Abstract
Objectives.
To examine whether stability of living arrangements and changes in household composition—both types and frequency—influence nursing home placement or timing to entry among older people.
Methods.
Data from the first 5 waves of the Assets and Health Dynamics of the Elderly (1993–2002) on 8,093 persons aged 70 years or older at baseline are used in probit and hazard models to predict nursing home entry and time to entry.
Results.
Stable living arrangements carry different risks of institutionalization. Those living continuously alone or with others were at highest risk; at lowest risk were those living continuously with a spouse or the same child (lowest overall). Changes in household composition were protective against nursing home entry and slowed time to entry; types of change were not influential when number of changes was taken into account.
Discussion.
Results suggest that stability of living arrangements in and of itself is not protective against institutionalization. Having options that allow one to change living arrangements over time in response to changing needs for assistance is of importance if the goal is to avoid institutional care or extend community residence prior to entry.
Keywords: Generational studies, Health services use, Living arrangements
STUDIES of nursing home admission among older people have consistently identified some living arrangements as being protective against nursing home entry and others as putting people at greater risk. Coresident family members, spouses in particular, are potential sources of support in the event of disabling physical or cognitive health changes that otherwise might lead to institutional residential care. Although many studies of nursing home entry have included living arrangements as a determinant, most use a single point in time measure and few have examined the role of living arrangement transitions.
Conceptually, nursing home entry is a form of health service use and can be viewed through the lens of a general model of health service use (Andersen, 1995). As an outcome, nursing home entry is a function of individual characteristics and resources (economic and social), need (for assistance due to disability), and supply of services (nursing home beds). Living arrangements and marital status have been included as covariates in many studies of nursing home entry (Freedman, 1996). A recent meta-analysis (Gaugler, Duval, Anderson, & Kane, 2007) found that being married lowered the odds of entry, whereas living alone resulted in twice the odds of being admitted. An earlier systematic review by Miller and Weissert (2000) found that two thirds of studies that examined living alone found that it increased likelihood of nursing home entry, and about two fifths of studies that included marital status found being married reduced risk. Other types of community living arrangements, for example, co-residing with an adult child or other relatives, have received less attention, although the availability of children as actual or potential caregivers is frequently examined (Spillman & Pezzin, 2000; Wolff & Kasper, 2006).
Among other individual characteristics, race and ethnicity have been of particular interest because of historical patterns of underrepresentation of minorities in the nursing home population (Akamigbo & Wolinsky, 2007). Konetzka and Werner (2009) in a comprehensive review of disparities in long-term care note that several studies have tried to sort out the extent to which these patterns reflect differences in support networks (Cagney & Agree, 1999; Chatters, Hardison, & Riley, 2001), culture (Dunlop, Manheim, Song, & Chang, 2002), and supply of services (Mor, Zinn, Angelelli, Teno, & Miller, 2004; Reed & Andes, 2001).
The relationships of economic resources, need, and supply of services to nursing home entry have also received attention. Both public (primarily Medicaid that covers nursing home care for low-income people) and private economic resources influence nursing home use, with increasing probabilities of use at the low and high ends of the income spectrum (Jenkins, 2001). Severe disability and cognitive impairment (measures of need in Andersen’s framework) have consistently been associated with nursing home entry as documented by reviews of the literature (Miller & Weissert, 2000; Gaugler et al., 2007) and cross-national studies (Himes, Wagner, Wolf, Aykan & Dougherty, 2000; Nygaard & Alberktsen, 1992). Finally, supply of nursing home beds is associated with use, one reason that most states have policies regulating the supply of nursing homes (Harrington, Anzaldo, Burdin, Kitchener, & Miller, 2004). Recent research has focused on variations in availability of high-quality nursing homes (Mor et al., 2004) as opposed to bed supply.
The study of determinants of nursing home entry is driven by the desire to find targets for intervention to prevent unnecessary institutionalization. Extending time in the community even when placement is unavoidable is also important, however, for reasons that include documented preferences for maintaining community residence and reducing the high costs associated with long-term care. Studies of factors associated with extending time in the community among persons at high risk of nursing home entry are relatively few. A high proportion of persons with diagnosed dementia, for example, are ultimately admitted to a nursing home (Hebert, Dubois, Wolfson, Chambers, & Cohen, 2001; Smith, Kokmen, & O’Brien, 2000; Yaffe et al., 2002), but considerable variability in time to admission has been observed, and living arrangements and marital status are among factors that affect timing (Heyman, Peterson, Fillenbaum, & Pieper, 1997; Miller et al., 1998; Yaffe et al., 2002). A recent article also suggests that timing of nursing home placement affects the survival time of persons with Alzheimer’s disease (McClendon, Smyth, & Neundorfer, 2006).
The objectives of this study are to examine whether living arrangements, and frequency and types of changes in household composition, influence nursing home placement or timing to entry among older people. We include as covariates other individual characteristics and measures of economic resources and need, as suggested by Andersen’s model and previous research. Three questions are addressed. First, is stability (or instability) of living arrangements associated with risk of nursing home entry? Because living arrangements represent one crucial component of community support to older individuals needing functional assistance, one might hypothesize that stability of living arrangements is protective. The concept of “aging in place” reflects the desire to stay in a chosen environment regardless of health changes, although risks associated with aging in place are also being recognized (Lau, Scandrett, Jarzebowski, Holman, & Emanuel, 2007). Living arrangement changes, and multiple changes in particular, may indicate lack of stability in community living arrangements and carry greater risks of nursing home placement. On the other hand, living arrangements and caregiving are often intertwined, and changes in living arrangements may be made to enhance access to needed assistance (Hays, 2002). Second, we investigate whether some types of living arrangement changes are more beneficial than others (e.g., moving in with an adult child relative to moving in with others) in reducing probability of nursing home entry and extending time to admission. Finally, we examine the extent to which types of living arrangements and frequency of living arrangement transitions extend months of community residence among persons who were admitted to a nursing home. Whether some types of living arrangements, or changes over time, are more effective in delaying nursing home entry even when admission cannot be prevented has important implications for policies designed to encourage certain types of community long-term care arrangements (e.g., coresidence with children through tax policies).
METHODS
Sample
Data from the first five waves (1993–2002) of the Assets and Health Dynamics of the Elderly (AHEAD), part of the U.S. Health and Retirement Survey (HRS), are used in this analysis. Designed as a nationally representative biennial study of Americans aged 70 years or older, the AHEAD-HRS study is particularly well suited for this analysis as it was specifically designed to enable a comprehensive understanding of how family and formal support interact to affect the well-being of the elderly respondents as they age (Soldo, Hurd, Rodgers, & Wallace, 1997). Although the AHEAD data starts with a sample of noninstitutionalized individuals, the panel tracks people when they enter a nursing home or similar institutions. Each wave provides information on the current economic status, health status, and living arrangement of respondents. Wave 1 consists of 8,219 respondents. Our study sample of 8,093 persons excludes 126 persons who were missing any of the key variables used in the multivariate analyses.
Measures
Community living arrangements: frequency and types of changes.—
Detailed wave-specific living arrangement indicators were constructed for each of the five waves of data collection. These indicators are mutually exclusive at each wave and include: respondent lives alone, with spouse only, with adult (biological or step) children only, with spouse and adult children, with other relatives only (neither spouse nor adult children present), with spouse and other relatives (no adult children present), with adult children and other relatives (no spouse present), with other nonrelatives only, with spouse and other nonrelatives (no adult children present), with children and other nonrelatives only (no spouse present), or in a nursing home. Among those who live with an adult child at a given wave, we further ascertained whether they lived with the same or a different child in the subsequent wave. These wave-specific indicators covering 10 years were used to create two variables: a count of living arrangement changes and a classification of types of changes.
The variable for frequency of changes reflects a count based on wave-to-wave observations of changes in living arrangement indicators over time. This “count” reflects changes from one wave to the next in the indicators listed previously or the “loss” of an adult child who had been present in the previous wave among those living with (at least one) adult child(ren). Although most respondents could experience only one change per wave pair, respondents who lost a spouse and made an additional living arrangement change from one wave to the next (e.g., now lived with an adult child or with others) were coded as having made two changes (maximum number of possible changes between waves was two). Throughout the study period, the minimum number of changes was zero; at the extremes, 24 people experienced four changes, 4 experienced five changes, and 1 experienced seven changes over the 10-year observation span. Nursing home entry—an end point for the study—works as a censoring mechanism; therefore, moving from any community living arrangement in one wave to a nursing home in the next was not counted as a change in living arrangement.
The variable for types of changes in living arrangement or patterns over time was also developed from the wave-specific indicators. Four patterns of “stable” living arrangements emerged: older adults who were always alone, those always with a spouse (including a small number of respondents who lived with a spouse and at least one adult child), always with the same adult child (but not with a spouse), and those always with others (predominantly relatives, including grandchildren, but excluding a spouse or adult child; <5% were nonrelatives). People in these four patterns are referred to throughout the article as living “continuously” in these arrangements, although changes in household composition over the study period are possible in some of these continuous living arrangements (e.g., persons who always lived with a spouse may experience changes in household composition that reflect an adult child moving in or out).
Three predominant patterns of change also were identified: persons who lost a spouse and remained alone; persons who moved in with an adult child having lived independently previously (either alone or with a spouse only who is now deceased); and persons who transitioned to living with others, having previously lived alone or with a spouse or adult child. Persons who exhibited other less common patterns, for example, living alone and then remarrying or living with a relative other than a spouse or adult child then changing to a nonrelative or to living alone, are grouped into “other patterns” because these did not occur with enough frequency to be examined individually.
Nursing home entry.—
The key dependent variables examined in this study are the likelihood and timing of permanent nursing home entry. It is therefore important to distinguish short postacute discharge nursing home stays from nursing home entry as a permanent residential transition due to the need for 24-hr care on a continuing basis. Information collected retrospectively at each interview was used for this purpose: Respondents were coded as having made a permanent transition if their nursing home stay (a) lasted across two waves or (b) ended in death with no return to community between waves.
Other covariates.—
The multivariate analyses include additional self (or proxy)-reported variables that are expected or known to be related to nursing home entry. These include age (in years), gender, race/ethnicity (Black non-Hispanic, Hispanic vs. other races, and primarily White non-Hispanics), number of living children, presence of any in basic or instrumental activities of daily living (ADL and IADL) limitations, number of chronic conditions (from zero to eight based on diabetes, hypertension, cancer [other than minor skin cancer], chronic lung disease, heart disease [heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems], stroke, psychiatric problems, and arthritis), cognitive functioning (values indicating fair/poor in a Likert scale rating the subject’s memory based on respondent’s and proxies’ assessments), and net worth (expressed as a continuous variable). In addition to these variables measured at baseline, we also include an indicator of whether the elderly respondent experienced a decline in physical and/or cognitive functioning over the study period, based on the trajectory of ADLs, IADLs, and general cognition over the 10-year period.
The number of years in the survey was also used in the analysis of nursing home entry to control for differential exposure (time to make living arrangement changes) due to death—the primary source of attrition in this sample—or survey nonresponse.
Analytic Strategy
We first estimate the effect of living arrangements on nursing home entry (n = {0,1}) for individual j,
 |
as a function of socioeconomic and health/disability variables (Xj), and household living arrangement types and number of household composition changes (Aj), described above, as a probit regression model. We then use the estimated coefficients to calculate the marginal effect on the probability of nursing home entry associated with each covariate.
To incorporate the dynamic aspect of our data, we estimate an individual’s time to nursing home entry—measured in months—via a competing risks, Gompertz model (Lee & Wang, 2003), such that an individual’s instantaneous risk, or hazard, of nursing home entry (h) between time t and t + 1 is
 |
in which β1, β2, and γ are parameters estimated within the model. The survival analysis corrects for bias in estimates related to censoring through death or the end of the survey.
We then use the estimated parameters to predict the time to nursing home entry for persons who entered nursing homes under several simulated living arrangement scenarios. Comparing the simulated time to nursing home entry allows us to quantify, in terms of changes in months to nursing home entry, the effect of altering an individual’s community living arrangements and their stability prior to entry.
We tested the sensitivity of our findings by examining several variants of the model. In some, we control for additional regressors, such as current income; in others, we experimented with alternative approaches to capturing time-varying physical and cognitive functioning (e.g., by including separate terms for declines in ADLs, IADLs, and cognition) or interaction terms between certain characteristics of interest (e.g., baseline disability and living arrangements). Our findings were remarkably robust to these alternative specifications, none of which yielded improved model fit. For all variants, the Gompertz model achieved the best fit among several alternative parametric survival models as measured by the Akaike’s information criterion (Akaide, 1974).
RESULTS
Stability and Change in Living Arrangements
In this cohort of individuals, 71.6% had stable living arrangements over 10 years (or until death or institutionalization)—consisting of continuously living alone, with a spouse, or with the same adult child (Table 1). A relatively small percentage of those living with a spouse or adult child did experience changes in household composition in the form of an adult child or children, who moved in or out, or other relatives or nonrelatives, who joined or left the household. Nonetheless, 89.8% of those living with a spouse had no changes to his/her household; the same was true for 86.6% of those who lived with the same child over the study period.
Table 1.
Living Arrangements and Number of Changes in Household Composition Over 10 years: Assets and Health Dynamics of the Elderly/Health and Retirement Survey 1993–2002
| Number of changes in household compositiona |
|||||
| Living arrangements | % (n) | 0 | 1 | 2 | 3+ |
| Continuously | |||||
| Alone | 28.6 (2,315) | 100.0 | — | — | — |
| With spouse | 37.9 (3,070) | 89.8 | 7.2 | 2.5 | 0.5 |
| With same adult child | 7.7 (619) | 86.6 | 13.4 | — | — |
| With othersb | 2.4 (190) | 100.0 | — | — | — |
| Changed | |||||
| From spouse or alone to adult childc | 4.6 (368) | — | 55.7 | 32.1 | 12.2 |
| From spouse to alone | 11.6 (940) | — | 82.6 | 13.6 | 3.8 |
| From spouse/alone/with adult child to othersb | 3.0 (242) | — | 43.4 | 40.1 | 16.5 |
| Other patterns | 4.3 (349) | — | 67.9 | 24.6 | 7.5 |
| Total % | 100.0 | 71.6 | 20.1 | 6.3 | 2.0 |
| Total N | 8,093 | 5,797 | 1,626 | 507 | 163 |
Changes in household composition were counted from one wave to another. For example, adding or losing an adult child from a household and adding or losing a spouse (usually through death) from a household. These changes are irrespective of physical relocation. For example, adding a child includes both a child who moves into a parent’s home and a parent who moves into a child’s home. Changes in household composition of persons living “continuously” with a spouse or with the same adult child reflect an adult child or children, who moved in or out, or other relatives or nonrelatives, who joined or left the household during the study period.
Predominantly relatives (excluding a spouse or adult child).
Also includes persons with a pattern from spouse to alone to adult child.
Among persons who changed living arrangements, the most common type of change was from spouse to living alone (11.6% of the sample). Persons who took up residence with an adult child, having previously been with a spouse or alone, were 4.6%, and those who took up residence with others were 3.0%. Other patterns of living arrangement change represented 4.3% of the sample.
Most people with changes in household composition over this 10-year period experienced only one change (20.1% of the total sample). Another 8.3% of the sample experienced two or more changes.
Nursing Home Entry Over 10 Years
About 1 in 10 individuals became nursing home resident over this 10-year period (Table 2). As expected, a higher percentage of those who continuously lived alone (19.6%) and a lower percentage (7.4%) of those who lived continuously with a spouse entered a nursing home. The highest percentage of persons transitioning to nursing home residence (21.1%) was among those who lived continuously with others (primarily relatives other than a spouse or adult child). These unadjusted estimates also indicate that more living arrangement transitions were associated with lower rates of nursing home entry, with persons with no changes in household composition having the highest percentage of persons entering a nursing home (13.2%).
Table 2.
Nursing Home Entry by Living Arrangements, Number of Changes in Household Composition, and Other Characteristics: Assets and Health Dynamics of the Elderly/Health and Retirement Survey 1993–2002
| Characteristics | Sample, % (n) | Entered nursing home (%)a |
| Total | 100 (8,063) | 11.6 |
| Living arrangements | ||
| Continuously | ||
| Aloneb | 28.6 (2,315) | 19.6 |
| With spouse | 37.9 (3,070) | 7.4** |
| With same adult child | 7.7 (619) | 11.2** |
| With othersc | 2.4 (190) | 21.1 |
| Changed | ||
| From spouse or alone to adult child | 4.6 (368) | 12.2** |
| From spouse to alone | 11.6 (940) | 5.6** |
| From spouse/alone/with adult child to othersc | 3.0 (242) | 9.5** |
| Other patterns | 4.3 (399) | 8.6** |
| Number of changes in household composition | ||
| 0b | 71.6 (5,797) | 13.2 |
| 1 | 20.1 (1,626) | 8.7** |
| 2 | 6.3 (507) | 5.1** |
| 3+ | 2.0 (163) | 3.1** |
| Individual characteristicsd | ||
| Age (in years) | ||
| 70–75b | 40.0 (3,233) | 5.8 |
| 76–80 | 23.8 (1,925) | 13.2* |
| 81+ | 26.8 (2,171) | 21.9* |
| Gender | ||
| Male | 36.8 (2,974) | 8.1* |
| Female | 63.2 (5,119) | 13.7 |
| Race/ethnicity | ||
| Black non-Hispanic | 13.3 (1,078) | 12.3 |
| Hispanic | 6.0 (482) | 11.9 |
| Other race/ethnicity (primarily Caucasian) | 80.7 (6,533) | 6.0** |
| Net worthe [M; SE] | [1.8; 4.1] (8,063) | [1.1; 1.6]* |
| Number of living children | ||
| 0b | 12.3 (995) | 18.7 |
| 1 | 16.2 (1,307) | 13.9* |
| 2+ | 71.6 (5,791) | 9.9* |
| Any ADL limitations | ||
| Yes | 14.1 (1,143) | 18.4* |
| No | 85.9 (6,950) | 10.5 |
| Number of chronic diseases | ||
| 0b | 22.1 (1,790) | 11.0 |
| 1 | 33.1 (2,675) | 10.6 |
| 2+ | 44.8 (3,628) | 12.7 |
| Cognitive functioningf | ||
| Fair/poor | 10.6 (860) | 22.7** |
| Excellent/very good/good | 89.4 (7,233) | 10.3 |
n = 939.
Reference category.
Predominantly relatives (including grandchildren) but excluding a spouse or child.
Baseline measures.
Continuous variable constructed by dividing net worth in dollars by $100,000.
Based on self-rating or proxy rating of memory at baseline.
*0.01< p ≤ .05; **p ≤ .01.
As expected, the percentage of persons who entered a nursing home was highest among those at older ages (21.9% among persons older than 80 at baseline) and among women (13.7%). Other characteristics associated with greater likelihood of entering a nursing home were lower net worth, no living children, ADL limitations, presence of chronic diseases, and an assessment of current memory as fair or poor.
Multivariate Results for Nursing Home Entry and Time to Placement
Each change in household composition significantly reduced likelihood of nursing home entry controlling for initial living arrangement and other characteristics—reducing one’s probability by 28% (from unadjusted probability of 11.6%–8.3%; Table 3). Relative to living continuously alone (adjusted probability = 14.4%), living continuously with a spouse significantly reduced likelihood of placement (−28% to 10.2%) as did living with the same adult child continuously (−48% to 7.5%). When controlling for number of living arrangement changes, loss of a spouse to alone or taking up residence with others (having previously lived alone or with a spouse or adult child) were associated with a change in likelihood of nursing home entry, both reducing probability of entry relative to those living continuously alone (by 36% and 24%, respectively). Other characteristics associated with an increased probability of nursing home entry were age, having ADL limitations, a fair/poor memory assessment at baseline, and experiencing a decline in physical or cognitive functioning over the 10-year study period. The most significant of these were declining physical or cognitive functioning, which increased the probability of nursing home entry by 463% (from 3.5% for those with no such decline to 19.7% for those with a decline); presence of ADL limitations at baseline, which increased probability of nursing home entry by 113% (from 10.2% for those without ADL limitations to 21.7%); and baseline cognitive functioning (fair/poor memory assessment), which increased probability of placement by 67% (from 10.7% to 17.9%). Being male, having higher net worth and a greater number of living children, on the other hand, were associated with significantly decreased likelihood of institutionalization.
Table 3.
Effects of Living Arrangements and Household Composition Changes on Nursing Home Entry and Time to Entry: Assets and Health Dynamics of the Elderly/Health and Retirement Survey 1993–2002
| Nursing home entry |
Time to entry |
|||
| Coefficient (SE) | Predicted probability (%) | Marginal effects (as % change)a | Hazard ratio (CI) | |
| Number of changes in household composition | −0.24** (0.06) | 8.3 | −28 | 0.60** (0.47–0.77) |
| Living arrangements | ||||
| Continuously | ||||
| Aloneb | — | 14.4 | — | — |
| With spouse | −0.27** (0.05) | 10.2 | −28 | 0.77** (0.64–93) |
| With same child | −0.48** (0.08) | 7.5 | −48 | 0.54** (0.41–0.70) |
| With othersc | −0.07 (0.12) | 13.1 | −13 | 0.91 (0.65–1.28) |
| Changed | ||||
| From spouse or alone to child | −0.16 (0.14) | 13.4 | −6 | 0.81 (0.52–1.24) |
| From spouse to alone | −0.31** (0.11) | 9.7 | −36 | 0.61** (0.41–0.80) |
| From spouse/alone/with child to othersc | −0.33** (0.16) | 9.4 | −24 | 0.69 (0.40–1.18) |
| Other patterns | −0.17 (0.14) | 11.7 | −26 | 0.85 (0.52–1.38) |
| Years observed in survey | 0.01** (0.00) | 14.2 | 22 | — |
| Age (in years)d | 0.05** (0.00) | 12.6 | 0.8 | 1.09** (1.08–1.10) |
| Maled | −0.15** (0.05) | 10.0 | −23 | 0.81** (0.68–0.99) |
| Black, non-Hispanicd | −0.11* (0.08) | 10.5 | −13 | 0.74** (0.60–0.90) |
| Hispanicd | −0.42** (0.01) | 6.8 | −44 | 0.44** (0.30–0.64) |
| Net worthd,e | −0.50** (0.01) | 10.9 | −7 | 0.91** (0.87–0.95) |
| Number of living childrend | −0.04** (0.01) | 11.0 | −6 | 0.93** (0.90–0.96) |
| Any ADL limitations (yes)d | 0.65** (0.06) | 21.7 | 117 | 2.72** (2.28–3.20) |
| Number of chronic diseasesd | 0.01 (0.02) | 12.2 | 5 | 0.99 (0.94–1.10) |
| Cognitive functioningd | 0.42** (0.06) | 17.9 | 67 | 2.40** (2.01–2.86) |
| Experienced decline in physical or cognitive functioning | 1.13** (0.55) | 19.7 | 463 | 4.99** (4.15–6.01) |
| Constant | −6.62 (—) | — | — | — |
Notes: Results for nursing home entry are from a probit model (dependent variable = permanent nursing home entry yes/no). Results for time to entry are from a competing risks Gompertz hazard model (dependent variable = months to permanent nursing home entry). CI = confidence interval.
Marginal effects represent percentage changes in adjusted (predicted) probability of nursing home entry relative to the baseline predicted probability of 11.6%. For continuous variables, adjusted probabilities are calculated by adding one unit to the relevant characteristics (e.g., number of changes in household composition) while holding other factors constant at their original levels. For binary variables, marginal effects represent the difference (measured as percentage change) between adjusted probabilities calculated by first assuming the presence and then the absence of a specific characteristic (e.g., fair/poor memory assessment) while holding other factors constant at their original levels.
Marginal effects for living arrangement variables represent percentage change between adjusted probabilities of nursing home entry for each category relative to the adjusted probability of 14.3% for the reference category “alone.”
Predominantly relatives (excluding a spouse or adult child).
Baseline measures.
Continuous variable constructed by dividing net worth in dollars by $100,000.
*0.01< p ≤ .05; **p ≤ .01.
Results for time to entry largely mirror those for placement (Table 3). Time to nursing home entry is slowed by greater numbers of changes in living arrangements, with almost a one-third reduction in the likelihood of entering a nursing home earlier with each increase. Similarly, relative to living continuously alone, living continuously with a spouse or with the same child was protective against earlier nursing home placement (a one-quarter and a one-half reduction, respectively). With the exception of loss of a spouse, none of the other types of living arrangement changes affected likelihood of experiencing earlier nursing home entry relative to living continuously alone. Men and those with higher net worth and more living children also were protected against earlier nursing home entry. Persons with ADL limitations, those with fair/poor memory rating at baseline, and those who experienced declines in their physical or cognitive functioning entered a nursing home more rapidly than others.
Living Arrangement Changes and Months of Community Residence Among Persons Who Transitioned to Nursing Home Residence
The simulations shown in Table 4 adjust for all variables in the multivariate models and show for persons who entered a nursing home over the 10-year study period, the predicted number of months of community residence gained or lost prior to nursing home entry by different living arrangement patterns and household composition changes. Persons living continuously with a spouse had 12.9 additional months in the community relative to those living alone continuously, and persons living continuously with the same child had 31.9 additional months in the community prior to nursing home entry. Although living with others was more protective against nursing home entry relative to living alone, it was the least protective among other stable living arrangements, associated with quicker rates of nursing home entry than those living with a spouse or with an adult child by 8 and 27 months, respectively.
Table 4.
Predicted Months of Community Residence Prior to Nursing Home Entry Among Persons Who Entered a Nursing Home by Living Arrangement and Number of Changes in Household Composition: Assets and Health Dynamics of the Elderly/Health and Retirement Survey 1993–2002
| Months of community residence gained (lost) relative to these living arrangements |
|||
| Types of living arrangements and number of changes in household composition | Alone continuously | With spouse continuously | With child continuously |
| None | |||
| Alone continuously | — | — | — |
| With spouse continuously | 12.9 | — | — |
| With same adult child continuously | 31.9 | 19.0 | — |
| With othersa, continuously | 4.6 | −8.5 | −27.5 |
| 1 Change | |||
| With spouse continuously | 57.3 | 44.4 | 25.4 |
| With same adult child continuously | 77.8 | 65.0 | 45.9 |
| From spouse or alone to adult child | 51.9 | 39.0 | 20.0 |
| From spouse to alone | 70.7 | 57.8 | 38.8 |
| From spouse/alone/child to othersa | 62.1 | 49.2 | 30.2 |
| Other patterns | 54.2 | 41.8 | 22.8 |
| 2+ Changes | |||
| With spouse continuously | 100.9 | 87.9 | 69.0 |
| With spouse or alone to adult child | 105.9 | 93.0 | 74.0 |
| From spouse to alone | 114.9 | 102.0 | 83.0 |
| From spouse/alone/child to othersa | 95.3 | 82.3 | 63.3 |
| Other patterns | 98.1 | 85.2 | 66.2 |
Notes: Simulations are based on parameter estimates from the Gompertz duration model shown in Table 3. Predicted months of community residence for each living arrangement/number of changes in household composition combination are calculated for each individual by substituting the values of the relevant characteristics (e.g., living arrangement = continuously with spouse and changes in household composition = 1) while holding other factors constant at their original levels. Individual level probabilities were calculated for all persons in the Gompertz duration model (n = 8,093), with means calculated for the relevant subpopulation of persons who entered a nursing home over the 10-year study period (n = 939).
Predominantly relatives, including grandchildren (excluding a spouse or adult child).
Additional changes in living arrangements increased months of community residence regardless of living arrangement change patterns. Those living continuously with a spouse and experiencing one total change gained an additional 44.4 months over those with no changes. Those with two or more changes gained an additional 87.9 months. The magnitude of the gains associated with number of changes was not constant across all living arrangements, however. Relative to those living alone continuously, the gain from zero to one household change for those living with a spouse continuously was a 344% increase in months (12.9–57.3) compared with a 144% increase for those living continuously with the same child (31.9–77.8).
The advantages conveyed by certain living arrangements also persisted. Although those living continuously with a spouse who also had one household change gained months of community residence prior to nursing home entry over those who lived continuously alone, with spouse or with the same child, the gains were greater relative to those continuously alone (57.3 additional months) than to those continuously with the same child (25.4 additional months).
DISCUSSION
This study examined the relationship between stability and changes in community living arrangements and nursing home entry. The results indicate a more complex picture than the one that has emerged from studies that have relied on baseline measures of living arrangement. Living alone or with a spouse predominate as living arrangements among people aged 70 years or old. These and other long-time stable living arrangements have different implications for risk of nursing home entry, however, indicating that stability of living arrangements in and of itself is not protective against institutionalization. Findings that persons living continuously alone were at greater risk, and those living continuously with a spouse were at lower risk, of nursing home placement are consistent with many previous studies. At the lowest risk of nursing home placement, however, were those living continuously with the same adult child, perhaps an indication that children who enter into coresidence with an aging parent are more capable of offering adequate full-time care than others. Those living with relatives other than a spouse or children were at equal risk of placement to those living alone. These arrangements may represent a “last resort” in terms of community living arrangements.
Once number of changes was accounted for, the most common patterns of change did not (for the most part) influence risk of placement relative to living alone. The exception was for those who lost a spouse and then lived alone, who were less likely to enter a nursing home relative to those living continuously alone. One recent study of older Finnish adults over a 5-year period found that risk of institutionalization was high immediately after the death of a spouse (in the first month) but decreased over time (Nihtila & Martikainen, 2008) using a comparison group of married individuals. Hays, Pieper, and Purser (2003) in a longitudinal analysis of the Duke Established Populations for Epidemiological Studies of the Elderly found that death of a spouse was associated not only with an increased likelihood of residential change (either household expansion or nursing home entry) but also with an increased likelihood of household expansion versus institutionalization. Our analysis indicates that most people do not make a change upon losing a spouse, however. As Table 1 indicates, 11.6% of persons aged 70 years or older over this 10-year period lost a spouse and subsequently lived alone for the remainder of the observation period.
Greater numbers of changes in living arrangements were the most important protective measure against nursing home entry. Types of change (e.g., from spouse to alone or from spouse or alone to adult child) appear not to matter as much as simply being able to make a change (with one exception—loss of spouse). If number of living arrangement changes is taken out of the models, all the “changed” living arrangements then reach significance (data not shown) in the direction of reducing risk of nursing home entry and slowing time to entry. These results suggest that having options that allow one to change living arrangements over time in response to changing needs for assistance is of greater importance than the type of change that is made, if the goal is to avoid institutional care. Among those who are no longer making changes or no longer have options, however, certain living arrangements are higher risk for both entry and quicker entry to nursing home care.
The importance of being able to make changes in living arrangements is also seen in the additional months in the community gained by examining the living arrangements prior to entry of the subgroup of individuals who transitioned to nursing home residence. Relative to living alone, other stable living arrangements (with no changes, the vast majority as seen in Table 1) resulted in more months of community residence. However, those who made living arrangement changes all gained substantially in terms of additional months prior to entering a nursing home, regardless of the type of change.
This analysis has a number of limitations. Although we examine living arrangements and changes over a 10-year period, changes in living arrangement were assessed at 2-year intervals. As a result, there may be changes between these points that went undetected. In addition, our study, like most examining living arrangement change, does not distinguish between changes that involve an older person relocating to live with others versus others moving in—a distinction that may have implications for nursing home entry. We explicitly consider only two generations (parents and their adult children) in our measures of living arrangements and group the third generation (grandchildren) in the “other relatives” category. Because grandchildren are treated as attributes of the elderly respondent’s adult child in the HRS/AHEAD data and do not have a specific person identifier, we are unable to distinguish among grandchildren or track movements of individual grandchildren over time. How explicit inclusion of a third generation would affect these results is not known. We also had to strike a balance between creating completely homogeneous categories of living arrangement patterns—many of which were too small to study separately—and groupings that are not perfectly homogeneous but reflect highly prevalent common characteristics.
Our results are generalizable to the U.S. population aged 70 years and older. There is evidence that changes in living arrangements may vary within subgroups, however, including among minority populations (Angel, Angel, & Himes, 1992; Hays, 2002; Peek, Koropeckyj-Cox, Asembik, & Coward, 2004), and relationships between stability and change in living arrangements to nursing home entry could vary as well. Although clearly important, we opted not to pursue such interactions given the complexity of our models and the relatively small sample of minority persons. Finally, nursing home entry is a rare event, although one with major financial and quality-of-life implications for individuals and their families. Because it is a rare event, we were limited in our ability to assess risk for living arrangement patterns that were less common among older people.
Richer longitudinal data that allow examination of living arrangements over longer periods, and for subgroups of older people, enhances our understanding of living arrangement and household composition changes among older people and the role these play in meeting needs for functional assistance and maintaining community residence. Much research has focused on the risks of nursing home entry associated with living alone versus living with a spouse. Our study used 10 years of data and shows that long-term stable living arrangements carry different levels of risk for nursing home entry and that making changes in living arrangements both reduces risk of entry and slows time to entry. From a policy perspective then, it appears important to continue to explore and expand alternative living arrangement options for older people. Although coresidence with an adult child was the most protective against institutional care, the viability of this option over the long term may decline with changes in the American family both in terms of size and in terms of composition (Cherlin, 2009). Here too, policy proposals (and some existing public programs) that provide incentives to family caregivers may help shore up our societal reliance on family to meet the needs of older people who experience health declines.
FUNDING
This study was supported by National Institute on Aging R01 AG025475 (L.E.P.).
References
- Akaide H. A new look at the statistical model identification. IEEE Transactions on Automatic Control. 1974;19:716–723. [Google Scholar]
- Akamigbo AB, Wolinsky FD. New evidence of racial differences in access and their effects on the use of nursing homes among older adults. Medical Care. 2007;45:672–679. doi: 10.1097/MLR.0b013e3180455677. [DOI] [PubMed] [Google Scholar]
- Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav. 1995;36(1):1–10. [PubMed] [Google Scholar]
- Angel RJ, Angel JL, Himes CL. Minority group status, health transitions, and community living arrangements among the elderly. Research on Aging. 1992;14:496–521. [Google Scholar]
- Cagney KA, Agree EM. Racial differences in skilled nursing care and home health use: The mediating effects of family structure and social class. Journal of Gerontology: Social Sciences. 1999;54:S223–S236. doi: 10.1093/geronb/54b.4.s223. [DOI] [PubMed] [Google Scholar]
- Chatters LM, Hardison CB, Riley A. Informal social support networks and subjective well-being among African Americans. Journal of Black Psychology. 2001;27:439–463. [Google Scholar]
- Cherlin A. The marriage go-round: The state of marriage and the family in America today. Baltimore, MD: Knopf; 2009. [Google Scholar]
- Dunlop DD, Manheim LM, Song J, Chang RW. Gender and ethnic/racial disparities in health care utilization among older adults. Journal of Gerontology: Social Sciences. 2002;57:S221–S233. doi: 10.1093/geronb/57.4.s221. [DOI] [PubMed] [Google Scholar]
- Freedman VA. Family structure and the risk of nursing home admission. Journal of Gerontology: Social Sciences. 1996;51:S61–S69. doi: 10.1093/geronb/51b.2.s61. [DOI] [PubMed] [Google Scholar]
- Gaugler JE, Duval S, Anderson KA, Kane RL. Predicting nursing home admission in the U.S: A meta-analysis. BMC Geriatrics. 2007;7:13. doi: 10.1186/1471-2318-7-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Harrington C, Anzaldo S, Burdin A, Kitchener M, Miller N. Trends in state certificate of need and moratoria programs for long term care providers. Journal of Health & Social Policy. 2004;19:31–58. doi: 10.1300/J045v19n02_02. [DOI] [PubMed] [Google Scholar]
- Hays JC. Living arrangements and health status in later life: A review of recent literature. Public Health Nursing. 2002;19:136–151. doi: 10.1046/j.1525-1446.2002.00209.x. [DOI] [PubMed] [Google Scholar]
- Hays JC, Pieper CF, Purser JL. Competing risk of household expansion or institutionalization in late life. Journal of Gerontology: Social Sciences. 2003;58:S11–S20. doi: 10.1093/geronb/58.1.s11. [DOI] [PubMed] [Google Scholar]
- Hebert R, Dubois M, Wolfson C, Chambers L, Cohen C. Factors associated with long-term institutionalization of older people with dementia: Data from the Canadian study of health and aging. Journal of Gerontology: Medical Sciences. 2001;56:M693–M699. doi: 10.1093/gerona/56.11.m693. [DOI] [PubMed] [Google Scholar]
- Heyman A, Peterson B, Fillenbaum G, Pieper C. Predictors of time to institutionalization of patients with Alzheimer’s disease: The CERAD experience, part XVII. Neurology. 1997;48:1304–1309. doi: 10.1212/wnl.48.5.1304. [DOI] [PubMed] [Google Scholar]
- Himes CL, Wagner GG, Wolf DA, Aykan H, Dougherty DD. Nursing home entry in Germany and the United States. Journal of Cross-Cultural Gerontology. 2000;15:99–118. doi: 10.1023/a:1006797731500. [DOI] [PubMed] [Google Scholar]
- Jenkins CL. Resource effects on access to long-term care for frail older people. Journal of Aging & Social Policy. 2001;13:35–52. doi: 10.1300/j031v13n04_03. [DOI] [PubMed] [Google Scholar]
- Konetzka RT, Werner RM. Disparities in long-term care: Building equity into market-based reforms. Medical Care Research and Review. 2009;66:491–521. doi: 10.1177/1077558709331813. [DOI] [PubMed] [Google Scholar]
- Lau DT, Scandrett KG, Jarzebowski M, Holman K, Emanuel L. Health-related safety: A framework to address barriers to aging in place. The Gerontologist. 2007;47:830–837. doi: 10.1093/geront/47.6.830. [DOI] [PubMed] [Google Scholar]
- Lee ET, Wang JW. Statistical methods for survival data analysis. New York: John Wiley and Sons; 2003. [Google Scholar]
- McClendon MJ, Smyth KA, Neundorfer MM. Long-term-care placement and survival of persons with Alzheimer’s disease. Journal of Gerontology: Psychological Sciences. 2006;61:P220–P227. doi: 10.1093/geronb/61.4.p220. [DOI] [PubMed] [Google Scholar]
- Miller EA, Weissert WG. Predicting elderly people's risk for nursing home placement, hospitalization, functional impairment, and mortality: a synthesis. Medical Care Research & Review. 2000;57(3):259–297. doi: 10.1177/107755870005700301. [DOI] [PubMed] [Google Scholar]
- Miller SC, Prohaska TR, Furner SE, Freels S, Brody JA, Levy PS. Time to nursing home admission for persons with Alzheimer’s disease: The effect of health care system characteristics. Journal of Gerontology: Social Sciences. 1998;53:S341–S353. doi: 10.1093/geronb/53b.6.s341. [DOI] [PubMed] [Google Scholar]
- Mor V, Zinn J, Angelelli J, Teno JM, Miller SC. Driven to tiers: Socioeconomic and racial disparities in the quality of nursing home care. Milbank Quarterly. 2004;82:225–256. doi: 10.1111/j.0887-378X.2004.00309.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nihtila E, Martikainen P. Institutionalization of older adults after the death of a spouse. American Journal of Public Health. 2008;98:1228–1234. doi: 10.2105/AJPH.2007.119271. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nygaard HA, Alberktsen G. Risk factors for admission to a nursing home: a study of elderly people receiving home nursing. Scandinavian Journal of Primary Health Care. 1992;10:128–133. doi: 10.3109/02813439209014049. [DOI] [PubMed] [Google Scholar]
- Peek CW, Koropeckyj-Cox T, Asembik BA, Coward RT. Race comparisons of the household dynamics of older adults. Research on Aging. 2004;26:179–201. [Google Scholar]
- Reed SC, Andes S. Supply and segregation of nursing home beds in Chicago communities. Ethnicity and Health. 2001;6:35–40. doi: 10.1080/13557850124941. [DOI] [PubMed] [Google Scholar]
- Smith GE, Kokmen E, O’Brien PC. Risk factors for nursing home placement in a population-based dementia cohort. Journal of the American Geriatrics Society. 2000;48:519–525. doi: 10.1111/j.1532-5415.2000.tb04998.x. [DOI] [PubMed] [Google Scholar]
- Soldo BJ, Hurd MD, Rodgers WL, Wallace RB. Asset and health dynamics among the oldest old: An overview of the AHEAD study. Journal of Gerontology: Social Sciences. 1997;52 doi: 10.1093/geronb/52b.special_issue.1. Spec No, 1–20. [DOI] [PubMed] [Google Scholar]
- Spillman BC, Pezzin LE. Potential and active family caregivers: Changing networks and the “sandwich generation.”. Milbank Quarterly. 2000;78:347–374. doi: 10.1111/1468-0009.00177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wolff JL, Kasper JD. Caregivers of frail elders: Updating a national profile. The Gerontologist. 2006;46:344–356. doi: 10.1093/geront/46.3.344. [DOI] [PubMed] [Google Scholar]
- Yaffe K, Fox P, Newcomer R, Sands L, Lindquist K, Dane K, Covinsky KE. Patient and caregiver characteristics and nursing home placement in patients with dementia. Journal of the American Medical Association. 2002;287:2090–2097. doi: 10.1001/jama.287.16.2090. [DOI] [PubMed] [Google Scholar]