|dc.description.abstract||Background: Home and community care is a critical part of an effective health care system. For many clients and families, home and community care services provide the necessary supports so they can manage various short- and long-term needs effectively and safely in their homes. In Ontario, personal support and homemaking (PS/HM) services account for three-quarters of all publicly funded home care services. PS/HM services assist clients with basic self-care and other tasks known as Activities of Daily Living and Instrumental Activities of Daily Living. Yet the processes for determining eligibility, priority, and allocation of publicly funded PS/HM services are neither consistent between Local Health Integration Networks (LHINs) nor accessible to clients and families. Client outcomes attributable to PS/HM service provision are also poorly understood. The overarching goal of this thesis is to develop and refine decision support tools to guide the allocation of publicly funded PS/HM services, and to characterise the relationship between the quantity of publicly funded PS/HM services and outcomes.
Study 1: Across Canada, Ontario is the sole province that has implemented the interRAI Home Care (HC), interRAI Community Health Assessment (CHA), and interRAI Contact Assessment (CA). The HC and CHA are standardised comprehensive assessments developed to assess the needs, values, and preferences of adults receiving services in home and community-based settings. The CA although much briefer follows the same interRAI standard, allowing direct comparisons across the three populations. To date, there is little published evidence on Ontario’s CA- and CHA-assessed populations. This chapter comprised of four sub-chapters based on a single retrospective cohort of unique clients (age ≥18 years) newly admitted to Ontario’s publicly funded home care program between April 1, 2016 and March 31, 2017 and assessed with the CA or HC (n=268,667) and unique clients assessed with the CHA between April 1, 2015 and March 31, 2016 (n=15,307).
Sub-study A identified unique characteristics and service use patterns among Ontario’s public home and community care clients assessed with the CA, HC, and CHA. Sub-study B modelled the relationship between the Assessment Urgency Algorithm (AUA) and time to HC assessment using cumulative incidence competing risk and Kaplan-Meier methods. Higher AUA levels are strongly associated with greater likelihood of receiving an HC assessment and shorter time to HC assessment, although 26.6% of clients in the highest AUA level were not subsequently assessed. The AUA calculated from the CA at intake is also moderately positively correlated with the Method for Assigning Priority Levels (MAPLe) algorithm that is used to guide decisions related to eligibility and priority for services and long-term care placement following the HC assessment. Sub-study C investigated the agreement between the receipt of publicly funded PS/HM services after the CA and HC. Three multivariable logistic models were fit to identify predictors of clients receiving significantly more or less service after the HC. As expected, measures of need are most strongly associated with service plan adjustments although enabling characteristics, especially the LHIN in which a client lives, are also highly influential. Sub-study D compared the self-reported and billed services data over the same seven-day lookback period and found that formal PS/HM services accounted for a small fraction of the total help that most home and community care clients received.
Study 2: In 2018, Ontario’s LHINs formally adopted the Personal Support (PS) Algorithm as a standard approach to identify need for PS/HM services. The PS Algorithm classifies clients based on functional and cognitive impairment and other need characteristics known to be associated with need for PS/HM services. Recent publications have suggested additional characteristics (“modifiers”) that may be relevant. The study sample consisted of 126,001 unique HC assessments completed between April 1, 2016 and March 31, 2017 that is a representative sample of Ontario’s public long-stay home care client population. To test the relevance of additional modifiers to the PS Algorithm, the median publicly funded PS/HM hours and total (i.e., formal and informal) home support hours per month were compared across PS Algorithm groups and selected modifiers. The PS Algorithm explains 25.5% and 33.4% of the variance in publicly funded and total PS/HM hours, respectively. Clients living alone receive more publicly funded PS/HM hours, but clients living with their primary informal caregiver receive much more total home support hours. Publicly funded and total PS/HM hours increase with the severity of cognitive impairment and caregiver distress, but generally do not respond to health instability except for very high health instability. Finally, comparison of the distribution of publicly funded PS/HM hours between FY 13/14 and FY 16/17 suggests that allocations have begun to cluster as LHINs move away from local allocation practices toward a common provincial standard.
Study 3: While the PS Algorithm is helpful for guiding the allocation of PS/HM services for HC-assessed home care clients, there is no equivalent tool to guide the allocation of PS/HM services for short-stay clients and within short-term service plans for long-stay clients. The goal of this study is to create a conceptually similar algorithm based on the CA that differentiates need for PS/HM services. The derivation sample consisted of 228,354 unique CA assessments completed between April 1, 2016 and March 31, 2017. Among CA-assessed clients, 15.4% received any publicly funded PS/HM services after the CA. Given the zero-inflated nature of the dependent variable, bivariate logistic models predicting the odds of receiving any publicly funded PS/HM services were fit for the full derivation sample, and bivariate linear models predicting the amount of services were fit for clients receiving any publicly funded PS/HM services. Automatic and interactive decision trees were developed based on need characteristics identified in exploratory analyses. An out-of-time validation sample was used to assess each model’s explained variance of the amount of publicly funded PS/HM services received after the CA and weighted kappa of the PS Group at the time of HC assessment. Consistent with the derivation of the PS Algorithm, measures of functional impairment, cognitive impairment, and caregiver distress are strongly associated with the amount of PS/HM hours received after the CA. Similar performance statistics were observed across the candidate trees; thus, the model replicating the PS Algorithm was selected as the final algorithm (“PS Algorithm for the CA”). In the validation sample, the PS (CA) Algorithm explains 20.4% of the variance in publicly funded PS/HM hours and is moderately associated with the PS Group at the time of HC assessment (weighted kappa statistic=0.36). In comparison, the AUA only explains 11.6% of the variance in publicly funded PS/HM hours.
Study 4: Derivation of the PS Algorithm and PS (CA) Algorithm was based on the premise that the average historical allocation of PS/HM services is an indicator of need. While the relative differences in allocation can be reliably used to differentiate levels of need, there is concern that the status quo may not represent the “right” amount of services and therefore the average historical allocation should not serve as a benchmark for future allocation practices. To address this concern, a multi-state analytic approach was used to test the hypothesis that some level of service below a threshold would increase the risk of poor outcomes or some level of service above a threshold would decrease risk of poor outcomes. Data for this study was provided by the Hamilton Niagara Haldimand Brant LHIN. Clients referred on or after January 1, 2010 and subsequently admitted for home care services were eligible for the study. For each eligible client, all HC assessments completed on or after January 1, 2013 were retrieved. Each HC assessment up to December 2017 was assigned to one of three initial states based on the presence of caregiver distress. A period of up to 456 days (15 months) was allowed to observe a follow-up HC assessment or home care episode discharge. The sample consisted of 57,208 observation pairs representing 30,625 unique clients. The independent variable of interest was the quintile of publicly funded PS/HM services, where the reference group was the 3rd quintile that represents the median allocation within a given PS Group. Adjusting for baseline client characteristics, providing less than the median PS/HM services significantly increases the odds of new caregiver distress, moving to long-term care, and death. Among distressed caregivers, providing less than the median PS/HM services significantly decreases the adjusted odds of resolving caregiver distress. Among clients with non-distressed caregivers, providing more than the median PS/HM services significantly decreases the adjusted odds of moving to cluster residence.
Conclusions: This thesis sought to provide actionable evidence on the predictors and outcomes of publicly funded PS/HM service allocation in Ontario. It is the first comprehensive study of the CA since its province-wide adoption in 2010. The CA is part of an efficient assessment process that identifies clients who should be at the highest priority to receive the more comprehensive HC assessment. As well, information from the CA can be used in a structured way to guide the allocation of PS/HM services for short-stay clients as well as within short-term service plans for long-stay clients. Together, the PS (CA) Algorithm and PS Algorithm provide a unified evidence-informed approach for allocating publicly funded PS/HM services throughout the home care episode. To date, Ontario’s LHINs have adopted the PS Algorithm without the corresponding Framework of Hours for specifying hours of service. The final part of this thesis demonstrates that the Framework of Hours identifies minimum thresholds below which publicly funded PS/HM allocation may lead to poorer client and caregiver outcomes. The findings provide compelling evidence for policy-makers to set standard service guidelines and monitor PS/HM-sensitive outcomes. Doing so will ensure that clients and families know what supports to expect from the public home and community care system, that public resources are distributed fairly, that investments in home care can be demonstrated, and that the valuable contributions of personal support workers can be properly recognised.||en