Abstract
Delays to institutionalization were compared between elderly individuals who differed in the amounts (“dosages”) of adult day services (ADS) they attended. A Kaplan–Meier survival analysis revealed higher dosages of ADS to be associated with greater delays to institutionalization. Retrospective data from financial and service utilization systems and from the Resident Assessment Instrument for Home Care (RAI-HC) were then used to fit a Cox regression model that was adjusted for potential selection biases. This model also found systematically lower hazards for institutionalization at higher ADS dosages. The ADS effect did not appear to be an artifact of increased utilization of additional home health services. Results suggest a beneficial effect of ADS on delay to institutionalization that cannot be attributed to home support, respite, or case management services.
Keywords
As the population ages and health-care costs escalate, viable community-based solutions are imperative to meet older adults’ functional, social, clinical, and mental health needs. Moreover, given the current policy push to delay or avoid institutional care, community programs are increasingly being called on to support individuals in their homes for as long as possible. Adult day services (ADS) are situated amid the continuum of home support services, which are designed to support older adults with functional and/or cognitive impairment so that they can continue to live at home.
Researchers have contended that the effect of ADS operates through mechanisms related to (a) attendance by the participant, (b) specific interventions such as social programs and health monitoring for the participant or information and support for the caregiver, or (c) respite for the caregiver (Fields, Anderson, & Dabelko-Schoeny, 2012; Gaugler & Zarit, 2001; see also Diesfeldt, 1992; Gottlieb & Johnson, 2000). The actions of these mechanisms can be evident in participant-based outcomes, such as improved mood and functioning, or caregiver-based outcomes, such as reduced burden. Effects can also be evident in system-based outcomes, such as utilization of heath care services. For instance, ADS is believed to influence participants’ use of emergency department, acute care hospitalization, and survival in the community before long-term institutionalization is required.
With system-based outcomes, research has focused extensively on the effect of ADS on time to nursing home placement (e.g., Cho, Zarit, & Chiriboga, 2009; Gaugler, Kane, Kane, & Newcomer, 2005; McCann, Hebert, Li, Wolinsky, Gilley, Aggarwal, Miller, & Evans, 2005), but studies have produced ambiguous results. In some research, placement was delayed by ADS (Gaugler et al., 2005) or particular types of ADS (Droes, Breebaart, Meiland, van Tilburg, & Mellenbergh, 2004; Gitlin, Reever, Dennis, Mathieu, & Hauck, 2006). In other studies, ADS appeared to actually hasten placement (Lyons, Zarit, & Townsend, 2000; Wattmo, Wallin, Londos, & Minthon, 2011). Still other research demonstrated no effect of ADS on nursing home placement (Diesfeldt, 1992; Gilleard, Gilleard, & Whittick, 1984) or disclosed ADS effects that are complicated by other factors, such as the participant’s relationship to the caregiver (Cho et al., 2009), the time point of the caregiving “career” (Gaugler et al., 2005), or the rate of short-term cognitive decline (Wilson, McCann, Li, Aggarwal, Gilley, & Evans, 2007). The amount, or “dosage,” of ADS is another factor that has been shown to modulate the effect on nursing home placement (Gaugler, Kane, Kane, Clay, & Newcomer, 2003).
There has long been interest in the effects of ADS dosage (Gaugler & Zarit, 2001; Harder, Gornick, & Burt, 1986), but here again research results have been inconclusive: The relationship between ADS dosage and likelihood of nursing home placement has been both positive (Wilson et al., 2007) and negative (McCann et al., 2005). Paradoxically, Gaugler and his colleagues (Gaugler, Kane et al., 2003) uncovered a curvilinear relationship, leading them to suggest that low amounts of ADS are non-therapeutic, whereas high amounts may actually represent a gradual transition to placement in nursing homes—essentially, an indication that caregivers are beginning to surrender their caregiving responsibilities (also see Hope, Keene, Gedling, Fairburn, & Jacoby, 1998).
How can the inconsistent effects of ADS and ADS dosage on nursing home placement be reconciled? It is possible that conflicting results may be due, at least in part, to design and methodological differences between the various studies. For instance, in some studies, use of ADS may have been too low, and consequently treatment effects on nursing home placement may have been too weak (Gaugler et al., 2005). Other studies used small sample sizes (e.g., Hope et al., 1998), had research goals other than direct investigation of ADS effects on nursing home placement (e.g., Lyons et al., 2000), or relied on self-report to determine the amount of ADS use (Gaugler, Kane et al., 2003; Gaugler et al., 2005; McCann et al., 2005).
The objective of the present study is to provide the first direct and systematic investigation of ADS dosage on time to placement in an assisted living or nursing home facility (i.e., or, more generally, institutionalization in any long-term care facility). We used a large and broad set of administrative data from home care clients, including assessment and outcome data from the Resident Assessment Instrument for Home Care (RAI-HC) as well as electronic records from financial (family income) and service utilization systems (home health program enrollment, home support and respite utilization, ADS days attendance, and number of emergency room registrations and hospital admissions).
We designed a quasi-experimental study, in which ADS attendance data were used to quantitatively establish four levels of ADS dosage as the independent variable. Levels were then labeled as “High,” “Moderate,” “Low,” and “None,” and individuals were retrospectively assigned to one of the dosage levels, based on their actual usage of ADS. Next, individuals’ baseline levels on a number of key covariates were derived from administrative data, including service utilization and family income data as well as from their initial RAI-HC assessments. We then compared ADS groups’ days between home health program enrollment and institutionalization using a Kaplan–Meier survival analysis. Finally, a Cox Proportional Hazards Model was fitted, adjusting for potential preexisting covariate differences between groups, to examine the isolated effect of ADS dosage on institutionalization.
Similar to previous research on the effects of respite and ADS on outcomes (e.g., Iecovich & Biderman, 2013; Kosloski & Montgomery, 1995; McCann et al., 2005), this study was couched in terms of Andersen’s (1995) Behavioral Model of Health Service Utilization. According to the Anderson model, health service utilization is determined by three general categories of factors: predisposing (e.g., age, sex, and marital status), enabling (e.g., family support, availability, and current use of health services), and need factors (i.e., perceived and evaluated need for health care services). Within this framework, we were able to unambiguously evaluate the effect of ADS dosage on time to institutionalization in either an assisted living facility or a nursing home.
Method
Context
The Fraser Health (FH) Authority provides acute care, public health, mental health, home, community, and residential care services to a population of approximately 1.6 million people within a large area of the lower mainland of the Province of British Columbia in Canada. Residents of FH who have chronic health care needs, mostly frail elderly persons, are eligible to receive subsidized home health services including formal case management/assessment, home support (i.e., personal care), facility respite, and day programs for older adults (i.e., ADS). At any given time, there are between approximately 10,000 and 11,000 clients enrolled in the home health program, with a yearly turnover of between approximately 5,000 and 6,000 clients. Clients typically receive their initial financial and RAI-HC assessments shortly after enrollment in the home health program.
Approximately 267 clients attend ADS per day in 1 of 17 different day programs across the authority. Most programs are open 5 days per week, operate near capacity, and maintain a lengthy waitlist. FH makes every effort to standardize ADS programs and procedures throughout the region. The program includes standardized admission and discharge procedures, health monitoring, personal care, meals, and social/recreational activities. To enter the day program, client and/or caregiver needs are assessed by a case manager, using the RAI-HC. Day program attendance is then authorized, if needed. Although ADS is subsidized by FH, clients are also charged a nominal fee. Discharge from the home health and ADS programs usually occurs concurrently, usually because a client has been institutionalized or has passed away.
Sample
Individuals who met the following criteria were included in this study: their first enrollment in the FH home health program occurred in the 4-year period between January 1, 2009, and December 31, 2012; their initial financial and RAI-HC assessments had been completed; they were 65 years of age or older on the reference date of the assessment; their sex, birth date, marital status, and level of education had been recorded on the assessment; they were not designated as “very high needs” clients (i.e., clients whose care options are self-managed by their caregivers—such clients are extreme outliers for home support utilization); and their current status with respect to the home health program (currently enrolled or discharged and, if discharged, the date and reason for the discharge) could be determined. In total, there were 16,012 individuals who met these criteria.
Independent Variable
For each individual, total days of ADS attendance (i.e., participation) after admission to the home health program was obtained. ADS staff and previous research was consulted to define four levels of ADS dosage that would be likely to have noticeably differential effects on rate of institutionalization. We were also guided by the notion that effectiveness of ADS should be a function of both the frequency and the overall duration of attendance. In other words, for a person attending ADS frequently over a shorter period of time, effectiveness should be relatively similar to a person attending less frequently but over a longer period of time. We quantified this “consistency of attendance” construct by computing an ADS index for each individual. An individual’s ADS index was obtained by multiplying the total number of days of attendance by the total number of months between their first and last recorded date of attendance, then dividing by 30 (i.e., the approximate number of days in a month).
Under this consistency assumption, we defined a “High” ADS user as the equivalent of someone who attended ADS at least 96 days over 12 months (i.e., an index of 38.4 or greater). This would also include, for examples, individuals who had attended ADS at least 128 days over 9 months or at least 72 days over 16 months. At least over the shorter term, this level of attendance surpasses the “2 days and 8 hr per week” dosage that Gaugler and his colleagues reported as therapeutic (Gaugler, Zarit et al., 2003; Gaugler, Zarit, Townsend, Stephens, & Greene, 2003; see also Wilson et al., 2007).
By contrast, we defined a “Low” ADS user as the equivalent of someone who attended ADS at least once, but no more than 18 days over 12 months (i.e., with an index greater than 0 but less than or equal to 7.2). Again, this level of use appears to fall within the presumably “subtherapeutic” range, according to investigators who reported weak or null ADS effects (e.g., Gaugler, Zarit et al., 2003).
Consequently, a “Moderate” user was defined as an individual whose index score fell between those of a High and Low user (i.e., more than 7.2 but less than 38.4). Finally, individuals who had never attended ADS (i.e., index = 0) were assigned to group “None.” With these definitions, the numbers (and percentages) of clients retrospectively assigned to the High, Moderate, Low, and None groups were, in order, 513 (3.2%), 434 (2.7%), 530 (3.3%), and 14,535 (90.7%).
Covariates
Individuals’ initial RAI-HC assessment was the first of two sources of covariates. From these assessments, the following baseline variables were obtained: For clients: sex; age; marital status; education; behavioral symptoms of wandering, verbal abuse, physical abuse, social disruption, and resistance to care; relationship to primary caregiver; diagnosis of dementia, either Alzheimer’s or non-Alzheimer’s; diagnoses of cerebrovascular accident, congestive heart failure, coronary artery disease, hypertension, cancer, diabetes, emphysema/chronic obstructive pulmonary disorder/asthma, and renal failure. For clients’ caregivers: ability to continue the caregiving role; feelings of satisfaction and distress. And for both clients and/or caregivers: feelings that clients would be better off in a different environment.
Clients’ outcome scores from the initial RAI-HC assessment were also computed: viz., Cognitive Performance Scale (CPS), Changes in Health, End-Stage Disease, Signs and Symptoms (CHESS), Method for Applying Priority Levels (MAPLe), Activities of Daily Living Self-Performance Hierarchy (ADL SELF), Instrumental Activities of Daily Living Difficulty (IADL DIFF), Depression Rating Scale (DRS), and Pain Scale (PS).
Financial and service utilization systems provided the second set of covariates. From these systems we obtained clients’ family income on admission to the home health program; months of home support and number of uses of respite; months of enrollment in the home health program; and the number of times registered in an FH emergency room or admitted to an FH hospital in the 90-day period prior to enrollment in the home health program.
To simplify the analysis, many baseline variables were converted to covariates and then coded. Some continuous variables were converted to categorical covariates in such a way to keep the percentage of clients in each category relatively equal (e.g., age was converted to an age group covariate with four categories: 65 to 78 years, 79 to 84 years, 85 to 88, and 89+ years, each category having between 22.2% and 27.3% of clients). However, continuous variables that were markedly skewed (months of home health enrollment and home support, and uses of respite) were not converted.
Most discrete variables were dichotomized, again in such a way to best equalize the percentage of clients over the two categories of the covariate (e.g., CPS was recoded 1 if the CPS score was greater than 1, otherwise recoded 0). For some nominal variables, the covariate was coded exactly the same as the RAI-HC variable (e.g., sex was coded 0 for female, 1 for male). Other nominal variables were categorized in such a way to ensure a substantial percentage of clients in each category, while retaining at least some meaning for the covariate (e.g., marital status was coded 1 if “married” and 0 if any other marital status [e.g., single, widowed, divorced, etc.] had been recorded in the assessment). Finally, in some instances, two or more variables were merged to create a single covariate (e.g., dementia was coded 1 if either Alzheimer’s or non-Alzheimer’s had been diagnosed, otherwise dementia was coded 0). Table 1 displays the covariates used in the analysis, grouped according to the three factors of the Anderson Behavioral Model. The table also shows from which system and underlying variables the covariates were derived, and the coding scheme and the percentage distribution of clients over the various categories for each covariate.
Covariate Coding Scheme.
Note. RAI-HC = Resident Assessment Instrument for Home Care; CPS = Cognitive Performance Scale; CHESS = Change in Health, End Stage Disease, Signs and Symptoms; MAPLe = Method for Assigning Priority Levels; ADL SELF = Activities of Daily Living Self Performance Hierarchy; IADL DIFF = Instrumental Activities of Daily Living Difficulty; DRS = Depression Rating Scale; PS = Pain Scale.
Dependent Variable
For each individual, it was determined whether or not the individual had been admitted to an assisted living facility and/or residential care (i.e., nursing home) on or before June 25, 2013. If they had been admitted to one or sequentially to both types of facilities, the total number of days between their enrollment in the home health program and their admission(s) to the facility(s) was determined. For individuals admitted to both types of facilities, the smaller number of days between enrollment and admission was used as the number of days to institutionalization. Individuals not institutionalized but no longer in the home health program due to death or some other reason were censored, and the total number of days they were enrolled in the home health program was obtained. Finally, individuals currently active in the home health program were also censored, and the current total number of days of enrollment as of June 25, 2013, was obtained.
Analysis
First, a Kaplan–Meier analysis was conducted and the Mantel–Cox log rank χ2 statistic was used to test group differences in survival to institutionalization. Next, differences between groups on each covariate were evaluated using a c2 test for categorical covariates and an F test for continuous variables. Finally, all covariates that showed significant differences in percentage distribution or means were included in a Cox Proportional Hazards analysis that modeled time to institutionalization.
Covariates were grouped into blocks that corresponded to the three factors of the Anderson Behavioral Model (see Table 1). Two separate Cox models were developed: the first model fitted data from all four groups, the second model fitted data from the three ADS treatment groups only. In both Cox models, the three blocks of covariates were entered in sequence and tested simultaneously. The enabling factor was entered first, followed by the need factor and then the predisposing factor. Prior testing indicated that this order of block entry ensured the greatest omnibus reduction of −2 log likelihood (−2LL).
In both Cox models, the ADS factor was entered alone in the fourth block. Entry of covariates within each block was forward stepwise, and the probabilities for entry into and removal from the model were set to 0.05 and 0.10, respectively. Reductions in −2LL were examined after each block using a c2 test, to determine whether or not the block factor was significant in the Cox model. Hazard ratios (HRs) were computed for all covariates within each block. (To interpret HRs, refer to the covariate coding in Table 1.) HRs for the different levels of the ADS factor were compared using repeated contrasts, in which the lower ADS dosage served as the reference group for the next higher dosage.
For all tests, significance was defined as p < .001. All analyses were conducted using IBM SPSS Statistics v. 19.
Results
The Kaplan–Meier survival analysis (Figure 1) displays an orderly and significant relationship between ADS dosage and days to institutionalization (c2[1] = 128.05). The 4-year (i.e., 1,460-day) survival function for the None group began to decline immediately, whereas there were periods of maintenance in the three ADS groups before their respective functions began to decline. Moreover, the longest period of maintenance occurred in the High group, followed by the Moderate group, then the Low group. Interestingly, with increasing days to institutionalization along the x-axis, there was a trend for the three ADS group survival functions to converge with and, in the case of the Low group at about the 365-day mark, even cross over that of the None group function. Of course, these differences in survival functions could have been due to preexisting differences between individuals in the four groups (i.e., a group self-selection bias).
Kaplan–Meier survival curves for individuals at each level of adult day service dosage.
Table 2 displays the number and percentage distribution of individuals over each level of each covariate within each Behavioral Model factor, cross tabulated by ADS group. In regard to the predisposing factors, individuals not attending ADS as a group were generally older, more likely to be female, less likely to be married, and had lower levels of education. The percentage of individuals with dementia was low in all four groups.
Group Totals (Percentages) and Means (Standard Deviations).
Note. CPS = Cognitive Performance Scale; CHESS = Change in Health, End Stage Disease, Signs and Symptoms; MAPLe = Method for Assigning Priority Levels; ADL SELF = Activities of Daily Living Self Performance Hierarchy; IADL DIFF = Instrumental Activities of Daily Living Difficulty; DRS = Depression Rating Scale; PS = Pain Scale.
p < .001.
With respect to the enabling factors, months of home support utilization and home health enrollment and number of respite uses was much higher in groups attending ADS, especially in the High condition. By contrast, non-attendees were more likely to have registered in an emergency room or been admitted to hospital in the 90-day period prior to admission to home health. As a group, ADS attendees were more likely to have a spouse as primary caregiver, no caregiver burden, and a higher household income.
Finally, regarding the need factors, groups were similar in percentages with problem behaviors, IADL difficulties (IADL DIFF > 4), and depression (DRS > 0). Non-attendees of ADS had greater percentages with chronic disease, clinical instability (CHESS > 1), ADL dependencies (ADL SELF > 0), and pain (PS > 1), whereas ADS attendees had greater percentages with cognitive difficulties (CPS > 1) and risk of institutionalization (MAPLe > 3). The last two columns of Table 2 indicate the significant covariates with an asterisk; for Model 1, only behavioral symptoms, IADL DIFF and DRS, were not significant and, hence, were excluded from this model. By contrast for Model 2, only dementia, home support utilization months, home health enrollment months, respite service uses, and chronic disease were significant and, hence were included in this model.
Table 3 displays covariate order of entry (if entered), final HRs, and omnibus reductions in −2LL for each block of the two Cox models. Each of the first three blocks of Model 1 produced a significant reduction in −2LL, and most covariates revealed a significant HR. For instance, presence of caregiver burden was associated with 2 times the risk of institutionalization compared with absence of caregiver burden, whereas each additional month of home support utilization was associated with a decreased risk of institutionalization, by 0.95 times (or conversely, each decrement in home support months increased institutional risk by the reciprocal of the HR, i.e., 1 / 0.95 = 1.05 times).
Cox Proportional Hazards Analyses.
Note. ADS = Adult Day Services; LL = log likelihood; CPS = Cognitive Performance Scale; ADL SELF = Activities of Daily Living Self Performance Hierarchy; MAPLe = Method for Assigning Priority Levels; CHESS = Change in Health, End Stage Disease, Signs and Symptoms; PS = Pain Scale.
p < 0.001.
The final four rows for Model 1 display the HRs for successive levels of the ADS factor. This block produced a further significant reduction in −2LL, but only the Moderate dosage had a hazard that was significantly (0.53 times) lower than its reference (Low) group. Statistically, there were no HR differences between the None and Low dosage or between the Moderate and High dosage.
Table 3 also displays results for Model 2. Reductions in −2LL were smaller but still significant for the enabling and need factors, but not for the predisposing factor. Only home health enrollment months and home support utilized months revealed significant HRs. Nevertheless, entry of the ADS dosage block still produced a further significant reduction in −2LL. In this model, the HR for High dosage was significantly (0.61 times) lower than its reference (Moderate) group, and the HR for the Moderate dosage was significantly (0.58 times) lower than its reference (Low) group.
Discussion
The present study provides compelling evidence for the beneficial effects of day program attendance, and shows that these effects increase systematically as consistency of day program attendance increases (Figure 1). When a statistical model was fitted to account for a considerable number of key covariates on which there may have been preexisting group differences, the beneficial effects of ADS and ADS dosage were still evident. Specifically, the risk of institutionalization was lower for individuals who attended ADS at a Moderate or High dosage (Model 1), and the hazard decreased systematically with increasing ADS dosage (Model 2). These findings are consistent with other research in which day program attendance was shown to reduce the risk of nursing home placement (e.g., Gaugler et al., 2005; Wilson et al., 2007). The present study extends these previous findings to include risk of institutionalization in general, viz., placement in either an assisted living facility or a nursing home. In addition, to our knowledge this is the first study that directly displayed an orderly relationship between increasing ADS dosage and decreasing risk of institutionalization.
Nevertheless, it is difficult to reconcile these findings with studies that found day program attendance to increase the risk of nursing home placement (e.g., Lyons et al., 2000; Wattmo et al., 2011) or that found increasing amounts of day program attendance increased the risk of placement (Gaugler, Kane et al., 2003; McCann et al., 2005). Perhaps closer examination of the Kaplan–Meier analysis of the present study (Figure 1) might shed some light on one possibility for the difference in findings. Prior to the Cox model adjustments, the Kaplan–Meier analysis demonstrated higher rates of survival in the three ADS treatment groups, with groups at higher ADS doses surviving longer than those at lower doses. But as time passed, the survival functions for the three groups receiving ADS approached and, in the case of the Low group, crossed over the function for the group never attending the day program. Note that these unadjusted data are similar to what Gaugler, Kane et al. (2003) and McCann et al. (2005) reported, viz., at higher ADS dosages the risk of nursing home placement increases, at least after a period of time has passed.
So why then did the association persist in Gaugler, Kane et al.’s (2003) and McCann et al.’s (2005) adjusted Cox models but not in the adjusted models of the present study? One possibility is there was greater precision in key baseline variables that were used as covariates in the present study. Specifically, we had access to highly reliable administrative data sets to determine the exact amounts of home support, respite, and day program utilization throughout each individuals’ enrollment in the home health program, as well as individuals’ exact durations in the home health program, registrations in emergency rooms and admissions to hospital in the 90-day period prior to enrollment. By comparison, other studies tended to rely on caregivers’ self-report for such data, and thus these covariates likely lost much of their precision. It is notable that these covariates fell in the enabling factor category in the present study, and that this category accounted for most of the reduction in −2LL in both Cox models (see Table 3). In other words, these covariates appear to be robust predictors of institutionalization. As a consequence, the fit of the statistical model is a function of the precision of the covariates. Whether this is an adequate explanation for the difference in findings between this study and Gaugler, Kane et al.’s (2003) and McCann et al.’s (2005) is uncertain.
An observation that deserves some attention is the relationship between ADS dosage and months of home health enrollment, months of home support, and number of times respite was used. It stands to reason that the longer an individual remains in the home health program (i.e., avoids institutionalization), the greater likelihood the individual will more frequently use the services of the program: case management, home support, respite, and day program. Certainly, the data in Table 2 are consistent with this reasoning. So how can we be certain that the day program bestowed additional benefit, over and above that conferred by these other home health services? Admittedly, in the context of a quasi-experimental study, it is impossible to completely disentangle the confounded effect of ADS dosage from those of home health and home support durations and uses of respite.
We acknowledge that there likely were benefits of added case management, home support, and respite that accrued with continued avoidance of institutionalization by individuals in the home health program. Nevertheless, we also believe that the day program bestowed additional benefit on its attendees. To test this possibility, we did a further analysis on data within the high ADS dosage group. All individuals in this group attended the day program consistently, and thus all had been enrolled in the home health program for substantial periods of time. Therefore, all individuals would have had similar periods of time to utilize other home health services. Consequently, there should be substantially reduced relationships between months of day program attendance and months of home health enrollment, months of home support, and number of times respite was used. So if the within-group analysis still showed increasing duration of day program attendance to reduce the risk of institutionalization, this would support the interpretation that the day program bestowed benefit, over and above that conferred by the other home health services.
For the high ADS dosage group only, logistic regression was used to predict institutionalization from months in the home health program, months of home support, months of day program attendance, and the number of respite uses. This analysis revealed that only months in the home health program and months of day program were significant. Interestingly, the odds of institutionalization increased with increasing months in the home health program (odds ratio = 1.05), but decreased with increasing months in the day program (odds ratio = 0.93). In other words, as months in the home health program increased, individuals became more and more at risk of institutionalization. However, institutional risk was reduced if they attended the day program. In fact, of individuals in the high ADS dosage group, descriptive analyses indicated that those who were institutionalized (n = 174) received even slightly more months of home support and used respite more often (M = 11.7 and 1.5, respectively) compared with those who were not institutionalized (n = 339, M = 11.6 and 0.9, respectively). Presumably, the additional day program attendance by individuals in the non-institutionalized group compared with the institutionalized group (M = 25.7 and 23.3 months, respectively) mitigated the risk of institutionalization experienced by individuals in the non-institutionalized group.
In conclusion, this study provides compelling evidence for the beneficial effect of ADS on a system-based outcome—that is, reduced risk of institutionalization—on a population of community-dwelling seniors. Results further indicate that benefits increase systematically with higher dosages, and are not likely due to additional home health services that are typically provided to consistent ADS attendees. Further studies should determine if similar ADS effects prevail with other system-based outcomes (e.g., number of emergency room registrations and acute care hospitalizations, overall length of stay in hospital). It would be of interest as well to use individuals’ subsequent RAI-HC assessments to reveal the ADS effect on client- and caregiver-based outcomes (e.g., changes in client ADL function or clinical status, or changes in caregiver burden), and whether or not effects are similarly dosage dependent.
Limitations
The main limitation of the present study is the absence of random assignment to treatment and control groups. Due to the retrospective nature of this research, there is the possibility that individuals self-select their own groups. Although a statistical attempt was made to reduce effects due to self-selection bias, such effects can never be practically eliminated short of a controlled study using random assignment.
Another limitation pertains to the generality of the present findings. There appears to be considerable variation in what constitutes “adult day programs,” not only in practice but also in the published literature. Evidently, adult day programs can vary in the extent to which they emphasize social and emotional health versus clinical monitoring and therapeutic well-being, or mix the two components (Forster, Young, Lambley, & Langhorne, 2008; Harder et al., 1986). Insofar as day programs in FH gravitate more to the social and emotional model, present results are likely more applicable to similar types of programs in other jurisdictions.
There is at least one practical limitation to this study, as well. In FH, attendance in day programs does not offset utilization of other home health services. In fact, allocation of all home health services, including day program, is governed by a policy that recommends all service options must be exhausted prior to consideration of institutionalization. Consequently, case managers encourage clients and their families to maximize all other non-institutional service options, and referring clients to assisted living facilities or nursing homes is discouraged unless absolutely necessary. Thus, it is possible that results of the present study reflect a sampling bias, wherein higher users of day program are more motivated to utilize all available resources before considering an institutional alternative. In addition, findings may reflect the proclivity of case managers to encourage high service use as well as variations in case manager referral practices. We did not attempt to account for these policy effects or referral-service allocation practices of case managers. Further research needs to incorporate contextual and system-level factors to explore these complex issues.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.