In Canada, the United States, and internationally, home care is playing an increasingly prominent role in the health care system [1–5], with the aim of reducing costs related to other services, including acute hospitalization and nursing home placement. For example, the Commission on the Future of Home Care in Canada [6] recommended the extension of public funding to cover the cost of four types of home care: post-acute medical care, post-acute rehabilitation, community-based palliative care, and mental health services for clients with behavior-management needs. Although home care accounts only for a small fraction of total expenditures in the Canadian and US health care systems, it has been the sector with the most rapid relative rate of growth in costs [7, 8]. In the United States, state 'waiver' programs were introduced with the aim of reducing nursing home utilization among the frail elderly [9]. For example, Michigan's MI-Choice program uses a targeting approach to identify and intervene with those elderly persons at greatest risk of imminent institutionalization in order to prolong their stay in the community.

Despite the widespread belief that home care has the potential to reduce expenditure in other parts of the health care system, there is evidence to suggest that targeted home care probably holds the greatest promise for cost-effective service delivery [10–15]. For instance, if services are provided to a predominantly healthy population, it may be difficult to show any benefits related to home care given that the clients are relatively independent and are at low risk of transition to poor health [16]. On the other hand, if services are provided too late in the disablement process, it may be difficult to reduce the risk of institutionalization [17, 18]. The relevant risk factors may be less modifiable and/or the informal network may no longer be able or willing to continue given the level and duration of burden they have endured. Moreover, for persons who are most severely functionally and cognitively impaired, home care may not be a sustainable means of prolonged service delivery due to the demands on formal and informal resources.

Home care case managers are charged with making decisions about the allocation of community and institutional resources for the care of the frail elderly in the community. Their decisions are commonly made on the basis of comprehensive assessments; however, in many jurisdictions these assessments are not standardized, have not been tested for reliability and validity, and are not accompanied by decision-support algorithms that aid in interpretation of the assessment results. Therefore, one might argue that, despite the use of geriatric assessment instruments, the decisions of case managers are often subjectively based and inconsistent due to idiosyncratic differences among clinicians. Clinicians themselves are also often inconsistent over time. A consequence of this type of approach is that limited resources may be allocated ineffectively. There may be a mismatch between needs that are either undetected or discounted inappropriately and the services offered to home care clients. The cost of health care can be driven up by poor communication among health professionals, exacerbations of illness or disability, unnecessary admission to long-term care settings, and avoidable hospitalizations. It is particularly important to provide effective home care to persons at imminent risk of nursing home placement because, once admitted, the likelihood of return to the community declines substantially and the formal costs of care grow substantially.

In 2003, the province of Ontario began to open numerous long-term care facilities across the province. Approximately 15,000 new beds were added to a system of about 50,000 beds in response to perceptions that persons requiring long-term care were using excessive amounts of acute hospital resources. With the opening of these new facilities, the provincial government recognized that there was a need to employ a systematic method of admitting new residents into those beds in order to ensure that the available spaces were taken by persons with the highest level of need and to minimize the chance of premature institutionalization.

In 2002, Ontario mandated the implementation of the Resident Assessment Instrument – Home Care (RAI-HC) [3, 4, 19, 20] for use with all long-stay home care clients in the province. The implementation of the RAI-HC provided a new information source upon which decisions regarding nursing home placements and allocation of community-based services could be made. A review of the literature and consultation with the field was undertaken in order to identify decision-support systems that may aid in the allocation of community and institutional resources. Three main candidate methodologies were identified. First, Miller [21] developed a crosswalk from the Ontario eligibility criteria for long-term care to the RAI-HC. However, this study showed that, using the existing algorithm, approximately 80% of Ontario long-stay home care clients would be eligible for long-term care placement, making this algorithm relatively uninformative.

The INST-RISK system developed by Morris et al [22] is an indexing system that considers approximately 20 functional, medical, social, and psychological measures to develop a summary score of risk of institutionalization, which is then stratified into four risk levels. The Regina Risk Indicator Tool [23] appears to be a derivative of this earlier work in that it makes only minor modifications to the system previously developed by Morris et al. There are a number of limitations with the INST-RISK approach, including: (a) the use of a number of independent variables (such as gender, income) that appear to have little or no predictive power for institutionalization; (b) the application of identical weights to variables of differing levels of importance for institutionalization; and (c) compatibility concerns due to measurement inconsistencies between the INST-RISK items and the RAI-HC.

The third system considered was the MI-Choice algorithm [9, 24], which was designed specifically for use with the RAI-HC. MI-Choice was developed for the Michigan Waiver Program using a sample of elderly persons considered to be at imminent risk of institutionalization. Clinical vignettes were used to train case managers on state eligibility rules to standardize their approach to matching clinical need to corresponding community or institutional services. They completed RAI-HC assessments on about 800 home care clients and rated them according to the level of care they believed the client required, including information and referral, homemaking, personal care, professional home care services, and nursing home placement. RAI-HC items were then used to model case manager decisions, beginning with those identified as eligible for nursing home places and information and referral (that is, the extreme values) and ending with homemaking as a residual category. Fries and James [9] reported moderate levels of overall agreement (weighted kappa = 0.55) between the case managers' ratings and the levels of care suggested by the MI-Choice algorithm, independent of the availability of informal support (see [25] for a discussion of kappa values). In the MI-Choice waiver program, this algorithm identified approximately 8% of clients as eligible for nursing home placement; however, in preliminary analyses performed as part of the present study, it classified less than 2% of Ontario home care clients in that category. Therefore, in contrast to Ontario's existing eligibility criteria, the MI-Choice system would allow for very few clients to be considered for entry into long-term care homes. A second concern was that the algorithm appeared to be insensitive to clinical complexity among younger home care clients, who were not part of the original MI-Choice study.

Since no existing system was found to be appropriate for the Ontario context, an effort was launched to develop a new decision-support algorithm for allocating home care resources based on the RAI-HC. The Method for Assigning Priority Levels (MAPLe) was created to assist case managers in determining the relative priority that should be attached to a client regardless of whether he or she needs community or institutional services. Instead of attempting to match client care characteristics to specific venues or types of care, the MAPLe system prioritizes clients to identify those in most urgent need of services, irrespective of the care setting. The aim of this paper is to describe the development and validation efforts undertaken as part of the development of the MAPLe system.

MAPLe provides an empirically sound decision-support system that will allow case managers to make more systematic evaluations of the needs of clients and the urgency with which they should respond to those needs. Using three different outcomes, with validation results in six countries, MAPLe clearly differentiated the risk of adverse outcomes, including institutionalization. Case managers who have completed an RAI-HC assessment can obtain the MAPLe results automatically from software in which the algorithm is embedded and these results then provide a context against which person-specific service recommendations may be made. It should be noted, however, that the intent is not to use MAPLe as an automated decision-making system devoid of clinical judgment. Instead, case managers considering MAPLe scores should also engage in a full discussion with the client, family, and other formal service providers to develop person-specific recommendations that take into account the individual's strengths, preferences, and needs. For example, some clients who score low on the MAPLe algorithm may rate themselves as being in poor health, may be showing signs of depression, or may have had an overnight hospital stay or frequent emergency room visits. It would be inappropriate to treat such an individual as an 'information and referral only' client if the case manager believes these other clinical considerations to be of critical importance. On the other hand, individuals in the very high MAPLe category do not necessarily require immediate nursing home placement because, for example, they may have family members who are both willing and able to address their current level of need in the community. Similarly, alternative housing arrangements may provide these individuals with the appropriate resources to remain relatively independent with access to formal supports when needed.

While MAPLe may be used to make person-centered resource allocation decisions at the individual level, it may also be used as a monitoring system at the regional, organizational, national, and international levels to evaluate practice patterns. That is, one may stratify populations according to MAPLe levels and then compare the performance of home care agencies with respect to outcomes of care within MAPLe levels. Such a benchmarking system may be used to identify jurisdictions where MAPLe-adjusted nursing home admissions, for example, are higher than expected based on the experience of other regions. Similarly, MAPLe levels at intake can be used to examine regional variations in access to services by level of need.

The introduction of MAPLe into normal, daily use in home care can be expected to have an impact on the nature of services provided by agencies. That is, clients with lower MAPLe levels will be less likely to receive services than they may have been in the past, and clients with higher MAPLe levels will presumably be more likely to access those services because they are more readily identifiable. That being said, it is not necessarily true that subgroups of home care clients need to be excluded from all services based on low MAPLe scores. The choice of who receives services and what type of services are provided remains a value-based decision to be made by policymakers and clinicians in each jurisdiction. In that regard, the cross-national results for MAPLe shown in Table 3 do not imply that some countries are providing services correctly while other countries erroneously target light-care clients. The main benefit of implementing MAPLe would be that persons with higher levels of need should be at a relatively higher level of priority for access to services than those with lower-level needs. That does not preclude the possibility of persons at the lowest level receiving appropriate services.

The present study made some important decisions regarding the role of informal support in the development of the prioritization system for community and institutional services. The choice to include measures of the informal support system (that is, caregiver stress) as a dependent variable rather than as an independent variable was important for methodological, clinical, and policy-related reasons. From a methodological perspective, the use of measures of informal support as independent variables would create a situation where those variables become highly vulnerable to systematic response bias in order to gain access to desired services when used in normal clinical practice. That is, if family members believe that saying they could do more to help a client would place their relative in a lower priority level, there would be a strong disincentive to making such claims. Therefore, any items reflecting the capacity of the informal network to provide care could become less valid after the implementation of such a system. The introduction of this type of bias will make it more difficult to differentiate between families who are showing clear signs of distress and those who have continued capacity to provide more services. From a policy perspective, one must be concerned about the potential unintended consequences of a steering effect that would result in family members being less likely to provide informal care. The capacity of the home care system to function adequately is heavily dependent on the provision of informal care by family members. Extreme caution should be taken with the introduction of policy measures that may provide disincentives to families to be involved. At the same time, distress among caregivers should be a concern to home care service providers and policy makers.

Although the present study showed that the MAPLe system is related to both the formal costs of service provision and hours of informal care, MAPLe is not intended to be an alternative to case-mix systems (for example, as an alternative to the RUG-III/HC [38] system). That being said, the introduction of MAPLe will have important implications for case-mix distributions in both home care and in long-term care facilities. As MAPLe begins to be used, one might expect the relative resource intensity of home care clients to increase because greater priority will be given to clients with higher levels of need. Similarly, to the extent that MAPLe is used as a decision-support system for nursing home placement, it is likely to result in a shifting of the distribution of new admissions to long-term care settings towards those community-based clients with the highest levels of need. Therefore, the introduction of MAPLe may have the effect of increasing case-mix or resource intensity in both home care and nursing home settings.

This study also demonstrates the benefits of using a grouping methodology, as with decision-tree analysis, compared with indexing systems that simply assign a linear increase in scores for any combination of variables. Unlike systems such as the RRIT or INST-RISK, MAPLe is able to consider the impact of specific combinations of variables (for example, cognitive impairment combined with functional impairment).

There was a remarkable level of cross-jurisdictional consistency in the MAPLe results. Although the slope of the relationship between the MAPLe scores and the dependent variables of interest were not always identical between jurisdictions, the relative increments in risk between MAPLe levels was very consistent in all of the countries examined. This is particularly surprising given the value differences underlying the provision of community-based services in these jurisdictions. Japan and Italy, for example, tend to be traditional in their expectation that family members will provide the bulk of care. It is therefore not surprising to see lower absolute levels of ratings that clients would be rated by themselves or by others as being better off elsewhere in these jurisdictions, compared with the Canadian results. On the other hand, in all jurisdictions, there was a rather consistent increase in the proportion of clients rated as being better off elsewhere with each increment in the MAPLe level. It would be useful to conduct future research in these other countries to determine the extent to which MAPLe is indeed predictive of actual institutionalization in those settings. In addition, it is important to recognize that clients, family members, and home care professionals may not always share the same view about the most appropriate care setting. These potentially different perspectives should be taken into account when determining what services would be most appropriate for a given client.

An important next step for research with the MAPLe system is to conduct intervention studies to determine the extent to which the level of risk associated with a given MAPLe category can be reduced. For example, could additional measures be put in place that would have the result of yielding a lower than expected institutionalization rate within a MAPLe category, or could the risk factors that have placed the client in that MAPLe category be reversed? For instance, falling is a major factor contributing to higher MAPLe scores. If a falls intervention was put in place that actually prevented future falls, conceivably, the individual's MAPLe scores would decline over time.

It will also be important to conduct research into the extent to which MAPLe is relevant to home care outcomes other than caregiver distress or nursing home placement. For example, it is not necessarily the case that clients with higher MAPLe scores feel more socially isolated or lonely or that they would benefit from interventions such as friendly visiting programs or transportation services. Therefore, it would be helpful to consider what comprises the full spectrum of benefits that could be realized through high-quality home care. This might be done through qualitative studies of home care clients, caregivers, case managers, and administrators. Once these benefits are identified, further research should be conducted on the effectiveness of MAPLe as a targeting system for persons at risk of adverse outcomes in other domain areas.

MAPLe also has potential benefits for the support of cross-national research. The problem that population-level information cannot be equated across jurisdictions is an important limitation of any home care research that is not based on individual-level observations. Indeed, results in Table 1 show that it would be a faulty assumption to believe that home care clients in Italy are comparable with home care clients in Ontario, for example. However, by using the MAPLe system to stratify samples with individual-level data, one could be more confident in the equivalence of samples being compared across two or more jurisdictions.

The MAPLe algorithm provides an empirically based decision support tool that may be used to inform choices related to the allocation of home care resources and prioritization of clients needing community or facility-based services. MAPLe is a valid predictor of nursing home placements, caregiver distress and ratings that the client would be better off elsewhere, and it has been shown to perform well in a variety of international jurisdictions. MAPLe may be used at the individual level to support clinical decision-making, but it may also be used with aggregated data to inform policy development and planning.