NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Wysocki A, Butler M, Kane RL, et al. Long-Term Care for Older Adults: A Review of Home and Community-Based Services Versus Institutional Care [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Nov. (Comparative Effectiveness Reviews, No. 81.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Long-Term Care for Older Adults

Long-Term Care for Older Adults: A Review of Home and Community-Based Services Versus Institutional Care [Internet].

Show details


Key Findings and Strength of Evidence

The literature base for head-to-head comparisons between recipients of Home and Community-Based Services (HCBS)—including assisted living (AL) residents—and residents of nursing homes (NHs) is limited. Although a considerable body of literature in the long-term care (LTC) field separately addresses HCBS and NH populations, few studies directly compare the two. Some literature examines the relative effectiveness of various HCBS programs in reducing admittance to NHs, thus assuming NH admission to be a negative outcome. Yet, in reality, some individuals may not benefit more from HCBS than from NH care, and HCBS may not justify increased private and public costs. (This is especially true if the individuals are unaware of their environments.) Moreover, NHs could have theoretical advantages in quality of care domains compared with HCBS programs.

This review addresses the direct comparison of HCBS and NHs. Included studies showed that on average NH residents were more impaired than HCBS recipients (including AL residents) in terms of physical and cognitive function, but results were mixed for mental health and clinical status measures. The distributions of HCBS recipients and NH residents on these measures overlapped. This overlap suggests that many people could be cared for by either approach. However, the difference in setting (HCBS versus NH, or AL versus NH) likely reflects other concerns, especially the availability of informal support for HCBS recipients, availability of housing (including residential homes), and state and provider policies that render HCBS flexible and affordable.

Table 29 summarizes the key findings and strength of evidence for the outcome domains addressed in the review. We found very few studies from which to draw conclusions for any given outcome of interest. Sparse literature and design problems resulted in low-strength or insufficient evidence for the seven outcomes we examined in detail.

Table 29. Summary of key findings and strength of evidence for outcome domains.

Table 29

Summary of key findings and strength of evidence for outcome domains.

Overall, the studies provided low-strength evidence that the rate of change in physical function, cognition, and mental health did not differ significantly between AL and NH residents. Evidence was insufficient for these outcomes for the HCBS versus NH comparison. Two studies provided low-strength evidence that mortality did not differ between AL and NH residents, but evidence was insufficient for the HCBS versus NH comparison. Evidence was also insufficient for use of acute care for both the AL versus NH comparison and the HCBS versus NH comparison.

Two studies provided low-strength evidence that harms differed between HCBS recipients and NH residents. HCBS recipients experienced higher rates of some harms and NH residents experienced higher rates of other harms. Evidence was insufficient for harms for the AL versus NH comparison.

One study provided insufficient evidence that Medicaid expenditures were higher for HCBS recipients compared with NH residents, but other program and individual expenditures were not analyzed. No studies analyzed expenditures for AL versus NH residents.

Several factors must be considered in interpreting this body of evidence. Most studies did not sufficiently describe the settings and services received. Few studies specified the type, frequency, or intensity of services. For NHs and AL, particular settings studied will dictate the services available. However, individual variation in care plans remains possible (if not likely). In HCBS, the service variation is potentially much greater.

Policy debates about the relative merits of HCBS and NH have typically been presented as if NH and HCBS were simple, consistent concepts, and the populations served were homogeneous. However, the underlying empirical evidence requires closer examination. Specifically, comparisons between these modes of care are made more complicated by the considerable diversity in HCBS. The nature of HCBS clients and services can vary widely. Rarely do the analyses present complete descriptions of either. AL likewise varies widely in clientele and services. Analyses of the effectiveness of these modalities must describe both the clientele and the services in enough detail to provide necessary context and to allow for judging applicability.

Additionally, most studies did not report whether participants received any informal care or examine how informal care may have affected the type, frequency, or intensity of formal care services or how it may have interacted with outcomes. None of the studies examined recipients' or caregivers' experiences of care.

Most studies do not adequately address the problems of selection bias or attrition. Further, the settings where individuals receive LTC may not reflect personal choice or fit but instead be largely a function payment systems and policy environments. Issues of selection, case mix, and attrition complicate attempts to make indirect comparisons of the effects on trajectories. If not adequately addressed by the study design, these issues also impede direct comparisons.

Given the variation in clientele and the low likelihood of random assignment to HCBS versus NH, selection bias will continue to be a major concern, as it was with the studies reviewed here. Typical efforts to adjust for this bias, such as propensity scores, may be impeded by the large degree of heterogeneity, which reduces the accuracy of predictive equations. Multivariate analyses would encounter similar problems. Strong candidates for instrumental variables will be hard to identify.

One promising alternative may be to use “ideal types” as subgroups.74 This approach uses frontier analysis to identify groups with specific clusters of key patient characteristics that pertain to a specific health issue. Patient subgroups identified in this manner are sometimes called “archetypes.” Frontier analysis techniques may be useful in addressing how variation in processes of care delivery—such as forms of HCBS or NH—combine with variation in patient characteristics to produce targeted health outcomes. Ideal types have been used in addressing diabetes care and could be applied here.74

Distinguishing between long-standing and new LTC users is difficult with most data sources since individuals may not be followed across different providers or settings over time. Few studies used an admission cohort. Outcomes may differ for newcomers to the LTC system compared with those who have lived with limitations or received services for a long period of time. However, studies infrequently distinguished between the two. Related to this, followup times for tracing outcomes of interest were often short. Outcomes may not continue on the same trajectory over longer periods of time, so it is necessary to follow individuals over many years.

Most studies were not explicit about the composition of the people included within each group. For instance, most studies specified neither inclusion nor exclusion of individuals receiving postacute care (short-stay). Postacute care is provided to individuals upon discharge from a hospital; it is intended to aid their recuperation and rehabilitation. Postacute care typically lasts less than 30 days. In contrast, LTC is integrated into a person's life over an extended period of time. These groups may experience different outcomes and have different characteristics and preferences. Thus, investigators must be explicit about who is included or excluded.

The variation in outcome measures and data collection used across studies made it difficult to compare results. Any comparison of outcome trajectories must take into account the differences in care environments, including which measurements are used and who collects the data. NH policies and environments restrict residents' activities. For example, few, if any, NH residents are allowed to bathe by themselves. Thus, NH residents' activities of daily living (ADL) measures reflect less ability for self-care.

Another concern related to ceiling or floor effects is that HCBS recipients and NH residents may have different starting points for their respective trajectories. Given their greater level of disability at the outset, NH residents may have less opportunity to decline and may show little change in the typical measures used. We must also acknowledge the potential for right censoring, whereby the higher mortality rate in NHs may remove the most disabled residents. We could not determine whether these important issues affected the results of any studies in this review. The review of costs was limited given the insufficient evidence from published studies. The supporting evidence from grey literature sources suggests that cost comparisons are typically incomplete and do not include many relevant sources, including other public program expenditures, individual expenditures, and family burden.


Applicability is limited. Many of the samples were not generalizable because they were often small or limited by geography or because they were drawn from specific programs that do not apply to populations outside the study.

Implications for Policy Decisionmaking

Given the generally weak literature to assess the relative effectiveness of HCBS and NHs, policy decisions will likely continue to be made on the basis of preferences and beliefs, largely at the policy level and perhaps at the consumer level. The paradigm for comparing HCBS and NHs has shifted over time. Initially, HCBS was seen as a potentially less expensive alternative to NHs. Increasingly, however, HCBS is viewed as a preferable care modality that enhances clients' quality of life. NH environments and living settings are frequently very restrictive, and few would now be surprised that quality of life was higher for those in HCBS.

HCBS and even NH services are in a state of rapid transition. However, competing trends may simply result in a continuation of the variability we have noted. One trend in HCBS is toward more flexible “consumer directed” services provided by self-employed “independent” providers. These developments may allow HCBS to offer more care at more flexible times. A slow but growing trend in NH services is toward either smaller settings in households within NHs or small self-contained houses licensed as NHs. Privately occupied rooms are becoming more available, as well. If these trends continue, we can hypothesize better quality-of-life outcomes for NHs than we would presently expect. The ever-shifting patterns of LTC settings make the careful measurement of the interventions even more important.

Research Gaps

The weakness of the literature stands in sharp contrast to the importance of the topic. Many stakeholders want to know about the relative effectiveness of alternative modes of LTC. As budgets tighten and as demographically driven demand increases, states and other entities are seeking more efficient ways to deliver LTC. Better research is needed.

Table 30 summarizes several issues for future research. Existing literature is difficult to review due to heterogeneity of services, settings, populations, and measures. Therefore, designs for future research should be prospective cohort studies, preferably with direct and comparable measurement and assessment in both settings.

Table 30. Summary of research gaps to address.

Table 30

Summary of research gaps to address.

Future research needs are extensive. They reflect both methodological issues and substantive clinical and policy questions, such as the tradeoffs individuals face in choosing one LTC setting over another. Addressing these issues will enable better analyses and will help consumers and policymakers make informed, evidence-based choices. Here we highlight important areas of future research, organized by Key Question (KQ), with specific needs for each.

KQ 1a. Similarity of Populations Served by HCBS and NHs

Studies that compare HCBS and NHs need to tease out differences in the intervention. Terms such as NH care, HCBS, and AL do not suffice. Studies must carefully describe the actual settings and services. Because of the considerable within-setting variation (in all settings, but especially HCBS), the descriptors of type, frequency, and intensity of service will vary with individual client characteristics. This necessitates larger sample sizes to allow for subgroup and interaction analyses. Additionally, studies must report whether participants received any informal care in order to increase understanding of how informal care affects the type, frequency, or intensity of formal care services and how it interacts with outcomes.

As AL continues to evolve, we may become more able to distinguish AL facilities that are essentially institutional in nature from those that are closer to community care. Refining such distinctions may mean that not all AL now counted as HCBS will be so counted in the future. For both AL and NH, we need to be able to discern the detailed nature of the setting in terms of privacy, autonomy, and independence-enhancing amenities. Reporting general characteristics of the setting does not suffice, because some residents in both NH and AL have no access to the features that render the setting less institutional. Cutler et al. developed a method of assessing the environment as it is actually experienced by each resident in an NH.75 Such methods, although superior to global ratings of an NH environment or even an NH unit, are expensive to implement.

Studies should be explicit about the populations served in each setting. Postacute care delivered in NHs and home settings has increased sharply over the last several decades. Outcome trajectories will likely be affected based on the composition of the populations served in each setting. Despite the difficulty in doing so, studies should, when describing their populations, distinguish between individuals receiving short-term postacute care and long-term users.

NH residents are more disabled than HCBS recipients across certain domains. However, the measures used to characterize NH and HCBS populations may not always be comparable and the method of ascertainment often varies by setting and program. Disability measures in NH and HCBS environments should be similar. Measures should not be keyed to the supposed limitations of functioning in NHs in a misguided effort to provide “a level playing field.” For example, NH residents may have a lower level of disability on the bathing item because of availability of bathing equipment or help with bathing, and the solution may be to avoid that comparison. Similarly, instrumental activities of daily living (IADLs) are typically omitted in head-to-head comparisons of NH and HCBS. Presumably this is because investigators assume that NH residents will have no opportunity to prepare food, spend money, take medications, and the like, making it impossible to meaningfully assess IADLs in the NH. However, omitting ADLs such as bathing and most IADLs forces the comparable measures to conform to the low expectations about nursing homes, thereby suppressing the potential benefits of HCBS.

KQ 1b. Outcomes Among Those Served by HCBS Versus NHs

Given the differences in case mix, comparing outcome trajectories of older adults across HCBS and NHs presents numerous analytical challenges. Randomization can increase the likelihood that HCBS and NH populations are comparable on measured as well as unmeasured factors. However, randomization to HCBS versus NH or even to AL versus NH is precluded for practical and ethical reasons. Few persons with a strong preference for community care would accept an NH placement. In addition, many who believe they need NH care have already decided to leave the community.

Therefore, studies need to use statistical techniques to adjust for selection to ensure that the comparisons between groups are unbiased. As noted previously, these techniques include multivariate analysis, propensity scores, instrumental variables, or frontier analysis. Factors considered for adjustment should go beyond demographics, physical status, and cognitive status to include LTC attitudes held by both clinicians and consumers and availability of informal support. Studies should adjust for case mix and account for changes in the composition of HCBS and NH populations that result from deaths or transfers to different settings. In some cases, HCBS recipients may transfer to NHs. The impact of such transfers should be analyzed and interpreted.

Outcome measures should define and measure individual experiences of care provided in NHs and through HCBS across multiple domains, including quality of life and social functioning. Outcomes such as functioning and safety are obviously important. However, quality of life and social functioning often play a large role when individuals decide between care settings. Therefore, these outcomes should also be examined.

Resident report is necessary when using certain outcomes, including psychological well-being, social well-being, pain, and satisfaction, among others. Arguably, we may need more sensitive measures of outcomes. Attention should be paid to particular problems, including accurate measurement of outcomes for persons with dementia who cannot self-report, choice of a reasonable proxy, and measures of the quality of dying and quality of life immediately before death. Care should be taken to use independent data collectors and identical instruments.

Studies of benefits tend not to examine constructs of increasing policy significance. One example of such a construct is community integration (meaning integration with a wider community than one's own home, AL, or NH). Community integration is an important policy goal under the Olmstead decision, which mandated access to community care whenever feasible. Measures of social support and family functioning are also weak. This is in part because of the tendency to rely on a single family member as the informant. Experience of care should be measured from both individual and family perspectives. Informal caregivers often play a large role in the lives of individual LTC recipients. Understanding how informal caregivers' needs are addressed is critical.

The characteristics of providers within each setting, as well across different services, may also impact the outcomes of HCBS and NH recipients. Future research should address how provider characteristics influence the outcomes of individuals receiving care through HCBS and in NHs.

Subgroup analyses, defined by such characteristics as diagnoses, race/ethnicity, and socio-economic indicators, could indicate whether outcomes for HCBS and NH recipients differ by groups. For instance, a growing use of NHs has been observed among minority populations.76 We need more understanding about whether outcomes differ for these individuals based on the setting of care.

Ultimately, the choice between NH and HCBS involves competing goals and necessary tradeoffs (for example, between independence and safety). Comparing NHs and HCBS along multiple domains will provide LTC users with information that helps them to better evaluate NH or HCBS settings based on their needs, preferences, and values.

KQ 1c. Harms From HCBS Versus NHs

Studies of comparative harms must take into account psychological and social harms as well as physical and functional harms. For NH care (and perhaps some AL environments), hypothetical risks include loss of identity, helplessness, and depression. For HCBS, risks might include isolation, loneliness, anxiety about needing help, and depression. Yet, these outcomes are rarely looked at as harms. Indeed, investigators sometimes use depressive symptoms to risk-adjust away the impact of negative self-report when in fact the setting may cause the depressed symptoms. A distinction must be made between endogenous and situational depression. Studies should examine which settings better address psychological and social harms for individuals with similar impairments.

Also, studies will need to tease out harms attributable to treatment, such as over-, under- or inappropriate medication. For example, cognitive function becomes a reasonable outcome to examine if one considers the possibility that treatments and environments increase confusion.

KQ 1. Research Needs Summary

  • HCBS and NH service packages need careful description.
  • Studies should be explicit about the populations included in each setting, such as whether any distinctions are made between short-stay and long-stay individuals in these settings.
  • Studies should use comparable measures to characterize NH and HCBS populations. Differences in disability in NH and HCBS settings should be measured in a way that allows for examining a full range of ADL and IADL outcomes in both settings rather than suppressing outcomes to match the assumed limitations in NHs. The measures should allow for meaningful comparisons. For example, IADLs are typically not assessed in NHs and hence comparisons with HCBS are not possible.
  • The timing of measurement—the point in LTC trajectory at which the populations are compared across settings—is as important as the type of measure used.
  • Longitudinal studies should adjust for selection bias using statistical techniques such as multivariate analysis, propensity score methods, or instrumental variables. Studies should adjust for attrition as a result of resident deaths or transfers to other settings.
  • Studies should follow admission cohorts through various stages of LTC service use and evaluate transfers between settings.
  • Benefits and harms need to be measured fully. Studies should account for psychological and social benefits and harms as well as physical and functional harms. Quality of life, social functioning, community integration, experience of care, and satisfaction are important domains to examine.
  • Subgroup analyses should capture major groups defined by diagnosis, functional capacity, socioeconomic indicators, and access to care.

KQ 2. Costs of HCBS Versus NHs

Costs of LTC are typically measured in terms of Medicaid expenditures. However, for two reasons, these expenditures represent only a portion of the total expenditures. First, LTC expenditures can be seen in both Medicare and private insurance (e.g., MediGap). Second, expenditures by public and private programs ignore several other sources of direct or indirect spending.

Thus, from a societal perspective, costs of LTC should include expenditures borne by (1) other public programs such as Medicare and (2) individuals and their families. This will enable policymakers to make true comparisons. It may also help them avoid shifting expenditures from one program to another, between state and federal funds, and from formal to informal care. True cost comparisons between HCBS and NHs should account for direct LTC expenditures and all other relevant expenditures on both sides of the HCBS/NH ledger including:

  • Expenditures on room and board (included in expenditures for NH but not usually for HCBS recipients).
  • Expenditures for other public subsidy programs such as housing subsidies and aging services through Older Americans Act (OAA) or Title XX (available to HCBS recipients but not to NH residents).
  • Costs of informal care (financed by family members or provided as in-kind services) are often substantial and should be considered. This is especially true if family contributions and care negatively affect the caregiver. Family contributions occur in both HCBS and NHs, although they are likely greater in HCBS. Measuring and monetizing these contributions are both very difficult analytic steps, however. In addition, investigators may overestimate costs of family care if they rely on accounts of what family members do. This is because many of the activities of family members both in HCBS and NHs fall into the realm of familial relationship rather than care per se (e.g., visiting relatives, accompanying them on outings, and supplying food). Further, some contributions of family or informal caregivers are discretionary (e.g., laundering clothes for an NH resident in the family home to improve quality of laundering). Live-in relatives present an even more difficult conceptual problem when estimating the cost of family care in HCBS.
  • Expenditures for hospital and emergency care and primary medical care (often covered by Medicare). Medicaid or individuals may also pay premiums, deductibles, and co-pays for these services, which are relevant expenditures.

Private pay clients should be further studied. Most of our information comparing expenditures for HCBS recipients and NH residents comes from the Medicaid population.

Notably, policymakers are also concerned about the aggregate costs to Medicaid. These aggregate costs are the product of both unit costs and volume. Additional concerns are raised by the “woodwork effect,” whereby the availability of more desirable HCBS may prompt some people to use the service who would otherwise forgo LTC in an institution. Some experts speculate that the “woodwork effect” could induce demand for LTC and ultimately raise expenditures. To the extent that HCBS induces use of LTC, the total Medicaid program expenditures could be higher even though the per-recipient expenditures are lower. However, the extent of this phenomenon has been a source of debate and is still being examined.77-80

KQ 2. Research Needs Summary

  • Studies should compare direct LTC expenditures between NH and HCBS and other relevant expenditures. Costs should be viewed from a societal perspective and should include expenditures from public programs such as Medicaid and Medicare, as well as individual expenditures and family burden.
  • Studies of expenditures for non-Medicaid LTC recipients are needed.

Current or Ongoing Studies

A study by a University of Pennsylvania team, headed by Mary Naylor, is currently underway. The study follows a cohort of older adults who recently began to receive LTC. The cohort is followed over 2 years. The final sample includes 468 English- and Spanish-speaking older adults (156 from nursing homes, 156 from assisted living facilities, and 156 from their homes), Data collection on the last few older adults enrolled in the study will be completed in summer 2012. Quarterly interviews are conducted with the study participants to elicit data regarding changes in multiple dimensions of health and quality of life. The data are supplemented by medical record data (e.g., use of health resources) and survey data (e.g., major organizational changes). In addition to a number of presentations, a paper on the conceptual model used to guide this effort is under review and an analysis of the first year of data is nearly complete.


The question of how the delivery of LTC through HCBS compared with NHs affects outcome trajectories of older adults is difficult to resolve based on limited evidence and the methodological limitations of studies reviewed. More and better research is needed to draw robust conclusions about how the setting of care delivery influences outcomes and costs for older adults using LTC.


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (984K)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...