This article explores the relationship between older adults’ HCBS use and their future residential transitions by analyzing a nationally representative dataset (The Second Longitudinal Study of Aging, 2002). The University of Washington Human Subjects Division approved this study.
Data source- The second longitudinal study of aging (LSOA II)
The Second Longitudinal Study of Aging (LSOA II, 2002) was a collaborative effort of the National Center for Health Statistics (NCHS) and the National Institute on Aging (NIA). Using the LSOA II data, this study analyzed nationally representative civilian non-institutionalized persons aged 70
years or older. The LSOA II followed a stratified, multistage probability design that permitted continuous sampling of the target population. After baseline face-to face interviews in 1994 (Time 1 [T1]; N
9,447), two follow-up interviews occurred using Computer Assisted Telephone Interviews [18
]: one interview between 1997 and 1998 (Time 2 [T2]; N
7,060), and one interview between 1999 and 2000 (Time 3 [T3]; N
5,294). The overall response rate was over 74% [19
]. Loss of respondents was due to attrition from death, hospitalization, and/or loss during tracking. The current study used one sample weight along with two sampling-related parameters (strata and psu) to account for the LSOA II’s sampling survey design.
A total of 3,085 older adults were included in the data analysis for this study. They were older adults who (1) had completed all three LSOA II interviews, and (2) had functional limitations at the baseline interview (T1), meaning that each had limitations in at least one disability in either Activity of Daily Living (ADLs) [20
], Instrumental Activity of Daily Living (IADLs) [21
], or Nagi’s functional limitation [22
]. We selected the respondents with functional limitations because these older adults would be more likely than those who did not have any functional limitations to seek support services such as HCBS.
The current study investigated associations between 13 different types of HCBS and four types of residential transition patterns. We further included 14 covariates, which were factors based on Anderson’s Health Behavioral Model (HBM).
Types of HCBS
In general, each specific service within the category of HCBS can be assigned to one of two categories: nondiscretionary or discretionary. Nondiscretionary services typically require prescriptions from health care professionals. Physical therapy and skilled nursing care are examples of nondiscretionary services. In contrast, discretionary services are typically used as a matter of individual choice; homemaker/companion services and personal care services (PCS) are examples of nondiscretionary services [2
The LSOA II data documented 13 services used by respondents between the T1 and T2 interviews. These 13 services were (a) senior centers, (b) Meals On Wheels, (c) meals at senior centers/facilities, (d) homemaker/companion services, (e) personal care services (PCS), (f) skilled nursing care, (g) physical therapy, (h) occupational therapy, (i) speech therapy, (j) dialysis, (k) tube feeding, (l) oxygen or respiratory therapy, and (m) hospice care. For the purpose of this study, the first five services were considered discretionary services; all the other services were considered nondiscretionary services. Two services (senior centers and meals at senior centers/facilities) were received outside the home, while all other services were received in the home. Because the T2 interview gathered the most detailed information on HCBS use, the current study used data from the T2 interview for information regarding HCBS use.
The LSOA II asked both 12-month and 2-year retrospective questions about HCBS use. The question asked at T2 interview for the first three services (services a through c) covered 12
months: “In the past 12
months, did you go to/use … [one of these services]?” [25
]. For the remaining 10 services [services d through m], the questions covered 2
years. There were two questions: “Since [month/year of last interview] did you receive any health care services IN YOUR HOME? This would include skilled nursing care, physical or occupational therapy, assistance with medications or personal care needs, and any other services provided IN YOUR HOME by a visiting nurse, nursing assistant, home health aide, personal assistant, therapist, or homemaker”; and “Which of the following services did you receive? Did you receive (01) Skilled nursing care (02) Physical therapy (03) Occupational therapy (04) Speech therapy (05) Dialysis (06) Tube feeding (07) Personal assistant services (08) Homemaker/companion services (09) Oxygen / respiratory therapy (10) Hospice care.” [25
]. All 13 HCBS variables were used to assess older adults’ use of HCBS between T1 and T2.
Patterns of residential transition
At each interview (T1, T2, and T3), each LSOA II respondent was living either in a home- and community-based setting (C) or in an institution (I). Home- and community-based settings included (a) single-family homes, (b) regular apartments, (c) retirement homes, (d) assisted living facilities, (e) supervised apartments, (f) group homes, (g) halfway houses, (h) boarding homes, and (i) developmental centers. Institutions included (j) nursing homes and (k) convalescent homes. All the older adults included in the LSOA II lived in communities (C) at the T1 interview. The questions asked at T2 and T3 interviews regarding residential status were these: “Is the place where you live a … [one of the 11 options described above]?” and “Since the last interview, have you been a resident in a nursing home/convalescent home?” (The Second Longitudinal Study of Aging—The Second Supplement on Aging, 1994). Each respondent who indicated residence in a nursing home, in answers to either of these two questions, was considered to have transited to an institution (I) during that period of time. Respondents whose answers did not indicate nursing home use were considered to be living in community (C). Respondents transitions between living arrangements were noted. The LSOA II collected data three times. Using these three time points, the current study defined four types of residential transitions: (1) CCC: older adults who resided in community from T1 to T3 and did not use any nursing home service during the entire study period (from 1994 to 2000); (2) CIC: older adults who resided in community at T1, had lived in an institution at some time between T1 and T2, including at T2, then had returned to community by T3 and had not used any nursing home services between T2 and T3; (3) CCI: older adults who resided in community between T1 and T2, including at T2, and did not use any nursing home services during this period of time, but had used nursing home services between T2 and T3, including at T3; and (4) CII: older adults who resided in community at T1 but in an institution at some time between T1 and T2, including at T2, and at some time between T2 and T3, including at T3. Figure shows how the four groups of older adults were defined by type of residence from T1 interview to T3 interview, and when the HCBS use data was collected.
Older Adults’ HCBS Use and Residential Transitions From Time 1 to Time.
Covariates from the health behavioral model
This study included 14 covariates based on factors from Anderson’s Health Behavioral Model (HBM), one of the most widely used behavioral models. The HBM posits that use of health services and residential transitions both are functions of three types of factors: factors that predispose people to use services, such as age, gender, marital status, race, region, and years of education; factors that enable or impede people’s use of such services, such as insurance information, family income, and size of family; and factors that affect people’s personal need for care, such as self-rated health, number of difficulties with Nagi’s Functional limitations, ADLs, and IADLs [20
]. The HBM covariates, as well as older adults’ use of HCBS, were assessed as influences on older adults’ residential transition patterns [28
]. Akaigbo and Wolinsky (2006) reported that older adults with a history of hospital use were associated with subsequent nursing home placement. As a result, we included among the covariates the variables of older adults being a hospital patient overnight either between T1 and T2 or between T2 and T3. All covariates were obtained from the T1 interview except for the variables of being a hospital patient overnight between T1 and T2 and being a hospital patient overnight between T2 and T3, which were drawn from the T2 and T3 interview data respectively.
Missing values in the study variables represented less than 5% of observations, with the exception of the HCBS variables (missing 8.8% to 15.6% of observations) and the income variable (missing 21% of observations). As a result, missing values from responses such as “not ascertained” and “don’t know or refused” were replaced using the Markov chain Monte Carlo method through the multiple-imputation procedure in LISREL 8.53 [33
The first research question was whether there are differences in service use among groups of older adults defined by four residential transition patterns. Since each of the 13 HCBS was a dichotomous variable, we used logistic regression analysis, with predisposing, enabling, and need factors (Including age, gender, education level, family income, size of family, number of functional limitation, number of ADLs, Number of IADLs, and being in a hospital patient over night between T1 an T2 interviews and T2 and T3 interviews) controlled, to answer this question. As a result, a series of 13 logistic regression analyses was performed. Each analysis assessed the relationship between one HCBS and four residential transition pattern. Variables were considered significant at P value ≤ 0.05, 2-sided (SPSS manual, version 16).
The second research question was which HCBS are associated with which residential transition patterns. This study intended to explore (1) what services might help older adults to stay in their communities longer, (2) what services might help older adults move back to their communities once they have been institutionalized, and (3) whether using different services in combination might associate with different residential transition outcomes. To address this question, we used estimated marginal means produced by ANCOVA, controlled for age, gender, education level, family income, size of family, number of functional limitations, number of ADLs, number of IADLs, and overnight hospital stay(s) between T1 and T2 interviews and T2 and T3 interviews, and then created bar graphs depicting patterns of service use in different transition groups. To address the complex sample design used in LSOA II, the STATA 9.0 survey suite was used for statistical analysis.