Pushed by aging populations and the related rise of chronic health and social needs, many jurisdictions, nationally and internationally, are now embracing the idea that it is socially and economically desirable for older persons to “age in place”. In addition to enhancing the well-being, quality of life, and independence of older persons themselves, care “closer to home” is seen to offer the potential to moderate demand for costly, and sometimes inappropriate, hospital and institutional care (Donner, Reference Donner2015; Mestheneos, Reference Mestheneos2011; Sinha, Reference Sinha2013; Walker, Reference Walker2011).
In Ontario, a recent high-profile expert panel concluded that if the province aims to provide “the right care, at the right time, in the right place” (Donner, Reference Donner2015), it should plan for more care to be delivered in the home where growing numbers of older persons, and their families, now manage increasingly complex needs, over longer periods of time (Donner, Reference Donner2015). In response, the province announced new multimillion-dollar investments in home and community care aimed at helping frail older persons remain in their own homes longer, while avoiding or delaying emergency department visits, hospital admissions, and placement in residential long-term care (LTC) homes (Ministry of Health and Long-Term Care [MOHLTC], 2015b).
Nevertheless, it seems unlikely that home care will be a workable option for all older persons, and that some will require placement in residential LTC. Even in “gold standard” countries such as Denmark, where comprehensive, integrated community-based care for older persons is widely accessible, and where the government has built no LTC beds since the late 1980s (Schulz, Reference Schulz2014), a minority of older persons continue to live in residential care settings (Organization for Economic Cooperation and Development [OECD], 2011).
This raises the question, Who requires residential LTC and who doesn’t? Conventional wisdom suggests that the needs of the care recipient are the sole or main driver of LTC placement; as people age and lose functional capacity, the likelihood of residential placement increases apace. However, as international comparative analysis suggests, needs alone do not account for variation in rates of residential care across the industrialized nations; such variation also reflects differences in the design and comprehensiveness of formal care systems, as well as access to informal care provided by mostly unpaid family, friends, and neighbours (OECD, 2011).
The impact of supply-side factors is well documented. With respect to formal community-based care, some observers in Ontario have commented that although older persons overwhelmingly wish to age at home, hospitalization and referral to LTC can occur by “default” when community-based care options are inadequate (Donner, Reference Donner2015; Drummond, Reference Drummond2012; Ontario Seniors’ Secretariat, 2013; Sinha, Reference Sinha2012; Walker, Reference Walker2011). Such arguments align with research findings from countries such as the United Kingdom, which suggest that the need for residential LTC beds is in large part determined by access to appropriate, cost-effective community-based care (Challis & Hughes, Reference Challis and Hughes2002; Challis, Hughes, McNiven, Stewart, & Darton, Reference Challis, Hughes, McNiven, Stewart and Darton1999; Clarkson, Hughes, & Challis, Reference Clarkson, Hughes and Challis2005; Hughes & Challis, Reference Hughes and Challis2004; Tucker, Hughes, Burns, & Challis, Reference Tucker, Hughes, Burns and Challis2008). They align also with recent research results from Ontario, which suggest that, particularly in rural and remote areas of the province with sparse community-based service infrastructures, older persons are more likely to be referred to residential LTC even at comparatively low levels of need (Bolin, Phillips, & Hawes, Reference Bolin, Phillips and Hawes2006; Kuluski, Williams, Berta, & Laporte, Reference Kuluski, Williams, Berta and Laporte2012; Kuluski, Williams, Laporte, & Berta, Reference Kuluski, Williams, Berta and Laporte2012; Williams, Kuluski, & Watkins, Reference Williams, Watkins and Kuluski2010; Williams, Challis, et al., Reference Williams, Challis, Deber, Watkins, Kuluski, Lum and Daub2009; Williams, Kuluski, et al., Reference Williams, Challis, Deber, Watkins, Kuluski, Lum and Daub2009). Lacking access to needed community-based care, LTC can thus become a costly, and potentially avoidable, “upward substitution”.
With respect to informal care, a growing weight of international evidence documents that family, friends, and neighbors are the “first line” of care in the community; they provide the bulk of the everyday emotional, instrumental, and personal supports required by care recipients who cannot manage on their own (Averett, Sikora, & Argys, Reference Averett, Sikora, Argys and Redmount2014; OECD, 2011). In addition, the scientific literature has investigated the impact of informal caregiving on the probability that a care recipient is placed in an LTC facility, concluding that access, or lack of access, to such caregiving plays an important role (Gaugler, Duval, Anderson, & Kane, Reference Gaugler, Duval, Anderson and Kane2007; Gaugler, Yu, Krichbaum, & Wyman, Reference Gaugler, Yu, Krichbaum and Wyman2009; Miller & Weissert, Reference Miller and Weissert2000). This literature also notes that the burden on informal caregivers can be substantial, both psychologically and physically (McKinlay, Crawford, & Tennstedt, Reference McKinlay, Crawford and Tennstedt1995), often leading to caregiver burnout and withdrawal, with referral to LTC as the only viable option.
The evidence regarding the relationship between formal and informal care, particularly as to whether informal care acts as a substitute or complement to formal care, is more nuanced. Some evidence suggests that for “heavier care” activities of daily living (ADLs) such as bathing and personal hygiene, informal and formal care may complement each other (Katz & Stroud, Reference Katz and Stroud1989) since informal caregivers are less likely to provide such care. In contrast, informal care appears more likely to substitute for formal care over a range of instrumental activities of daily living (IADLs), which include routine but crucial tasks such as managing finances and medications, phone use, and transportation. We note that the ability to perform such IADL tasks has been observed to affect the likelihood of referral to LTC in different regions of Ontario (Kuluski, Williams, Berta, et al., Reference Kuluski, Williams, Berta and Laporte2012; Williams, Challis, et al., Reference Williams, Challis, Deber, Watkins, Kuluski, Lum and Daub2009). Insofar as informal caregivers provide a good portion of IADL-related care, understanding their ability to do so is critically important for policy-makers.
A 2012 report by Health Quality Ontario found that 20 per cent of Community Care Access Centre home care clients who were placed in LTC could have stayed in their homes or been placed elsewhere in the community (Health Quality Ontario, 2012). Another report in 2011 found that 37 per cent of clients waiting in hospital alternate-level-of-care beds had needs no more urgent than those clients being cared for at home (Walker, Reference Walker2011). Thus, to gain a better understanding of the determinants of LTC placement in Ontario, our study looked at the factors associated with LTC wait-list placement in North West Ontario. We focused on referred individuals rather than those who had already been admitted to LTC – the rationale for this focus was that in some sense those who were on the wait list for LTC represented the “at risk” population. Although some clearly required the higher intensity of care provided in an institutional setting based on their health needs, a non-trivial proportion of the at-risk population could potentially have continued to be cared for in the community if the right supports became available.
In our study described in this article, we used a multivariable logit regression modeling approach to analyse data documenting the characteristics and needs of care recipients waiting for LTC placement in North West Ontario. Our aim was to contribute to the existing literature in a number of ways. First, we investigated factors that affect the likelihood of wait-list placement for institutional LTC; to our knowledge, this study is the first to do so. Second, we tested the hypothesis that in addition to the assessed care needs of the care recipient, LTC wait list placement is simultaneously impacted by supply-side factors such as access to formal community-based care and informal caregiving. Although a number of studies have examined these factors individually and a few in combination with regard to LTC admission, they have not been examined in combination in terms of their impact on LTC wait-list placement. We anticipate that where formal and informal care is more readily available, older persons are more likely to be able to age in place even at relatively high levels of assessed need; conversely, where community-based formal and informal care are less accessible, the needs threshold or “tipping point” for institutionalization will be lower, and older persons will be more likely to be referred to the wait list for LTC (Kuluski, Williams, Berta, et al., Reference Kuluski, Williams, Berta and Laporte2012; Williams, Challis, et al., Reference Williams, Challis, Deber, Watkins, Kuluski, Lum and Daub2009). The third aim of our study was that we explored the impact of IADLs and ADLs in disaggregated form to determine if there were particular IADLs or ADLs that are more important determinants of LTC wait-list placement.
Background Literature in Brief
Much of the literature has focused on factors affecting the likelihood that a care recipient is admitted to an LTC facility. The focus of our study, in contrast, was on the factors affecting the likelihood for institutional LTC wait-list placement. Here, the interest was to examine the at-risk population. It would not be unreasonable to expect that the factors that have been found to significantly affect LTC admission may also affect the likelihood of wait-list placement for LTC. Consequently, we provide here a brief overview of the factors that have been investigated and shown to have an important impact on LTC admission. For reasons of parsimony, we mainly limit our literature review to studies that modeled the likelihood of LTC admission rather than time to placement. Although the two are related, factors associated with time to admission may not necessarily be the same as factors related to the probability of an admission occurring (Gaugler et al., Reference Gaugler, Duval, Anderson and Kane2007). To the best of our knowledge, our study is the first to have looked at another related aspect – the wait-listed population rather than care recipients who have already been admitted to LTC, the difference being that the at-risk population may potentially be able to delay LTC admission if appropriate supports become available.
Studies examining risk factors for residential LTC began emerging in the literature more than three decades ago (Miller & Weissert, Reference Miller and Weissert2000). These studies typically focused on care recipient characteristics such as advanced age, limitations in ADLs and IADLs, cognitive deficits, health status, and living alone as risk factors for institutionalization (Andel, Hyer, & Slack, Reference Andel, Hyer and Slack2007; Branch & Jette, Reference Branch and Jette1982; Luppa et al., 2009; Luppa, Luck, Matschinger, König, & Riedel-Heller, Reference Luppa, Luck, Matschinger, König and Riedel-Heller2010; McFall & Miller, Reference McFall and Miller1992; Tomiak, Berthelot, Guimond, & Mustard, Reference Tomiak, Berthelot, Guimond and Mustard2000). More recent studies have continued this focus – for instance, in examining care recipient characteristics, such as cognition, leading to increased dependency on LTC in different jurisdictions (Boyd, Bowman, Broad, & Connolly, Reference Boyd, Bowman, Broad and Connolly2012). It is worth noting that a common approach in these studies is to aggregate numbers of ADL and IADL limitations into summary scores, rather than examining the impact of particular limitations; for example, a U.S.-based meta-analysis indicated that a combination of three or more ADL and cognitive deficits were strong predictors of nursing home admission (Gaugler et al., Reference Gaugler, Duval, Anderson and Kane2007).
The impact of informal caregiving has also been extensively investigated with the presence of an informal caregiver generally found to be a significant factor in reducing the risk of institutionalization (Boaz & Muller, Reference Boaz and Muller1994), even after controlling for the acuity of clients and whether or not formal services are received (Charles & Sevak, Reference Charles and Sevak2005). Other studies have demonstrated that type of caregiver (e.g., spouse vs. adult child) impacts risk of placement (Pearlman & Crown, Reference Pearlman and Crown1992; Kim, Cho, & Lee, Reference Kim, Cho and Lee2013). However, although some studies have shown that co-residence of the caregiver, irrespective of relationship to the care recipient, is an important consideration regarding risk of placement (Jette, Tennstedt, & Crawford, Reference Jette, Tennstedt and Crawford1995), other studies have failed to do so (McFall & Miller, Reference McFall and Miller1992).
The volume and mix of informal care provided, rather than the simple presence of a caregiver, have also been shown to affect risk of institutionalization. Pearlman and Crown (Reference Pearlman and Crown1992) made the point that measures such as living arrangements are really an indication of “potential” caregiving resources as opposed to actual care received; they found that informal care provided 5 or more days per week had a significant impact on reducing risk of admission. Van Houtven and Norton (Reference Van Houtven and Norton2004) reported that a 10 per cent increase in informal care hours led to a .77 percentage point decrease in the likelihood of admission.
Related to the number of hours and the roles performed is the issue of caregiver burden. McFall and Miller (Reference McFall and Miller1992) investigated the likelihood that an elderly person with a spouse or adult child caregiver would enter a nursing home, controlling for caregiver burden, where burden was measured in relation to 10 different problems associated with caregiving (such as “taking care of him/her is hard on me emotionally”). Risk for LTC increased commensurately with the number of problems identified by informal caregivers. Other studies have linked caregiver stress (Spillman & Long, Reference Spillman and Long2009) and “negative personal impact” (McKinlay et al., Reference McKinlay, Crawford and Tennstedt1995) to greater likelihood of nursing home placement. A literature review of studies concerned with the predictors of nursing home admission for people with dementia underscored the importance of caregiver stress as an independent driver of institutionalization (Gaugler et al., Reference Gaugler, Yu, Krichbaum and Wyman2009).
System-level factors can also play an important role in the use of residential care. For example, variations in the risk of LTC admission across regions independent of the age-sex and morbidity distribution of populations may occur resulting from the degree of rurality and differences in the availability of health care resources (Connolly & O’Reilly, Reference Connolly and O’Reilly2009; Keating & Phillips, Reference Keating, Phillips and Keating2008). Although rural areas tend to have fewer community care options than more urbanized communities (Chan, Hart, & Goodman, Reference Chan, Hart and Goodman2006), they also tend to have more LTC beds per elderly population (Rosenthal & Fox, Reference Rosenthal and Fox2000). Formal home and community care service capacity may be limited in rural and remote areas because of distance (Sims-Gould, Martin-Matthews, & Keating, Reference Sims-Gould, Martin Matthews, Keating and Keating2008), low population, and a lack of service infrastructure, resulting in a higher risk of institutionalization (Bolin et al., Reference Bolin, Phillips and Hawes2006; Coward, Netzer, & Mullens, Reference Coward, Netzer and Mullens1996; Foley et al., Reference Foley, Ostfeld, Branch, Wallace, McGloin and Cornoni-Huntley1992). In their investigation of LTC wait lists in North West Ontario (the same region we examined in our study), researchers Kuluski, Williams, Berta, et al. (2012) and Kuluski, Williams, Laporte, et al. (2012) concluded that individuals on LTC wait lists in rural areas experienced lower levels of physical and cognitive impairment than their counterparts in the region’s main urban area.
Of course, accessing needed community-based care may prove challenging even within urban areas. For example, a study conducted in Toronto, Ontario’s largest city, estimated that between a third and a half of the care recipients waiting for LTC could be “diverted” safely and cost-effectively to care in the family home or in supported housing (Williams, Challis, et al., Reference Williams, Challis, Deber, Watkins, Kuluski, Lum and Daub2009) if adequate resources were available (which was not the case). Lacking access to such resources, care recipients instead received LTC, which effectively served as an “upward substitution” for community-based care (Williams, Challis, et al., Reference Williams, Challis, Deber, Watkins, Kuluski, Lum and Daub2009).
The “balance” of formal care and informal care in relation to the risk of institutionalization has been investigated to a lesser extent than care recipient needs, and with mixed results. Some findings suggest that the amount of informal care is to some degree conditional on the amount and type of formal home care services provided (Spillman & Long, Reference Spillman and Long2009; Stabile, Laporte, & Coyte, Reference Stabile, Laporte and Coyte2006). Branch and Jette (Reference Branch and Jette1982) did not find any effect from the use of seven different services (ranging from meals to specialist care) on the risk of admission controlling for the health status and level of disability of the care recipient. Newman, Struyk, Wright, and Rice (Reference Newman, Struyk, Wright and Rice1990) found receipt of formal care to be related to increased risk of nursing home entry but noted that in their data the measurement of formal service receipt could have been up to two years before the observed date of admission. Jette et al. (Reference Jette, Tennstedt and Crawford1995) found that the increased risk of admission resulting from receipt of formal services was reversed for care recipients with cognitive impairment. Controlling for health status, living arrangements, and caregiver burden, McFall and Miller (Reference McFall and Miller1992) found that receipt of formal paid help or help from an organization was positive and significant in their model for probability of nursing home entry.
Less evident in the literature are studies that simultaneously examine care recipient needs, as well as access to formal and informal care. One exception is a study by Spillman and Long (Reference Spillman and Long2009), which examined needs, caregiver stress, hours of informal and formal care, and regional (system) characteristics; when all variables were controlled, nursing home beds per 1,000 persons emerged as the only significant predictor of nursing home admission.
Most studies in Canada have focused on individual factors in modeling the likelihood of LTC admission. For example, St. John, Montgomery, Kristjansson, & McDowell (2002) found that higher cognitive scores reduced the probability of institutionalization after adjusting for age, gender, education, marital status, ADL impairment, IADL impairment, and self-reported health. Hirdes, Poss, and Curtin-Telegdi (Reference Hirdes, Poss and Curtin-Telegdi2008) developed an index of individual characteristics called the Method for Assigning Priority Levels (MAPLe), which was found to be a strong predictor of nursing home admission, caregiver burnout, and the perception that the care recipient would be better off in another institution (by the individual and/or caregiver). One study that investigated the effects of individual, caregiver, and system-level characteristics jointly was that of Maxwell et al. (Reference Maxwell, Soo, Hogan, Wodchis, Gilbart, Amuah and … Strain2013), which looked at a sample of residents in assisted living in Alberta. Their caregiver level measure was hours of informal care from friends or family, which was not significant in their final model. They also looked at time to LTC placement rather than LTC wait-list placement. Our study aimed to contribute to the body of literature in Canada and abroad by looking at care recipient needs simultaneously with informal caregiver and formal system characteristics as determinants of the likelihood that an individual is placed on a wait list for institutional LTC.
Ontario Context
In Ontario, the bulk of publicly funded home and community care services are delivered through 14 Community Care Access Centres (CCACs), whose boundaries correspond to those of the provincial health regions or Local Health Integration Networks (LHINs). CCACs assess needs and consequently contract services such as nursing, social work, rehabilitation, and personal care from for-profit and not-for-profit providers on a competitive basis (MOHLTC, 2015a).
CCACs do not charge user fees, minimizing the likelihood that care recipients will experience financial barriers to access. Nevertheless, in Ontario, there is no guarantee that even eligible care recipients will receive home care services. Although Ontario residents have the right to an assessment using a standardized assessment tool (Resident Assessment Instrument – Home Care [RAI-HC]), there is no legislated entitlement; access will vary depending on a range of supply-side factors including service availability, service ceilings, and policy priorities.
For example, even when an adequate supply of services is available at the local level, which is not always the case particularly outside of urban areas, provincial service ceilings stipulate that with some exceptions, a care recipient may not be provided with more than 28 visits from a registered nurse or registered practical nurse in a 7-day period (MOHLTC, 2006). Moreover, if the services are provided by a registered nurse, then no more than 43 hours of service may be provided; however, up to 53 hours can be provided by a registered practical nurse and no more than 48 hours if both registered nurses and registered practical nurses provide care (MOHLTC, 2006) in that 7-day period. Similar limits are imposed for other services such as rehabilitation and home making.
Needs “thresholds” for home care services are rising as CCACs respond to provincial policy priorities that concentrate available resources on care recipients at imminent risk of LTC placement (aimed at reducing lengthy waits for residential LTC beds). Additionally, CCACs focus resources on hospital patients requiring post-discharge care (aimed at reducing high numbers of alternate level of care beds occupied by care recipients who no longer require hospital care but cannot be discharged because of a lack of community-based care options). As a result, “high needs” care recipients now account for an increasingly large share of the home care “pie”, with “lower needs” care recipients (including many requiring supports for IADLs) referred to mostly smaller-scale, local, not-for-profit, volunteer-driven community support services agencies, which vary widely in terms of access, service capacity, eligibility, and user fees (Donner, Reference Donner2015).
CCACs also control queues for LTC beds. In 2014, Ontario funded and regulated 78,120 LTC beds in a total of 627 municipal homes for the aged, charitable homes, and nursing homes (both for-profit and not-for-profit). However, there were significant wait lists; in May 2014, more than 20,000 seniors were awaiting placement in LTC, with an average wait time of about three months (89 days) (Ontario Association of Non-Profit Homes and Services for Seniors [OANHSS], 2015).
Provincial policy frameworks on Ontario rely heavily on summary indices of assessed needs to direct older persons towards care settings such as home care, assisted living, and LTC (MOHLTC, 2011). For example, a care recipient with an RAI-HC score of 11 or more is considered eligible for LTC placement (Auditor General of Ontario, 2012). Care recipients with a score of 8–10 require review from a senior manager, who generally makes the decision on the basis of caregiver burden. Care recipients with a score of 7 or less are usually not considered eligible. Beyond client needs and caregiver burden, CCACs are also required to review community-based alternatives before deeming a care recipient eligible for LTC (Auditor General of Ontario, 2012).
Once assessed as eligible for an LTC bed, the care recipient or their legal representative must select their top choices of residential care facility. When a suitable bed becomes available in a selected facility, the care recipient may elect to take the bed, or defer, albeit at the risk of losing their place in the queue. If an LTC placement is required urgently (e.g., because of a medical issue), a “crisis” placement may be made to another facility in the region with the possibility of later transfer to a facility of choice.
Like CCAC services, LTC beds are publicly funded through the LHIN (although LTC residents are responsible for a co-payment of about $57/day to cover “hotel” costs that may be reduced or waived based on ability to pay); this mitigates selection issues that may arise in multiple payer contexts. Moreover, because CCACs queue access to home and community care as well as to LTC, standardized assessment data are available for all care recipients assessed and deemed eligible for home care, including those placed on an LTC wait list (OANHSS, 2015).
North West LHIN
Given extensive variation in population needs, and in the supply of home care services across the province, we modeled home care assessment data from one region, the North West LHIN (NW LHIN). Although the NW LHIN is geographically Ontario’s largest region, accounting for 47 per cent of Ontario’s landmass, it has a relatively small population, with only 2 per cent of the provincial total (North West Local Health Integration Network [NW LHIN], 2013). Thunder Bay is the main city centre, with 55 per cent of the regional population (NW LHIN, 2013). The rest of the population is sparsely spread throughout rural and remote areas offering scope to test the impact of urban/rural differences in our analysis. The population of the NW LHIN is also one of the oldest in the province (due in large part to an outflow of younger persons seeking education and jobs in more southern regions), with a lower-than-average socioeconomic profile and a higher-than-average prevalence of poor health behaviours (NW LHIN, 2013). Hence, although we had data only on this one LHIN, the results of the analysis should be applicable to regions that are similar in rurality and population demographics. Moreover, the general methodological approach – which simultaneously investigates care recipient, informal caregiver, and system-level characteristics – can be applied across regions.
Data
For our study, we obtained assessment data from the North West CCAC for all care recipients aged 55 years and older who received publicly funded home care services in 2014 (n = 3,387). These data were derived from the RAI-HC, which, as noted, is a standardized and validated assessment tool used by CCACs across the province to determine eligibility for home care and LTC placement (Canadian Institute for Health Information [CIHI], 2002; InterRAI, Reference InterRAI2015). We removed 876 care recipients from the data set since they were not living in the community at the time of assessment (including those assessed in a hospital, retirement home, LTC facility, or assisted living residence); this yielded a sample of 2,511 care recipients living in private residences. Of these 2,511 care recipients, 16 resided in a region that could not be determined, and one care recipient was coded as “out of the region” at the time of the assessment. Because we could not determine the location of these individuals, they were removed from the analysis. In addition to the 16 respondents whose region could not be determined, 24 care recipients did not report the amount of informal hours of care they received in the 7 days prior to their assessment. Of these 24, 8 did not report formal hours of care and 1 additional case did not report nursing hours. We therefore removed 42 observations because of missing values, bringing the total sample from 2,511 to 2,469. Ethics approval for this study was obtained from University of Toronto Ethics Review Board on March 28, 2014.
Methods
Dependent Variable
We modeled the likelihood that a care recipient would be placed on the LTC wait list; the binary-dependent variable assumed a value of 1 if a care recipient was wait-listed and 0 otherwise. As the assessment is standardized and administered by trained case managers across the region, concerns regarding large variation in LTC wait list placements from different referral sources should be minimal (Kuluski, Williams, Laporte, et al., Reference Kuluski, Williams, Berta and Laporte2012).
Explanatory Variables
We included three “blocks” of explanatory variables measuring care recipient characteristics; informal caregiver characteristics; and formal system characteristics, which are all derived from the RAI-HC. We did this to quantify the effect of informal and formal factors on the likelihood of LTC wait-list placement while controlling for the individual needs of the care recipient. This is in contrast to previous studies in Canada, as well as other countries internationally, which have tended to focus solely on the role of individual factors on the likelihood of LTC admission.
Care Recipient Characteristics
We included multiple measures of care recipient needs in our analysis. Difficulty with ADLs was captured using the ADL hierarchy scale, which captures the recipient’s dependence in 4 areas: personal hygiene, toilet use, locomotion, and eating (Morris, Fries, & Morris, Reference Morris, Fries and Morris1999). The original ADL scale scores ranged from 0 to 6. For the purposes of this analysis, we recoded scores into three categories: low difficulty, medium difficulty, and high difficulty for ADL scores of 0, 1–2, and 3–6, respectively, since very few care recipients in the data had an ADL score greater than 3. Analysis with ADLs aggregated, as well as disaggregated, into its component parts (personal hygiene, toilet use, locomotion, and eating) were used in the regression in separate analyses (see Logit Regression Results section). We also included variables relating to IADLs, which measure the care recipient’s ability to perform meal preparation, ordinary housework, managing finances, managing medications, phone use, shopping, and transportation. Each of these was coded categorically as no difficulty, some difficulty, and great difficulty respectively.
Cognition was captured using three of the individual items that make up the Cognitive Performance Scale (CPS). Footnote 1 These include cognitive skills for daily decision making, short-term memory, and “making self understood” (Morris et al., Reference Morris, Fries, Mehr, Hawes, Phillips, Mor and Lipsitz1994). Cognitive skills for daily decision making was measured on a 5-point scale, indicating that the care recipient had no impairment or modified independence, and was minimally impaired, moderately impaired, or severely impaired. We also included an indicator for short-term memory problems, as well as an indicator for problems in making self understood. The latter was coded as a binary variable since very few care recipients had serious problems with this measure (i.e., beyond “usually understood”, see the Appendix for a detailed definition for each variable used in the analyses). Other measures of care recipient characteristics included in the analysis were reliance on an interpreter, unsteady walk, fear of falling outside, and poor health (self-assessed and/or diagnosed), as well as age and sex.
Informal Caregiver Characteristics
We included information about the informal caregivers in relation to the care recipient: relationship (e.g., spouse yes/no), whether or not they co-resided, and their roles in providing ADL and IADL support. The total amount of informal care received by the care recipient in the past 7 days of the assessment date was included as a continuous predictor. We also included information about the status of the caregiver, such as whether or not the caregiver was unable to continue in caring activities or expressed feelings of distress, anger, or depression. We constructed an indicator labeled “caregiver distress” that took on a value of 1 if the caregiver experienced either of the above circumstances, and 0 otherwise. The preferences on the part of the care recipient and informal caregiver as to whether the care recipient would be better off receiving care outside the home was coded categorically, taking on values of 1, 2, and 3 if the care recipient, caregiver, or both care recipient and caregiver thought that care would better be delivered outside of the home setting, respectively, and 0 otherwise.
Formal System Characteristics
The Rurality Index of Ontario (RIO) was the tool we used to capture the “medical” rurality of the local area in which the care recipient resided. The RIO score is determined almost exclusively by distance and travel times and comprises three main components: (1) community population and community density; (2) travel time to the nearest basic referral site; and (3) travel time to the nearest advanced referral site. Rural areas are less densely populated than urban areas, and so family doctors may have to travel longer distances to reach their patients. Rural residents may also have to travel longer distances to receive care, which may be further impeded by a lack of public transportation. Higher RIO scores, indicating higher degrees of rurality, are thus interpreted as indicating more limited access to health care resources in a region. Footnote 2 Each care recipient was assigned a RIO score based on which community planning area (CPA) they resided in, which is based on their postal code. CPA at the census subdivision level was provided by the NW LHIN. RIO scores less than 40 were classified as urban areas, and areas with RIO scores greater than or equal to 40 were classified as rural (MOHLTC, 2012). Footnote 3 This index has been validated for use in Ontario and has been used previously in research related to home care in Ontario (Laporte, Croxford, & Coyte, Reference Laporte, Croxford and Coyte2007).
The amount of formal care the care recipient received in the 7 days preceding their assessment was also included. This included hours from the following services: home health aides, visiting nurses, homemaking services, meals, volunteer services, physical therapy, occupational therapy, day care or day hospital, and social workers. We excluded speech therapy as less than 1 per cent of the sample (on and off the LTC wait list) was receiving this type of service. Each of these was coded with a value of 1 if the care recipient had received the service, and 0 otherwise. We did not include the supply of LTC beds in each census subdivision in the analysis as this was highly correlated with the RIO. We chose to use the RIO over LTC bed supply as it represented access to medical resources and acted as a proxy measure for the number of beds in a given region. Moreover, the relevant LTC bed supply could not be that in the immediate area in which the care recipient resided – it could, for example, be the bed availability in the area in which their children lived that mattered. Unfortunately, we did not observe in the data the location or other characteristics of the facilities that the recipient had selected or that had been identified by the LHIN staff for their placement.
Analysis
We employed a logit multivariable modeling approach. The three blocks of explanatory variables were entered into the regression in four separate models to assess their relative impact using backward induction. Each model is reported in Table 4. Model 1 includes the block of explanatory variables measuring care recipient characteristics. Model 2 adds cognition to the care recipient characteristics in Model 1. Model 3 adds informal caregiver characteristics to the variables in Model 2. Model 4 adds formal system characteristics to the care recipient and informal caregiver blocks from Models 1, 2, and 3. We explored a range of interactions of the variables in the models (i.e., between rurality and informal caregiver-level variables, as well as rurality and other formal system-level variables), but none were significant at the 5 per cent level. As a result, none of the 4 models include interaction terms.
A Ramsey’s Regression Specification Error Test (RESET) was performed for each of the models in the analysis, whereby squared fitted values were included as explanatory variables in an auxiliary regression of the dependent and independent variables. A finding of significance of the squared fitted values would suggest misspecification of the model or omitted variables (Studenmund, Reference Studenmund2006).
For each block of explanatory variables, we calculated a likelihood ratio statistic to test if the additional variables introduced in each block were jointly equal to zero.
In addition to the block likelihood ratio tests, we utilized commonality analysis to evaluate the contribution of each block of explanatory variables to the total R 2 in the regression. Commonality analysis is a method of decomposing the variance explained by each of the co-variates in a multiple regression framework (Ray-Mukherjee et al., Reference Ray-Mukherjee, Nimon, Mukherjee, Morris, Slotow and Hamer2014; Seibold & McPhee, Reference Seibold and McPhee1979). The R 2 was decomposed into unique and common effects, whereby the unique effects indicated how much variance was accounted for by a single co-variate, and common effects indicated how much variance was common to a set of co-variates. Rather than calculating a unique score for each co-variate, we calculated the unique contribution of each block of explanatory variables (care recipient, informal caregiver, and formal system level) to the overall R 2 in the final model. This was appropriate in our case since we were interested in the contribution of blocks of explanatory variables in addition to the contribution of the variables within the blocks.
In terms of the IADLs, as already noted, we chose not to aggregate them into one composite measure but to instead include each of them into the regressions. This brought about concerns of multi-collinearity, as the 7 IADL items were correlated not only with each other, but also with the cognition variables (see supplementary Appendix for correlation matrix of explanatory variables with IADLs). We therefore calculated an uncentred variance inflation factor (VIF) in the final model. A VIF shows the increase in the variance of a coefficient estimate attributable to the fact that this variable is not orthogonal to the other variables in the model (Greene, Reference Greene2012). Various cutoffs have been suggested in the literature, with 10 being the most common (O’Brien, Reference O’Brien2007). However, dropping variables simply due to a high VIF raises concerns about the potential for omitted variable bias. Having great difficulty with housework, meal preparation, and shopping produced a VIF in our analysis greater than 10, suggesting that the effects of these IADLs could be captured by the other variables (i.e., the other IADLs or cognition) in the analysis.
We observed no change in the signs or significance of the other coefficients when these variables were dropped in the final model (when all co-variates, including cognition, were controlled for), suggesting that they were redundant in the analysis. Since these were the only IADLs with a high VIF, these particular IADL variables were dropped from the analyses. In terms of the other variables in our regression, “number of medications taken in the last 7 days” also produced a high VIF. This variable was not significant in any of the models and produced a model with a higher Bayesian information criterion (BIC) score in the final model. As a result, we did not include this variable in the descriptive and regression analyses. Once this variable was dropped, the remaining independent variables had a VIF less than 10, suggesting that the potential for harmful multicollinearity was low.
Results
Care Recipient Characteristics
Of the 2,469 community-dwelling care recipients receiving publicly funded home care services in the NW region, 367 were on the wait list for institutional LTC. Table 1 provides summary statistics of the characteristics of these care recipients. We observed a greater proportion of care recipients who were listed as experiencing “medium difficulty” and “high difficulty” in ADLs on the wait list for LTC. For all IADLs, we observed that the majority of people who experienced “no difficulty” and “some difficulty” were not on the wait list for LTC, and people who experienced “great difficulty” were on the wait list for LTC. The exception to this was “phone use”, where more people in both the “some difficulty” and “great difficulty” categories were on the wait list for LTC. This difference seemed to support our approach of looking at each IADL individually rather than aggregating them into a composite index.
Table 1: Care recipient characteristics
Chi-squared tests used for categorical/ordinal variables. t-tests used for continuous variables.
*, **, and *** denote significance at the 5%, 1%, and 0.1% levels respectively.
We observed a greater proportion of older care recipients on the wait list for LTC. Sex did not differ significantly across wait list and non–wait list categories in our sample. Smoking and tobacco use were less prevalent among those who were wait listed. Those on the LTC wait list were more likely to be in poorer health across all the measured dimensions. We observed a greater proportion of wait-listed care recipients with an unsteady gait, as well as having a fear of falling outside. Not surprisingly, we observed a large proportion of people who were fully intact in making daily decisions in the non–wait list group. We also observed a greater proportion of care recipients with short-term memory problems and difficulty making themselves understood in the wait-list category.
Informal Caregiver Characteristics
Table 2 presents descriptive results for informal caregiver characteristics. We found that 97–98 per cent of the sample had a primary caregiver for wait-listed and non-wait-listed care recipients. We observed no variation in living status, as roughly the same proportion of wait-listed care recipients had a live-in caregiver, as well as a caregiver that was their spouse. A high proportion of caregivers indicating that they were experiencing significant burden were caring for people on the wait list (51%); however, 26 per cent of caregivers for those not on the list also experienced significant burden. A significantly larger proportion of those who were on the wait list received assistance with ADLs from their caregivers. In terms of the attitudes of the care recipients and caregivers, we observed that for 26 per cent of cases on the wait list, both the care recipient and caregiver felt that the care recipient would be better off receiving care in an environment other than the home setting. In an additional 47 per cent of cases on the wait list, both the care recipient and caregiver felt that the care recipient would not be better cared for in a setting other than the home. Interestingly, a rather higher proportion of those on the waiting list (23%) corresponded to cases in which the caregiver alone felt that it would be better if care were provided outside the home setting, compared to 3 per cent for the care recipient alone. Care recipients on the wait list were receiving significantly more informal hours of care (during the week and on the weekend, as well as overall).
Table 2: Caregiver-level characteristics
Chi-squared tests used for categorical variables. t-tests used for continuous variables.
SD = standard deviation
*,**, and *** denote significance at the 5%, 1%, and 0.1% levels respectively.
Formal System Characteristics
The data in Table 3 describe formal system characteristics. Rurality varied significantly, with a higher proportion of care recipients on the wait list in rural areas, recalling that our measure of rurality captures the availability of medical care resources. Formal services varied significantly across the categories for most of the services. A greater proportion of care recipients on the LTC wait list received support from home health aides, homemaking services, meals (Meals on Wheels), volunteer services, and social worker services. Interestingly, a significantly larger proportion of care recipients not on the wait list were receiving nursing care and occupational therapy. In addition, a significantly greater proportion of care recipients were drawing on services from a day care centre or day hospital. This program provides social, recreational, medical, and/or functional support to the client, and also provides respite for primary caregivers (CIHI, 2002).
Table 3: Formal system characteristics
Chi-squared tests used for categorical variables.
*,**, and *** denote significance at the 5%, 1%, and 0.1% levels respectively.
Meals: Prepared meals delivered to client for immediate or later consumption (i.e., Meals on Wheels).
Day care or day hospital: Program (out of home) where client receives social, recreational, medical, or function support. Also provides respite for primary care givers. See CIHI (2002) for a more detailed explanation.
Logit Regression Results
Table 4 presents the results of the regression analysis in the form of odds ratios. These results suggest that the odds of LTC wait-list placement were greater for those with higher levels of ADL impairment in comparison to those who have no difficulty in these activities. This relationship persists across all 4 models, with “high difficulty” becoming significant in Model 3. In a separate analysis with ADLs disaggregated into its components (personal hygiene, toilet use, locomotion in home, and eating) the only variable out of the four that remained statistically significant was hygiene. The variables as a group, however, were jointly significant (p < .01). For parsimony, and because the elements of the measure were jointly significant, we report aggregated ADL results with the disaggregated results available upon request. In terms of IADLs, care recipients with “some” or “great difficulty” in phone use, and “great difficulty” in managing finances and transportation were at higher odds of wait-list placement compared to those having “no difficulty”. Having “some” or “great difficulty” in managing medications was initially significant, but lost its effect when we introduced the cognition variables.
Table 4: Logit regression results
Odds ratios (OR) from logistic analysis.
Model 1 includes care recipient factors only. Model 2 adds measures of cognition to the care recipient factors in Model 1. Model 3 adds informal caregiver variables to the care recipient and cognition variables in Models 1 and 2. Model 4 adds formal system characteristics to the care recipient and informal caregiver variables in Models 1, 2, and 3.
*,**, and *** denote significance at the 5%, 1%, and 0.1% levels respectively.
LR = likelihood ratio
Care recipients having any form of impairment with cognitive skills for daily decision making were at greater odds of LTC wait-list placement in comparison to those who were independent, and this effect increased as the level of severity increased. As for age, those aged 65 years or older were at greater odds of LTC wait-list placement (in reference to those 55–64), at a relatively increasing rate (4–9 times greater in Model 4), but tapered off in the age 85 and older range in Model 5. Sex was not significant. Those requiring an interpreter were at lower odds of LTC wait-list placement (p < .05).
Regarding caregiver characteristics, a care recipient whose caregiver was distressed was at significantly greater odds of being placed on the LTC wait list. In addition, if the care recipient, caregiver, or both felt that the care recipient would be better off in another environment, then the care recipient was at much greater odds of LTC wait-list placement (between 3–5 times in Model 4) than a person who was perceived to be okay in the community (i.e., needs were being met in the community). Each additional hour of informal care was associated with a 1 per cent decrease in the odds of LTC wait-list placement, which was statistically significant.
As for formal system characteristics, care recipients who received support from health care aides were at greater odds of LTC wait-list placement (p < .01), as well as Meals on Wheels or day care or day hospital services (p < .05). Other formal care services were not statistically significant at the 5 per cent level.
As seen from the results for Model 4, rurality was a highly significant predictor of wait list placement. Care recipients in rural areas were 2.6 times more likely to be on the LTC wait list than their urban counterparts, all else being equal.
The significant likelihood ratio p-value (p < .001) for care recipient factors (in models 1 and 2) suggests that a model with care recipient characteristics is a better fit compared to a model with a constant alone (i.e., where the coefficients on the explanatory variables in models 1 and 2 are restricted to be 0). The significant p-value (p < .001) for Model 3 suggests that the addition of the informal caregiver characteristics provides a better fit than a model with care recipient characteristics alone. Similarly, the significant p-value (p < .001) for Model 4 suggests that the addition of formal system characteristics creates a model with a better fit than a model with care recipient characteristics and informal caregiver characteristics alone.
The results of the commonality analysis from the full model are presented in Table 5 (common effects omitted). The results suggest that in Model 4 (i.e., with all co-variates controlled for), care recipient characteristics uniquely accounted for 35 per cent of the total variation in LTC wait-list placement, followed by informal caregiver characteristics (22%) and formal system characteristics (13%).
Table 5: Regression pseudo R Footnote 2 decomposition
The RESET test (the test statistic on the squared fitted values) was not significant in any of the models at the 0.1 per cent level, suggesting no evidence of misspecification.
Discussion
This study represents a first attempt in the LTC placement literature to quantify the impact of individual, caregiver, and formal system factors in predicting the risk of LTC wait-list placement.
An important finding relating to care recipient characteristics is that the ability to perform everyday IADLs such as telephone use, managing one’s finances, or travelling beyond walking distance, remained significant determinants of wait-list placement even when all other variables were taken into account. These IADLs lose magnitude once cognition is controlled for, but all except medications management remain significant in Model 4. The loss of magnitude may be due to an omitted variable bias problem. Difficulty with IADLs is positively correlated with cognitive impairment, as we would expect a person to have difficulty with these tasks if they are cognitively disabled. Moreover, cognitive impairment increases the risk of LTC wait-list placement. We would therefore expect to see positive bias on the IADL coefficients without cognition as a control.
Although not considered health care per se, these activities clearly affect the likelihood that older persons can remain at home. Paradoxically, IADL supports have been among the first to be cut in Ontario, as home care resources have increasingly been diverted towards more intensive post-acute care (Canadian Healthcare Association, 2009; Donner, Reference Donner2015; Hermus, Stonebridge, Thériault, & Bounajm, 2012). New communication technologies and greater access to homelike care settings such as supportive housing offering 24/7 surveillance and support may provide additional scope to address IADL needs (Lum, Williams, Sladek, & Ying, Reference Lum, Williams, Sladek and Ying2010). Our analysis underlines the importance of addressing these simple everyday activities in the community to avoid residential LTC.
The results also confirm the key role of informal caregivers in delaying or avoiding LTC entry: hours of care provided turned out to be particularly significant even after controlling for the health status of the care recipient and formal services received. Evidence that a significant proportion of caregivers feel that they are over-burdened raises concerns about the long-run sustainability of informal care provision, especially given the shift in population demographics now occurring across developed countries towards greater numbers of older people with more complex care needs requiring more hours of care and fewer younger people to provide care.
Even though the data fail to show that co-residence is significant when other factors are controlled, they emphasize that caregiver burden matters: caregivers who said they were unable to continue in caring activities and/or who expressed feelings of distress, anger, or depression increased the odds that a care recipient was placed on the LTC wait list by 66 per cent (CI: 22–126%) even after controlling for the health status of the care recipient and formal services received. Potentially complicating this picture is evidence from countries such as the United Kingdom that the number of older persons requiring care will soon outstrip the number of family members available to provide care and that just to keep pace, current caregivers will have to increase their efforts by more than half over the next 15 years (McNeil & Hunter, Reference McNeil and Hunter2014).
Moreover, we found that when informal caregivers felt that the care recipient would be better off cared for outside the home, care recipients had a significant likelihood to end up on the wait list. In other words, regardless of assessed needs, or formal care provided, when caregivers alone or in combination with care recipients determined that community dwelling is no longer a workable option, the likelihood of referral to LTC increases significantly. Some caregivers, given the particular needs of the care recipient, the quality and appropriateness of formal home and community care accessible at the local level, and their own willingness and capacity to provide care, may reach the limit of caring sooner than others; when they do, our analysis suggests that the likelihood of referral to LTC is multiplied. This adds weight to recent expert reports in Ontario which have concluded that the presence of caregivers “is the reason why so many older Ontarians have been – and will remain – able to age in their places of choice for as long as possible” (Sinha, Reference Sinha2012, p. 21), and that without their continued contributions, the health care system would not be sustainable (Donner, Reference Donner2015). Nevertheless, in Ontario, home care does not recognize caregivers as “care recipients”. Thus, there are few dedicated caregiver support programs (Peckham, Williams, & Neysmith, Reference Peckham, Williams and Neysmith2014).
System factors were also significantly associated with the odds of wait-list placement. Confirming what has been observed in the literature and what we had observed in an earlier project, rural location was associated with greater odds of being on the LTC wait list than was urban location (Kuluski, Williams, Berta, et al., Reference Kuluski, Williams, Berta and Laporte2012).
One of the more surprising results was the lack of significance of types of home care services such as physical therapy; nevertheless, these services were significant as a group, suggesting that access to home care does affect the likelihood of aging in place. Footnote 4 Even more noteworthy was the positive and persistent impact of community support services – such as day care/day hospital programs, home health aides, and Meals on Wheels – on the likelihood of wait-list placement even when all other variables were controlled in the analysis. The positive and significant impact of day care or day hospital may indicate a client with a deteriorating health status or a caregiver approaching distress status, as these services provide support to care recipients and respite for caregivers. The positive effect of home health aides and Meals on Wheels is less clear – although it is worth noting that the data are cross-sectional (one point in time). Thus, it could be the case that these clients had been drawing on these services over a long period of time in order to keep them in the community. Access to longitudinal data would allow for a more detailed exploration of this observed association.
In sum, our results suggest the important interplay between the assessed needs of the care recipient, and access to informal and formal community-based care in determining whether older persons can continue to age in place, or be referred to residential LTC. Although care recipient needs account for a significant proportion of the variance in LTC wait-list placement, the explanatory power effectively doubles when informal and formal care are considered. Indeed, the results show that supply-side factors (e.g., access to community-based formal and informal care) predict as much of the variance in the likelihood of being wait listed, as do care recipient needs.
Besides providing a more complete and nuanced understanding of the drivers of loss of independence, these findings caution against mechanistic assumptions that population aging will require a proportionate increase in the supply of residential LTC beds. As other observers have noted (Walker, Reference Walker2011), recourse to institutional care is more likely where community-based care options are inadequate. They also caution against strategies that ignore support for non-medical IADLs such as telephone use, managing finances, and transportation, or that fail to consider the needs of informal caregivers who provide the bulk of the everyday care required to maintain the well-being and independence of older persons living in the community. The regression results underline that regardless of the care recipient’s level of need, when an informal caregiver decides alone, or in combination with the care recipient, that aging in place is no longer an option, the likelihood of referral to LTC rises significantly.
Two final points are worth considering. The first concerns the use of summary indices of care recipient needs, such as those commonly used in Ontario to direct care recipients to different care settings such as LTC, home care, or supportive housing. Although high scores on such indices can act as a warning, in much the same way as a warning light in an automobile, they do little to diagnose actual problems that, as we have seen, could derive from different sources including an inability to perform simple everyday activities of daily living – the IADLs. Moreover, these indices typically exclude information about informal caregivers, even though their willingness and ability to provide needed care appears crucial in determining whether older persons at all levels of need can continue to live at home.
The second point to consider concerns the importance of deriving better measures of system capacity. Although, as we have seen, available measures such as the RIO and hours of formal home care have a significant impact, they include little information about the range of community-based options that may be available, or not, at the local level. We believe that such measures are essential to assess the relative effectiveness of services and supports such as transportation and respite, which may assist care recipients as well as informal caregivers to stay at home, thus mitigating the need for LTC.
Strengths and Limitations
The use of the RAI-HC is a strength of our study in that it contains a large array of variables that describe the care recipient’s physical and mental status at the time of assessment. Moreover, the RAI-HC contains information on care recipient/caregiver attitudes as to whether the care recipient would be better off in another care setting. This variable is generally lacking in other data sets that are used to model the likelihood of LTC admission, and was shown to be a strong predictor of LTC wait-list placement in our sample. The RAI-HC also includes information about the nature of informal support provided to the care recipient, which includes the amount of informal care provided in the days prior to assessment, as well as the distress status of the informal caregiver. It also includes information about the level and range of formal home care services provided with an ability (by way of location information) to link to the RIO, which provides a sense of the regional resources on offer that may affect wait-list placement rates. Moreover, trained case managers conduct the assessment across the NW LHIN region (Kuluski, Williams, Laporte, et al., Reference Kuluski, Williams, Berta and Laporte2012), so variations in assessment data from different case managers should be minimal. In essence, the data set allows for the exploration of factors along these three dimensions simultaneously to a degree that has been done infrequently in the Canadian context and to a very small degree in the extant literature.
Nonetheless, we note a number of important limitations. First, although the RAI-HC is rich in care recipient–level characteristics, some detailed information on caregivers such as caregiver health status, age, and sex are not collected. This lack of data is an important limitation since there may be variation in the tasks caregivers are willing and able to perform depending on their age, sex, and health status, which could contribute to the likelihood of LTC wait-list placement for their care recipients. Moreover, our study did not look directly at the preferences of caregivers, as well as the benefits to caregivers of caregiving. The majority of caregivers do not suffer any adverse events (Hermus et al., Reference Hermus, Stonebridge, Thériault and Bounajm2012), and caregivers often prefer to take on the caregiving role (Hollander, Liu, & Chappell, Reference Hollander, Liu and Chappell2009). Thus, the caregiver’s preferences to caregiving may play an important role in the likelihood of LTC wait-list placement. Another important limitation is the lack of data on caregiver resilience, or their ability to continue or cope through adverse advents. The extent to which these factors are associated with LTC wait-list placement remains unclear. Future work with more detailed data on caregivers may explore this in greater detail.
System-level characteristics beyond the formal home care services funded by the LHIN were not collected as part of the assessment. In addition, the RAI-HC data cannot be used to see if care recipients were purchasing additional formal services out-of-pocket, which may have an influence on the likelihood of wait-list placement.
Finally, our data were limited to one LHIN, which was not fully representative of the Ontario population. Despite this, the results are generalizable to areas that contain large rural areas. Moreover, the analysis underscores the importance of assessing factors at the care recipient, informal caregiver, and formal system levels as they have all been shown here to play important roles in wait-list placement. It would be important to establish whether these relationships hold across regions.
Conclusion
Consistent with the literature, we found that care recipient characteristics and needs were a significant predictor of referral to LTC in North West Ontario. Age, cognition, as well as capacity to perform ADLs were all associated with the likelihood that a care recipient would be wait listed. However, care recipient needs turned out to be only one, albeit important, part of the story; as hypothesized, supply-side factors related to formal system and informal caregiver capacity also played a significant role.
Supplementary Material
To view supplementary material for this article, please visit https://doi.org/10.1017/S071498081700023X
Appendix
Table AI: Variable definitions
