a second set of alternative (narrow) estimates by imposing a minimum duration of receipt of help of 3 months or longer. See technical appendix for additional details.

and include those who never provide care. Thus, the estimates should be interpreted as population averages. Third, calculations apply current age-specific mortality and caregiving rates to a hypothetical cohort; hence, they are not intended to be forecasts of future experience. The stability of future caregiving rates will depend on a number of factors, including changes in late-life disability and mortality rates, average family size and composition, competing demands from work and family, the availability of formal caregivers, and cultural norms (Stone, 2015).

KEY FINDINGS

Proportion of Adults Providing Care to Older Adults

In 2011, approximately 18 million adults ages 20 and older—nearly 8 percent of all those age 20 and older—provided care to older adults with one or more activity limitations. The percentage of adults providing care ranges from less than 2 percent among those ages 20 to 29 to 16 percent among those ages 70 to 79 (Table D-1).

During mid-life (ages 40-69), women are more likely than men to provide care whereas men are more likely than women to provide care above age 80. Consequently, the chances of providing care peaks at different ages for men (nearly 16% older than age 70) and women (more than 18% among those ages 60 to 69).

About 8.5 million caregivers (48% of caregivers) provided care to an older adult with severe limitations. Percentages providing care are substantially lower using this narrower definition: the percentage ranges from less than 1 percent among those ages 20 to 29 to more than 7 percent among those ages 60 to 69 (last panel of Table D-1).

Number of Years and Percentage of Remaining Lifetime Providing Care to Older Adults

A 20-year-old adult can expect to spend on average 5.1 years—or nearly 9 percent of his or her remaining lifetime—caring for an older adult with an activity limitation (Table D-2). Over their lifetimes, women spend more years caring than men—on average 6.1 years or nearly 10 percent of their adult life—whereas men spend on average 4.1 years or just more than 7 percent of their adult life (p<.05 for difference in years).

The percentage of remaining life to be spent providing care peaks at different ages for men and women. For men, once they reach age 70, nearly 16 percent of remaining lifetime—or 1 to 2 years—is spent caring for an older adult. For women, this figure peaks between ages 50 and 69, when about 15 percent of remaining lifetime—or about 4 to 5 years—is spent caring.

TABLE D-1 Proportion Giving Care to Older Adults, by 10-Year Age Groups, 2011

NOTE: Caregivers are family members or unpaid non-relatives ages 20 and older who provided assistance in the past month with mobility, self-care, or household tasks; transportation; money matters other than bills or banking; or medical activities (sitting in with the sample person at physician visits; helping with insurance decisions).

^aAdults ages 65 and older who live in community or residential care settings (other than nursing homes) and received assistance in the prior month with self-care or mobility activities (eating, bathing, dressing, or toileting; getting out of bed; getting around inside; getting outside) or household activities (doing laundry, shopping for groceries or personal items, making hot meals, handling bills and banking, and keeping track of medications), the latter for health or functioning reasons.

^bAdults ages 65 and older who live in community or residential care settings (other than nursing homes) and either have probable dementia or received assistance in the past month with two or more self-care activities (eating, bathing, dressing, toileting, or getting out of bed).

SOURCES: Estimates calculated from the 2011 Current Population Survey and the 2011 National Study of Caregiving (NSOC) linked to the National Health and Aging Trends Study, unweighted n for NSOC.

TABLE D-2 Expected Number of Years and Percentage of Remaining Life Caring for an Older Adult, 2011

NOTE: Caregivers are family members or unpaid non-relatives ages 20 and older who provided assistance in the last month with mobility, self-care, or household tasks; transportation; money matters other than bills or banking; or medical activities (sitting in with the sample person at physician visits; helping with insurance decisions).

^aAdults ages 65 and older who live in community or residential care settings (other than nursing homes) and received assistance in the prior month with self-care or mobility activities (eating, bathing, dressing, or toileting; getting out of bed; getting around inside; getting outside) or household activities (doing laundry, shopping for groceries or personal items, making hot meals, handling bills and banking, and keeping track of medications), the latter for health or functioning reasons.

^bAdults ages 65 and older who live in community or residential care settings (other than nursing homes) and either have probable dementia or received assistance in the past month with two or more self-care activities (eating, bathing, dressing, toileting, or getting out of bed).

SOURCES: Estimates calculated from the 2011 Current Population Survey and the 2011 National Study of Caregiving (NSOC) linked to the National Health and Aging Trends Study, unweighted n for NSOC.

On average, 2.4 years—or nearly half of the years spent providing care to an older adult (2.4/5.1 years)—is spent providing care to an older adult with severe limitations, defined as receiving help with two or more activities of daily living or having probable dementia (second to last column of Table D-2).

TECHNICAL APPENDIX
Methodology for Calculating Average Number of Years and Percentage of Adult Life Spent Caregiving

DATA SOURCES AND CAREGIVER DEFINITIONS

Source of Caregiving Information. Age-specific estimates of the proportion caregiving are calculated from two sources.

Numerators are drawn from the National Study of Caregiving (NSOC), a follow-back to the 2011 National Health and Aging Trends Study (NHATS).³ NHATS is a nationally representative study of Medicare enrollees ages 65 or older living across all settings. The Round 1 response rate was 71 percent. NSOC is a follow-back telephone interview with all caregivers of eligible 2011 NHATS participants (see below for definition). NHATS respondents provided contact information for 68 percent of eligible caregivers. Sixty percent of those with contact information completed a telephone interview. NSOC provides non-response adjusted weights that are intended to adjust for the three levels of non-response so that the sample represents the total family caregiver population as identified in NHATS. For details see Kasper et al. (2013b).

Denominators (number of individuals in the non-institutionalized population by 10-year age groups) are drawn from the 2011 Current Population Survey (CPS) (U.S. Census Bureau, 2011).

Definition of Caregiving. NHATS participants were eligible for NSOC if they lived in the community or residential care settings other than nursing homes and received assistance in the past month with self-care or mobility activities (bathing, dressing, eating, toileting, getting out of bed, getting around inside, and going outside) or household activities (doing laundry, shopping for groceries or personal items, making hot meals, handling bills and banking, and keeping track of medications), the latter for health or functioning reasons.

Once eligible NHATS participants were identified, caregivers were eligible for NSOC if they were family members or unpaid non-relatives who provided assistance in the past month (according to the NHATS respondent) with mobility, self-care, household tasks, or transportation, or in the past year with money matters other than bills or banking or medical

___________________

³ NHATS and NSOC are sponsored by the National Institute on Aging (grant number NIA U01AG32947) and were conducted by the Johns Hopkins University.

activities (sitting in with the sample person at physician visits; helping with insurance decisions).

Of the 2,007 caregivers interviewed in NSOC, we excluded 11 respondents who did not provide care in the past month (according to NSOC) and 25 who were younger than age 20. Of the remaining 1,971 caregivers included in the analysis, 31 were missing age.⁴

Alternative (Narrower) Definition of Caregiving to Older Adults with Severe Limitations. We also generated estimates for a narrower definition of the caregiving population that includes only those who cared for an older adult with severe limitations. This group of care recipients is defined as living in the community or in residential care (other than nursing homes) and either (1) receiving help with two or more out of five activities (getting out of bed, eating, toileting, bathing, or dressing) or (2) being classified as having probable dementia.⁵ We also generated a second set of alternative (narrow) estimates that imposed a minimum duration of receipt of help of 3 months.⁶

CALCULATIONS

Choice of Age Interval. Ten-year age groups were chosen over smaller (e.g., 5-year) groups in order to ensure ample precision of estimates of the proportion providing care in each age group. For the broader definition of care for men and women together, there was also ample precision to repeat calculations using 5-year age intervals (presented at the end of this appendix).⁷

___________________

⁴ Age at the NSOC interview was calculated from month and year of birth from NHATS for spouse caregivers and from NSOC for other types of caregivers. For 36 cases where age was missing from NSOC, the information was filled in based on age in NHATS. An additional 31 cases were still missing age, and assumed to be missing age at random (i.e., we assumed knowing their ages would not change the age distribution).

⁵ NHATS participants were considered to have probable dementia if: the participant or the proxy reported a doctor’s diagnosis of dementia or Alzheimer’s disease; the participant received a score of 2 or more on a dementia screening instrument administered to a proxy; or the participant scored >=1.5 SD below the mean on at least two out of three domains on tests of memory, orientation, and executive functioning. These criteria have high sensitivity and specificity relative to a clinical diagnostic assessment (see Kasper et al., 2013b).

⁶ In the second set of calculations, duration of help was assumed to be 3 or more months if the NHATS respondent received assistance for 3 or more months with any self-care activities (if they reported receiving assistance with eating, toileting, bathing, or dressing) or with any mobility activities (if they only reported receiving help getting out of bed). This additional restriction is intended to approximate the 90-day requirement in the definition of disability in the Health Insurance Portability and Accountability Act (Drabek and Marton, 2015).

⁷ For all estimates, Relative Standard Errors (i.e., ratio of a standard error of an estimate to the estimate) are less than .30, a commonly used guideline in health surveys (Klein et al., 2002).

Proportion Caregiving and Standard Errors. To obtain estimates of the proportion caregiving, _nc_x, we divided the weighted number of caregivers from NSOC in each 10-year age group by the non-institutionalized population in each age group from the CPS for 2011 (see Table D-3).

Standard errors of proportions were calculated by taking the square root of the variance, according to the following formula: var(_nP_x*W*N/_nT_x) = (N^2) * [(W^2)*var(_nP_x) + (_nP_x^2)*var(W) + (var(_nP_x )*var(W))] / (_nT_x^2), where _nP_x is the proportion of caregivers in age group x to x+n, W is the average weight for the caregiving sample, N is the number of caregivers in the sample, and _nT_x is the number of adults in the population in age group x to x+n.⁸Table D-4 shows the unweighted and weighted sample sizes and the mean and standard error of the weight used in the calculations of the standard errors.

These calculations take into account uncertainty from two components in the numerators of the care rates: the distribution of caregivers across age groups (_nP_x) and the mean population weight (W). Standard errors for _nP_x and W were estimated using svy commands in Stata that take into account the complex design of NSOC. Population counts (from the CPS) are assumed to be fixed. The latter assumption should have minimal influence on the confidence intervals because the CPS relies on large sample sizes and produces point estimates very similar to the population counts from the 2010 Census.

Life Table Calculations. Unabridged (single year of age) life tables, available for 2010 for the entire population and by gender, were converted to abridged (10-year age category) life tables according to procedures described in Arias (2014). Because the focus of the caregiving calculations is adult life, we began the life table calculations at age 20; that is, the initial population (i.e., “radix”) of the life table was assumed to begin at age 20 with 100,000 people (see Table D-5).

Expected Years of Care and Percentage of Remaining Life Spent Caring. Life expectancy was apportioned into years spent caring using Sullivan’s method. First, we divided person-years expected to be lived in each age group (_nL_x in Table D-5) according to the proportion in each age group who provide care (_nc_x in Table D-6). Then, we calculated total years caring from age x forward by summing the person-years caring for the current age group to age 80+. We then calculated the expected number of years caring from age x by dividing the total years caring from age x forward by

___________________

⁸ For gender-specific estimates, we used the proportion of women (men) caregivers in age group i, the average weight for women (men), the number of women (men) caregivers, and the number of women (men) in the population in age group i.

TABLE D-3 Calculation of Age-Specific Proportions Caregiving and Standard Errors^a

^a See text for formula for calculating Standard Errors.

^b SOURCE: 2011 National Study of Caregiving linked to the National Health and Aging Trends Study.

^c SOURCE: U.S. Census Bureau 2011 Current Population Survey.

^d See text for definition of severe limitations.

TABLE D-4 Sample Sizes, Weighted Population, and Mean Weight for Caregiving Samples

^aSee text for definition of severe limitations.

SOURCE: 2011 National Study of Caregiving linked to the National Health and Aging Trends Study.

the number surviving to age x (column l_x in Table D-5). The percentage of remaining life to be spent caring was calculated by dividing the expected number of years caring from age x (in Table D-6) by the expectation of life at age x (in Table D-5). Step-by-step calculations (for active life expectancy) are available in Jagger et al. (2006).

Confidence Intervals for Expected Number of Years Caring.Table D-7 presents calculations of the standard error of the expected number of years caring. These calculations adopt the usual assumption that mortality rates (from vital statistics), which generate the life table estimates, are fixed. Step 1 (column 1) was to take the square of the number of person-years lived in each age group (_nL_x from Table D-5) and multiply that figure by the variance (squared standard error) of the proportion caregiving in that age group (SE(_nc_x) calculated in Table D-3). In column 2 we sum the figures in column 1 from age x forward. The variance of the expected number of years caring is then column 2 divided by the squared number of people surviving to age x (l_x from Table D-5), and the standard error is the square root of this calculation. Confidence intervals of 95 percent are calculated using the standard approach of plus or minus 1.96 times the standard error of the estimate. A test statistic for differences in number of years caring between men and women (3.87) was calculated by dividing the difference in years caring (6.1-4.1) by the sum of the square roots of the variances (.218+.299).

Alternative Estimates Using 5-Year Age Groups. To examine the sensitivity of calculations to age group width, Tables D-8 through D-11 provide calculations using 5-year age groups for (all) caregivers providing care to an older adult with activity limitations. Findings regarding percentage of life spent caregiving are consistent with calculations using 10-year and 5-year age groups. For example, at age 80 there is only a .2 percentage point difference between the estimates based on 10-year (11.3 percent) and 5-year (11.5 percent) age groups.

TABLE D-5 Abridged Life Table Calculations, Adults Ages 20 and Older, 2010

SOURCE: Based on National Center for Health Statistics (Arias, 2014).

TABLE D-6 Calculation of Expected Number of Years and Proportion of Remaining Life Spent Caregiving

^a See text for definition of severe limitations.

SOURCE: Proportion caring _nc_x and standard errors SE (_nc_x) calculated in Table D-3. _nL_x calculated in Table D-7. Standard errors of e_cx calculated in Table D-7.

TABLE D-7 Calculation of Standard Error of Caregiving Life Expectancy

^a See text for definition of severe limitations.

SOURCE: _nL_x and l_x calculated in Table D-5; SE(_nc_x) calculated in Table D-3.

TABLE D-8 Calculation of Age-Specific Caregiving Rates and Standard Errors: 5-Year Age Groups^a

^a See text for formula for calculating Standard Errors.

^b SOURCE: 2011 National Study of Caregiving linked to the National Health and Aging Trends Study.

^c SOURCE: 2011 Current Population Survey.

TABLE D-9 Abridged Life Table Calculations, Adults Ages 20 and Older, 2010: 5-Year Age Groups

SOURCE: Based on National Center for Health Statistics (Arias, 2014).

TABLE D-10 Calculation of Expected Number of Years and Percentage of Remaining Life Spent Caregiving: 5-Year Age Groups

SOURCE: Caregiving rates _nc_x and standard errors SE (_nc_x) calculated in Table D-8. _nL_x calculated in Table D-9. Standard errors of e_cx calculated in Table D-11.

TABLE D-11 Calculation of Standard Error of Caregiving Life Expectancy: 5-Year Age Groups

SOURCES: _nL_x and l_x calculated in Table D-9; SE(_nc_x) calculated in Table D-8.

REFERENCES

Klein, R. J., S. E. Proctor, M. A. Boudreault, and K. M. Turczyn. 2002. Healthy People 2010 criteria for data suppression, no. 24. http://www.cdc.gov/nchs/data/statnt/statnt24.pdf (accessed October 10, 2015).

National Health and Aging Trends Study. Produced and distributed by www.nhats.org with funding from the National Institute on Aging (grant number NIA U01AG32947).

National Study of Caregiving. Produced and distributed by www.nhats.org with funding from the U.S. Department of Health and Human Services’ Office of the Assistant Secretary for Planning and Evaluation in cooperation with the National Institute on Aging (grant number NIA U01AG32947).

Stone, R. 2015. Factors affecting the future of family caregiving in the United States, edited by J. Gaugler and R. L. Kane. In Family caregiving in the new normal. Waltham, MA: Elsevier.

Sullivan, D. F. 1971. A single index of mortality and morbidity. HSMHA Health Reports 86:347-354.

U.S. Census Bureau. 2011. Current Population Survey, Annual Social and Economic Supplement. Table 1.http://www.census.gov/population/age/data/2011comp.html (accessed October 10, 2015).