Does the Profit Motive Matter? COVID-19 Prevention and Management in Ontario Long-Term-Care Homes.

We introduce evidence that for-profit long-term-care providers are associated with less successful outcomes in coronavirus disease 2019 outbreak management. We introduce two sets of theoretical arguments that predict variation in service quality by provider type: those that deal with the institution of contracting (innovative competition vs. erosive competition) and those that address organizational features of for-profit, non-profit, and government actors (profit seeking, cross-subsidization, and future investment). We contextualize these arguments through a discussion of how contracting operates in Ontario long-term care. That discussion leads us to exclude the institutional arguments while retaining the arguments about organizational features as our three hypotheses. Using outbreak data as of February 2021, we find that government-run long-term-care homes surpassed for-profit and non-profit homes in outbreak management, consistent with an earlier finding from Stall et al. (2020). Non-profit homes outperform for-profit homes but are outperformed by government-run homes. These results are consistent with the expectations derived from two theoretical arguments-profit seeking and cross-subsidization-and inconsistent with a third-capacity for future investment. © Canadian Public Policy / Analyse de politiques.

Introduction

The novel coronavirus (coronavirus disease 2019, or COVID-19) pandemic has drawn public attention to the state of long-term-care homes (LTCHs) in Canada and around the world. As of July 2020, around 80 percent of Canada’s COVID-19 deaths were linked to long-term-care or retirement homes (Canadian Foundation for Healthcare Improvement 2020). Concerns about the conditions in LTCHs were elevated when a military report documented unsafe and unsanitary conditions in the five Ontario homes to which military personnel were deployed (Headquarters 4th Canadian Division Joint Task Force 14 May 2020). In this article, we apply evidence to a question that has garnered significant public attention: when it comes to the quality of care in LTCHs, does the profit motive matter? Specifically, have COVID-19 outcomes been worse in for-profit LTCHs than in non-profit and government-run LTCHs? We focus on Ontario, where COVID-19 outbreaks have occurred in at least 491 LTCHs.

In doing so, we make two contributions to the literature on contracting. First, we contribute evidence on a service quality outcome indicator that is underused in privatization research: outbreak prevention and management, which is critical to ensuring the safety of long-term-care residents. Second, we draw on the public service contracting literature to provide a clear theoretical basis for hypothesizing differences in service quality dependent on provider type. Our analysis introduces two sets of theoretical arguments: those that deal with the institution of contracting and those that address organizational features of for-profit, non-profit, and government actors. We contextualize these arguments through a discussion of how contracting operates in Ontario long-term care. That analysis leads us to exclude the institutional arguments while retaining the arguments about organizational features as our three hypotheses.

We find that COVID-19 outcomes were better in government-run LTCHs than in contracted private homes, especially for-profit LTCHs. This variation is linked primarily to outbreak management—government LTCHs had lower death rates in homes with outbreaks—whereas by February 2021, there was no statistically significant effect on outbreak prevention, perhaps because COVID-19 outbreaks had occurred in most LTCHs. Non-profit LTCHs occupy a middle ground, performing better than for-profit homes but worse than government-run ones.

Our article complements and builds on an earlier study by Stall et al. (2020). We confirm their finding that for-profits underperformed government-run and non-profit LTCHs in outbreak management. We also offer two distinct contributions. First, we provide updated data that account for the end of the first wave and the second wave of the pandemic. The data in Stall et al. captured an early snapshot from May 2020, whereas we use data from February 2021. Accordingly, the updated data affect the main finding of Stall et al.: whereas design standard was a key driver of divergent outcomes early in the pandemic, we now find that provider type is a more important variable. Second, our article contributes theoretical rationales for expecting variation in COVID-19 outcomes among for-profit, non-profit, and government-run LTCHs.

Introducing Long-Term Care in Ontario

LTCHs are places where adults live and receive access to 24-hour nursing and personal care along with help with most to all daily activities. These facilities deliver more care than a retirement home and are regulated under a different policy regime.[1] LTCHs in Ontario are regulated by the Long-Term Care Homes Act, 2007 (LTCHA) and Regulation 79/10 (MLTC 2019; Ontario 2017). Together, these provide detailed requirements for operating a licensed LTCH (Ontario 2008). Access is organized through 14 Local Health Integration Networks (LHINs), which are responsible for planning, integrating, and funding local health care in Ontario (Ontario LHINs 2014). LHINs determine eligibility, set out home options, explain costs, and help people apply.

Under the act, LTCHs can take several organizational forms. They can be operated by non-profits, municipalities, corporations, partnerships, or for-profit sole proprietors. To operate a LTCH, an organization must obtain a license from the Ministry of Long-Term Care (Ontario 2011). License applications are competitively assessed on the basis of pre-identified criteria and the target number of total bed spaces the ministry wants to add (MOHLTC 2020). Requirements under the LTCHA and the regulation spell out standards of care and apply to every organization running a LTCH, although there are some small differences depending on the size of a home. LTCHs have service accountability agreements with the ministry that provide further specificity (MOHLTC 2008).

All licensed LTCHs receive funding from the Ontario government. The precise funding formula is complicated; however, in general it consists of a per diem funding rate set on the basis of the number of licensed beds and annual top-up funding (MOHLTC 2008, 2019a). Government funding can only be put toward eligible administrative, operational, and maintenance costs (MOHLTC 2019b). LTCHs also charge accommodation costs, which are paid by residents but set and standardized by the ministry, regardless of LTCH organization type (Ontario 2020b). There are three rates: basic, semi-private, and private. Individuals with insufficient income to afford the basic rate receive a subsidy through the Long-Term Care Home Rate Reduction Program (Ontario 2020b).

Long-term care in Canada is contracted out through a mixed welfare model that includes non-profit, for-profit, and government service provision. Although other Canadian provinces also use contracting, long-term care in Ontario is among the most privatized (Daly 2015; Palley, Pomey, and Forest 2011). In Ontario, the majority of LTCH providers, 55 percent by total number of homes, are for-profit, although there are also a sizable portion of government (16 percent) and non-profit (28 percent) homes. Canadian long-term care is no outlier: privatization through contracting, and especially through competitive welfare markets, has become increasingly common since the 1980s (Greene 1996). In the context of a pandemic, long-term-care quality has bearings on mortality rates, and so this is a particularly important policy domain in which to study the effects of privatization.

Empirical studies on long-term care have consistently found that for-profit LTCHs underperform non-profit and government homes on a range of quality indicators (Comondore et al. 2009; Hillmer 2005). Although the long-term-care literature offers a clear empirical record linking for-profit homes to lower service quality, these studies largely do not offer theoretical reasons for why one should expect such a consistent divergence.[2] Theoretical explanations focus on factors, such as staffing intensity, regulation, and management style, that need not covary with provider organization type (e.g., Ågontnes et al. 2019; Amirkhanyan et al. 2018; Banerjee and Armstrong 2015; Berta, Laporte, and Valdmanis 2005; Daly 2015; McGregor et al. 2005).

To provide a theoretical basis for explaining these empirical findings, we draw from the literatures on public service contracting in public policy and comparative politics research. Those literatures identify two sets of theories providing differential expectations for for-profit, non-profit, and government service quality in contracted social service environments: explanations that focus on the institution of contracting and explanations that are premised on organizational features of providers. In the next section, we introduce two institutional arguments—innovative competition and erosive competition—and three arguments premised on organizational features—profit seeking, cross-subsidization, and ability to invest in the future.

Interestingly, however, the public service contracting literature does not establish a clear expectation that for-profit providers will underperform government and non-profit providers. Indeed, on both the institution of contracting and organizational features, the literature offers explanations that could lead one to expect each of the three provider types to perform best or worst. In the next section, we introduce the five theoretical arguments linking provider type to care quality. To reconcile the theoretical ambivalence on service quality and provider type with the very clear empirical expectations in long-term care, we discuss how well each of the five arguments applies to long-term care in Ontario. We also devote attention to the particularities of infection prevention and control (IPAC) as a quality indicator, especially given the extraordinary nature of the COVID-19 pandemic. Given the absence of competition in this case, we argue that neither of the institutional arguments is applicable. That leaves three valid theoretical expectations rooted in the organizational features of for-profit, non-profit, and government providers.

Social Welfare Privatization and the Quality of Care

Within the literature on social welfare service contracting, there is ongoing debate regarding service quality (Amirkhanyan 2008). Some argue that privatization brings efficiency gains and increases the responsiveness of public services to user needs (Carey et al. 2017; Savas 2000, 2005). Others argue that privatization undermines service quality through the erosion of spending on care, predatory or irresponsible practices, and corruption (Morgan and Campbell 2011a; O’Toole and Meier 2004; Zuberi 2013). Other research posits that for-profit providers offer more uneven service quality (Blomqvist 2004; Stolt, Blomqvist, and Winblad 2011); underperform on certain performance measures while overperforming on others (e.g., Winblad, Blomqvist, and Karlsson 2017); or offer low-quality services only under certain conditions (Ben-Ner, Hamann, and Ren 2018; Ben-Ner, Karaca-Mandic, and Ren 2012).

The public contracting literature provides two sets of explanations for potential differences in government, non-profit, and for-profit service quality: those rooted in the institution of competitive contracting and those that pertain to the organizational features of service providers. As we show, the two institutional arguments are not applicable to Ontario long-term care because both are premised on competition, which is not truly present in this case. However, all three arguments pertaining to the organizational features of for-profit, non-profit, and government providers are applicable to Ontario long-term care.

Two Competing Narratives on Contracting and Quality

Research on privatization presents two opposing arguments about how service contracting should affect service quality: innovative competition and erosive competition. These two arguments act as foils for one another: whereas the first points to salutary effects of competition, and therefore expects private organizations (for-profits and non-profits) to offer higher quality care, the second points to the harms of competitive service contracting and expects government to perform best. Both arguments rely on the institution of competitive service contracting as the operative concept; in each, relative exposure to competition produces different effects, rather than features inherent to each organization type.

The first argument, innovative competition, claims that competition drives improvements to productivity through innovation (Baumol 2002). First, innovative providers may submit more competitive proposals to win service contracts from government. Second, when providers rely on users to survive, user choice fuels quality improvements (Le Grand 2007). Users can exit poor-quality services, causing low-quality providers to fail (Morgan and Campbell 2011b; Prado and Trebilcock 2018). In mixed markets, private providers are expected to outperform government providers because governments are shielded from the forces of competition: they may not be exposed to direct competition and, even if a government loses consumers to an alternative provider, governments cannot cease to exist in the same way that a failing for-profit or non-profit would. Thus, the mechanism of creative destruction, which is posited to drive innovation, cannot fully function for government providers.

Because the innovative competition argument relies on competition as the key driver of service quality, it is necessary to discuss whether and to what extent competition exists in the case of Ontario long-term care. Long-term care in Ontario is a system of managed competition, in which LTCHs compete for licenses to establish facilities and, once operative, for patients (Denton et al. 2006). Prospective for-profit and non-profit providers bid for LTCH licenses through a competitive request for proposals process that awards contracts on the basis of “highest quality, best price” (Skinner and Rosenberg 2006). According to the LTCHA, licenses are issued for a period of up to 30 years; when the license expires, the government decides whether or not to issue a new license. We were unable to find research indicating the average license length or how common it is for existing licensees to have their licenses terminated or not renewed.[3] However, policy discussions around the large number of licenses set to expire in 2025 suggest a preference for supporting existing providers (e.g., Griffin 2016), consistent with the approach in other social policy areas (Pue 2019). As such, competition for licenses appears to be quite limited.[4] While the license is active, LTCHs in theory compete for patients. In practice, though, the scarcity of long-term-care beds renders consumer choice impossible: the occupancy rate of long-term care is 99 percent at any given time, with an average wait time of 161 days and waiting lists exceeding 34,000 Ontarians (OHC 2019b; OLTCA 2019). Thus, the long-term context provides scant opportunities for innovation-enhancing competition, and the first argument does not apply well to LTCHs in Ontario.

The second argument, erosive competition, contends that competitive contracting undermines long-run service quality. Here, privatization through contracting is understood to create incentives for bidders to produce services as cheaply as possible, which creates insufficient or uneven service quality (Smith and Lipsky 1993). Contracting out services usually involves some kind of competitive bidding system in which providers apply to deliver services in exchange for funding. There can be considerable variation in the bases on which bids are evaluated—for instance, some contracting systems prioritize cost reductions more than others—as well as in the rules of compensation—which can be, for instance, input based, output based, or performance based (Pue 2019). Although these particulars can matter a great deal, the erosive competition thesis argues that competitive contracting that emphasizes cost reduction can create challenges for service quality when these reductions are found not through productivity-enhancing innovation but rather through decreased care or the externalization of costs, such as the elimination of worker benefits and job security (Armstrong et al. 2020; Pue 2019). In a mixed welfare context, government providers are insulated from the effects of competitive contracting. The erosive competition argument thus predicts that the highest quality care will be offered by governments, whereas private providers will offer lower quality care as a result of being exposed to the institution of competitive contracting.

The erosive competition argument is salient in contexts in which providers compete on the price of the service that they will offer. Price-based competition is, however, only relevant in a very limited sense for Ontario long-term care. Providers do compete on the price of building a LTCH. There is evidence that competition for licenses has created incentives for for-profits and non-profits to make decisions that might reasonably be expected to erode the quality of care (Cloutier-Fisher and Skinner 2006; Skinner and Rosenberg 2006; Williams et al. 2010). Moreover, municipal LTCHs are approved through a separate process, which may shield them from the competitive licensing system (MLTC 2020). However, as discussed earlier, once established a LTCH very rarely loses its license. Also, there is no ongoing price-based competition for operating long-term care: government subsidy levels are defined by funding formulas that apply uniformly across all LTCHs in the province, and the cost for clients is fixed. Private providers cannot bid down the quality of care by proposing a smaller government subsidy.

Although innovative competition and erosive competition both establish plausible arguments connecting the institution of service contracting to care quality, neither supposition is well suited to the case of long-term care in Ontario. To better understand the impact of LTCH type on quality, we move next to arguments about organizational features.

Three Arguments Linking Organization Type to Service Quality

Three arguments present distinct expectations based on organizational features of for-profit, non-profit, and government providers: profit seeking, cross-subsidization, and investment in the future. These arguments better align with the realities of LTCH contracting in Ontario.

The first argument predicts that for-profits will provide the lowest quality care by linking care quality to the organizational feature of profit seeking shared by for-profits (O’Neill et al. 2003). For-profits are by definition profit seeking and so must seek to extract value to distribute to owners and shareholders (Friedman 1970). In contrast, government and non-profit social care providers spend all revenue on their programs and activities. To some, this creates an expectation that, all else being equal, in for-profit homes fewer resources will go toward service provision and crisis preparedness than in non-profit or government homes. (Of course, an alternate argument proposes that the profit motive gives businesses incentives to use resources more effectively, which some suggest may maintain service quality.) Extant evidence suggests a link between the profit motive and reduced care in LTCHs: for-profit LTCHs provide fewer hours of direct care per resident, which has been linked to higher rates of LTCH patient admission to hospital (Berta et al. 2005; McGrail et al. 2006). Moreover, average personal support worker wages are highest in government-run LTCHs (LTCSSAG 2020).

Next, the ability to cross-subsidize—to draw on external resources above the baseline provided by service contracts, user fees, or both—potentially enables some service providers to offer higher-quality services than others. By providing supplementary long-term-care funding through municipal budgets or philanthropic donations, government and non-profit LTCHs can effectively plan to operate the service at a loss, which is something that a for-profit provider cannot do. For-profits are, thus, least able to cross-subsidize. Governments can cross-subsidize using tax revenue. For example, the City of Toronto (2012) has estimated that it provides 20 percent of the operating costs of its municipal LTCHs. Non-profits that receive philanthropic donations can direct these funds to service provision where government funding might otherwise be minimal (Marwell and Calabrese 2015). We randomly sampled 20 non-profit LTCHs to understand the extent to which these organizations supplement their service with philanthropic revenue. See the online Appendix A for information on how we collected these data. The sampled non-profit LTCHs received a mean 66 percent of their revenue from the provincial government and a further mean 25 percent from earned income (largely user fees). As such, a mean 9 percent of non-profit LTCH revenue in this sample was from other sources, including philanthropy and occasionally grants from local or federal government.[5] The limited role of philanthropic revenue in the sampled non-profit LTCHs suggests that we might expect less of a positive effect from cross-subsidization on quality in non-profit providers, compared with government providers.

A final argument concerns the ability of different organization types to borrow or redirect money to invest in the future. Investments in, for example, equipment, facility upgrades, or research and development can improve the capacity of providers to offer innovative, high-quality services and respond to adverse events such as respiratory outbreaks. Although for-profits do not have an external source of funding with which to cross-subsidize services, they are able to borrow and therefore to make investments in the future.[6] For-profits are the most able to access capital and invest, and as such this argument predicts the highest service quality from for-profits. This is likely to be especially true for for-profit LTCHs that are part of a chain. Government LTCHs in Ontario are run by municipalities, which face considerable restrictions on borrowing. Non-profits face constraints in accessing capital (Salamon 2003) and donor preferences for direct expenditures (Lecy and Searing 2015), both of which severely limit these organizations’ capacity to invest in the future.

Hypotheses

In this article, we test the association between LTCH provider type and outcomes in preventing and managing COVID-19 outbreaks, which we present as a service quality indicator. The public services contracting literature provides three distinct theoretical expectations, set out in the previous section, for how organization type should affect a LTCH’s service quality—and therefore infection management and control outcomes. Accordingly, we test three hypotheses: On the basis of previous empirical findings on quality outcomes in long-term care (e.g., Comondore et al. 2009; Hillmer 2005), including a recent article on the COVID-19 pandemic (Stall et al. 2020), we expect to find evidence consistent with Hypothesis 1 and inconsistent with Hypotheses 2 and 3. However, there is a theoretical basis for all three hypotheses.

H1 (profit-seeking): Government-run and non-profit LTCHs will be best at preventing and managing COVID-19 outbreaks, whereas for-profit LTCHs will be worst.

H2 (cross-subsidization): Government-run LTCHs will be best at preventing and managing COVID-19 outbreaks, whereas non-profits will be middle performers and for-profits will be worst.

H3 (future investments): For-profit LTCHs will be best at preventing and managing COVID-19 outbreaks, whereas government-run LTCHs will be middle performers and non-profits will be worst.

Data and Methods

Infection Prevention and Management as a Care Quality Indicator

Social welfare service quality is typically measured using indicators such as regulatory violations, hospital admissions, and staffing ratios. Here, we use COVID-19 outbreak prevention and management as indicators of service quality. The capacity to prevent and manage outbreaks is a direct measure of LTCH quality, because the ability to keep residents safe is a critical aspect of care. Pandemics—and outbreaks more broadly—test whether organizations have sufficient resources and planning to cope with predictable but extraordinary situations. Respiratory infection outbreaks are common and can be especially dangerous for LTCH residents (Lee et al. 2019; Storr et al. 2017). As such, a high-quality LTCH should excel at minimizing the risk to a resident of contracting a respiratory virus while living in the facility.

Best practices for IPAC in Ontario LTCHs are articulated in The Control of Respiratory Infection Outbreaks in Long-Term Care Homes, 2018 (MOHLTC 2018a). This guidance was developed for respiratory viruses that commonly cause infection outbreaks, including influenza and coronavirus (MOHLTC 2018a). IPAC programs entail immunization, education, policy and procedures, active and passive surveillance to identify outbreaks, and a series of outbreak management activities (MOHLTC 2018a). These core activities in turn rely on a well-trained staff complement of sufficient size to allow for surge capacity in the event of an outbreak (MOHLTC 2018a). Successful IPAC also requires adequate facility space and access to sufficient supplies, for instance disinfectants and personal protective equipment (MOH 2020b; MOHTLC 2018a). As such, we should expect infection outcomes to be best for the organization type that is most able to offer high-quality services overall, especially when it comes planning for resource-intensive care scenarios.

Case Selection and Description

This study includes only one Canadian province: Ontario. Although the selection of a single province limits the study’s generalizability, Ontario is an appropriate choice for two reasons. First, Ontario has a sufficiently large number of LTCHs of all three organization types, as well as a large number of LTCHs with outbreaks, thus allowing for sophisticated quantitative analysis. Second, attention to problems in the quality of long-term care in Ontario has been particularly acute as a dimension of the pandemic. By focusing on a single province, we hold the regulatory framework for LTCHs (which varies from province to province) constant.

Data Collection

Of the 630 LTCHs in Ontario, 608 were listed in a database managed by the MOHLTC (2018b).[7] Using this base dataset, we then added data on COVID-19 in LTCHs from a Government of Ontario list of facilities with active and inactive outbreaks as of 11 February 2021 (Ontario 2020a). The Government of Ontario LTCH outbreaks data are self-reported by the LTCHs to the MOHLTC (2020). As of 11 February 2021, 491 homes had active and inactive COVID-19 outbreaks. Further details on data collection are included in the online Appendix B. An early round of data collection was done on 18 August 2020. Analysis using these data is included in the online Appendix C, capturing the results for only the first wave of the pandemic.

Variables

We use three dependent variables to measure LTCHs’ ability to prevent and manage outbreaks. A binary variable indicates whether an outbreak occurred, measuring the ability of LTCHs to prevent outbreaks. To capture the ability of LTCHs to manage outbreaks, we also examine resident COVID-19 deaths as a percentage of total beds. Finally, we used the number of resident COVID-19 deaths in a LTCH to capture the combined ability of LTCHs to prevent and manage outbreaks.

The main independent variables in this study concern the LTCH licensee organization type—government, non-profit, or for-profit—derived from an Ontario government database (Ontario 2008). To control for the higher chance that an outbreak occurs in a LTCH with more residents, the number of beds in a LTCH is used as a control variable. It may be more difficult for a LTCH to prevent and manage outbreaks if the needs of its clientele are particularly acute. Unfortunately, data on the overall case mix are not publicly available.

One can distinguish between two types of for-profit homes: those that are part of chains, where one company operates more than one LTCH, and those that are independently run. We present the results for tests that distinguish between chain and non-chain for-profit LTCHs in the online Appendix D. We find no statistically significance difference between the performance of chain and non-chain for-profit homes. These tests show a clear difference between the performance of for-profit chains and government-run homes in preventing deaths. Whether non-chain for-profit homes perform worse than government-run homes is dependent on the model and controls included.

We include three sets of control variables that encapsulate more traditional service quality indicators. The first is the design standard to which beds in the home conform. Design standard is an important variable because it indicates the maximum occupancy per room in a LTCH. For a detailed description, see the online Appendix E. We control for the design standard prevailing for a plurality of beds in the LTCH. This mirrors the control used by Stall et al. (2020) for LTCH design standard that they find, on the basis of data from May 2020, accounts for much of the poor performance of for-profit homes. We do not find that the design standard control variable accounts for the differences among for-profit, non-profit, and government-run LTCHs.

Our second control is performance data from Health Quality Ontario (2020), including the proportion of patients not living with psychosis given antipsychotic medications; with depression whose symptoms worsened; who developed new or worse pressure ulcers; who experienced falls; who were physically restrained; and who experienced moderate pain daily. Third, we control for the number of inspection violations and compliance orders that a LTCH received (using MOHLTC 2018b). Online Appendix F indicates that these control variables are not correlated with home type, suggesting that they capture independent factors that affect quality of care. We include controls for design standards in our second models and add controls for all other performance measures in our third models. Table 1 shows the means and standard deviations broken down by home type for each variable used in the analysis.

Table 1:

Variable Totals and Means by Home Type

Variable	Government	Non-Profit	For-Profit
Continuous variables, mean (SD)
Beds	162.061	122.269	119.911
	(79.404)	(93.973)	(58.316)
Anti-psychotics	17.807	18.765	18.746
	(6.112)	(9.184)	(7.620)
Pressure ulcers	2.968	2.723	2.684
	(1.475)	(1.767)	(1.423)
Falls	17.412	15.431	16.711
	(4.359)	(5.217)	(4.652)
Restraints	6.287	5.125	3.034
	(7.698)	(5.795)	(4.522)
Depression	26.966	23.258	22.320
	(9.681)	(10.478)	(9.621)
Pain	5.720	6.575	5.186
	(4.889)	(7.505)	(5.054)
Orders	0.434	0.663	0.620
	(1.051)	(1.339)	(1.248)
Violations	6.141	5.727	5.398
	(5.151)	(5.582)	(4.882)
Totals, no. (%)
Plurality new beds	53	87	125
	(53.54)	(5.10)	(37.20)
Plurality A standard	27	19	0
	(27.27)	(11.38)	(0.00)
Plurality B standard	6	16	34
	(6.06)	(9.58)	(10.12)
Plurality C standard	13	27	166
	(13.13)	(16.17)	(49.40)
Plurality D standard	0	3	II
	(0.00)	(1.80)	(3.17)
Plurality ELDCAP	0	15	0
	(0.00)	(8.98)	(0.00)
Total homes	99	167	336

Note: ELDCAP = Elderly Capital Assistance Program.

Sources: Data come from a data set compiled by the authors using data from Ontario (2008, 2020a), Ontario Health Coalition (2019), Health Quality Ontario (2020), and MOHLTC (2018b), as noted in the “Data Collection and Variables” section.

Regional Heterogeneity

The likelihood of an outbreak varies regionally because it is affected by the prevalence of COVID-19 in that region and the policies made by the LHIN. To account for this, we use fixed effects and clustered standard errors, both based on the LTCH’s LHIN. Because LHINs correspond to regions, controlling for them captures both administrative and regional factors that affect the likelihood of an outbreak. These fixed effects are preferable to controls for the demographics of the neighbourhood in which a home is situated, because LTCH populations do not always match their neighbourhoods. In negative binomial models, fixed effects can only be used in one part of the model. We use them for the likelihood of an outbreak because that is most likely to be affected by regional difference. Unlike Stall et al. (2020), we used fixed effects in place of a control for the extent of an outbreak in a region because fixed effects allow us to control for any other regional factors that may influence a home’s response to COVID-19. Models without fixed effects are included in the online Appendix G as a robustness check and do not produce substantially different results.

Ordinary Least Squares, Logit, and Zero-Inflated Negative Binomial Models

LTCH performance depends both on infection prevention and infection management. We test these variables separately because the ability of a home to prevent an outbreak is best measured as a binary variable, whereas a home’s ability to manage an outbreak is best measured by the deaths that occur.[8] Complicating these tests is the long right tail in the distribution of deaths across homes, demonstrated in Figure 1. As of 11 February 2021, 379 of the 608 LTCHs did not have a death. Most homes that had deaths were able to limit the number of deaths to between 1 and 20. The long right tail suggests that independent variables might not have a linear effect on the number of deaths in a home, violating the assumptions of an ordinary least squares (OLS) model.

Figure 1:

Number of Homes with Different Numbers of COVID-19 Deaths

We run three types of models. To test LTCHs’ ability to prevent outbreaks, a logit model is run, using the presence of an outbreak as the dependent variable. For ease of interpretation, we include a figure showing predicted probabilities along with the regression table.

To test the ability of LTCHs to manage outbreaks, a second set of analyses is run, examining death rates in homes with outbreaks, using OLS models. Because OLS models are relatively easy to interpret, coefficients are reported in the main body of the article.

A zero-inflated negative binomial model is then run to account for the long right tail in the distribution of deaths as well as the large number of homes with no deaths. Count models are a better fit for data with this kind of distribution. Because the distribution of deaths is over-dispersed (there is greater variance in the number of deaths than a Poisson model would assume), a negative binomial model is run. A zero-inflated negative model is used because there are a larger number of homes with no deaths than a negative binomial model would assume. The zero-inflated model estimates each variable’s effect on the likelihood of excess zeros and then fits a negative binomial model that accounts for different variables’ effects on the number of excess zeros. This produces two estimates for each independent variable. One estimates the effect each variable has on the number of excess zeros (in this case, the effect a variable has on the likelihood that a home has no deaths) and a second estimates the effect a variable has on the likelihood of an additional death, accounting for the likelihood that different homes will have no deaths. Doing so distinguishes between the effect each variable has on outbreak prevention (where outbreaks result in deaths) and the effect each variable has on a LTCH’s ability to manage an outbreak once a death has occurred. These models are commonly used for data with similar distributions, such as studies of the effects of price on cigarette smoking (Sheu et al. 2004). We include predicted probabilities based on the model for the number of deaths in different types of homes alongside regression tables.

Note: COVID-19 = coronavirus disease 2019.

Source: Ontario (2020a).

There is a trade-off between the OLS and zero-inflated binomial models. The percentage-of-deaths dependent variable in the OLS models does a better job of accounting for the number of deaths in a home, whereas the zero-inflated negative binomial model can better account for the distribution of deaths across homes. Fortunately, the models produce similar results. Online Appendix F shows that, to the extent there are differences in the average size of homes, government-run homes tend to be larger even though they also perform better in each of the models.

Results

Likelihood of an Outbreak

In the first analyses in this article, we examine the ability of different types of home to prevent outbreaks. Figure 2 shows the percentage of homes with outbreaks across the three different home types. At 84 percent, for-profit LTCHs had a slightly higher percentage of outbreaks than non-profit and government-run homes, at 76 percent and 80 percent, respectively. This analysis does not incorporate the region that the home is in (and as a result the extent of the COVID-19 outbreak in the region), the number of beds in a home, the design standard, or other indicators of care quality that may be independent of home type.

Figure 2:

Percentage of Outbreaks by Home Type

Notes: The figure includes means with no other variables controlled for. None of the differences between the means by each home type is statistically significant (statistical significance was tested using a t-test for differences in means).

Source: Ontario (2020a).

Logistic regressions are used to account for these factors. Table 2 shows that when these factors are accounted for, there is little difference in the likelihood of an outbreak among government-run, non-profit, and for-profit homes. In Model 1, government-run homes are slightly less likely to have an outbreak than for-profit homes, but this difference is only statistically significant at the 90 percent confidence level. Once controls for the bed standard are added (Model 2) as well as the other measures of home quality (Model 3), this difference is no longer statistically significant. Thus, there is no clear evidence of a difference between government-run, non-profit, and for-profit LTCHs in outbreak prevention. Notably, however, online Appendix C shows that government-run homes were less likely to see outbreaks than for-profit homes during the first wave of the pandemic. This suggests that the lack of a difference between government-run and for-profit homes in outbreak prevention is a result of the increase in outbreaks since August 2020—particularly because there have now been outbreaks in the vast majority of LTCHs.

Table 2:

Effect of Home Type on Likelihood of Outbreaks

	Model
Variable	1	2	3
Home type
Government run	−0.801*	−0.727	−0.730
	[−1.672, 0.070]	[−1.770, 0.317]	[−1.683, 0.224]
Non-profit	−0.419	−0.336	−0.320
	[−1.209, 0.371]	[−0.956, 0.283]	[−0.967, 0.326]
Beds(10)	0.223***	0.203***	0.215***
	[0.173, 0.273]	[0.143, 0.263]	[0.147, 0.283]
Plurality
New beds		−0.364	−0.364
		[−2.589, 1.862]	[−2.589, 1.862]
A standard		−0.928	−0.928
		[−3.306, 1.451]	[−3.306, 1.451]
B standard		−0.976	−0.976
		[−3.293, 1.342]	[−3.293, 1.342]
C standard		−0.681	−0.681
		[−3.012, 1.651]	[−3.012, 1.651]
ELDCAP		−1.342	−1.342
		[−5.363, 2.679]	[−5.363, 2.679]
Anti-psychotics			0.036**
			[0.002, 0.070]
Pressure ulcers			−0.010
			[−0.211, 0.192]
Falls			−0.028
			[−0.066, 0.011]
Restraints			0.001
			[−0.068, 0.069]
Depression			0.011
			[−0.018, 0.040]
Pain			−0.024
			[−0.061, 0.012]
Orders			0.070
			[−0.159, 0.300]
Violations			−0.006
			[−0.093, 0.080]
Health region fixed effects	Yes	Yes	Yes
Constant	1.661	1.140	0.688
R 2	0.322	0.326	0.330
No. of observations	607	602	587
Non-profits (compared with government)	0.382	0.390	0.409
	[−0.702, 1.465]	[−0.662, 1.442]	[−0.684, 1.503]

Notes: 95% confidence intervals are in brackets. D-standard beds serves as the base category for bed standards. Central serves as the base category for health regions. For these models, the Central and Central West health regions have been grouped together because the fact that there have been outbreaks in every Central West home would otherwise cause each Central West home to be dropped from the model. ELDCAP = Elderly Capital Assistance Program.

p < 0.1;

p < 0.05;

p < 0.01.

Sources: Results of regression analysis conducted by the authors. Data for the analysis come from a data set compiled by the authors using data from Ontario (2008, 2020a), Ontario Health Coalition (2019), Health Quality Ontario (2020), and MOHLTC (2018b), as noted in the “Data Collection and Variables” section.

Deaths in Homes with Outbreaks

Outbreak management quality affects how effectively a home can limit deaths once one has occurred. To assess this, we examine the number of deaths as a percentage of beds in homes that experienced an outbreak. Figure 3 compares the average percentage of deaths in each home type. At just less than 2 percent, government-run homes outperform non-profit and for-profit homes by a substantial margin on this measure. For-profits have the highest average percentage of deaths, at just more than 6 percent, and non-profits are in between, with just less than 5 percent.

Figure 3:

Percentage of COVID-19 Deaths in Homes with an Outbreak

As in analyses of outbreaks, the percentage of deaths may be affected by the regional prevalence of COVID-19, the number of beds in a home, and other LTCH characteristics. These are controlled for in the OLS regression models in Table 3. Model 1 in Table 3 shows that government-run LTCHs experienced an average number of deaths 4.6 percentage points lower than for-profits. This drops to 3.3 and then 3.1 percentage points lower when the controls are introduced in Models 2 and 3. Government-run homes also outperform non-profits, seeing between 1.8 and 1.9 percentage points fewer deaths depending on the model. Non-profits perform better than for-profits when one does not control for factors affecting LTCH quality, but the difference between non-profits and for-profit homes does not remain statistically significant when these controls are added.

Table 3:

Effect of Home Type on COVID-19 Deaths as a Percentage of Beds in Homes with Outbreaks

	Model
Variable	1	2	3
Home type
Government run	−4.596***	−3.341***	−3.125***
	[−6.544, −2.647]	[−4.841, −1.841]	[−4.726, −1.524]
Non-profit	−2.727***	−1.478*	−1.296
	[−4.537, −0.918]	[−3.090, 0.133]	[−3.086, 0.494]
Beds (10)	−0.060	−0.029	−0.035
	[−0.212, 0.091]	[−0.185, 0.128]	[−0.201, 0.131]
Plurality
New beds		−1.770	−1.885
		[−5.620, 2.080]	[−5.866, 2.095]
A standard		−0.667	−1.067
		[−4.878, 3.545]	[−5.457, 3.323]
B standard		−0.249	−0.471
		[−4.018, 3.519]	[−4.516, 3.575]
C standard		2.409**	2.141**
		[0.566, 4.253]	[0.401, 3.882]
ELDCAP		−2.730	−5.012
		[−7.497, 2.037]	[−12.825, 2.801]
Anti-psychotics			0.051
			[−0.034, 0.136]
Pressure ulcers			0.186
			[−0.366, 0.737]
Falls			0.008
			[−0.193, 0.208]
Restraints			−0.043
			[−0.211, 0.125]
Depression			−0.003
			[−0.125, 0.120]
Pain			−0.027
			[−0.226, 0.171]
Orders			0.143
			[−0.782, 1.067]
Violations			0.061
			[−0.136, 0.258]
Health region fixed effects	Yes	Yes	Yes
Constant	10.283	6.593	2.948
R2	0.138	0.169	0.168
No. of observations	491	489	480
Non-profits (compared with government)	1.868**	1.863**	1.829**
	[0.417, 3.319]	[0.335, 3.391]	[0.158, 3.500]

Notes: 95% confidence intervals are in brackets. D-standard beds serves as the base category for bed standards. Central serves as the base category for health regions. COVID-19 = coronavirus disease 2019;ELDCAP = Elderly Capital Assistance Program.

p < 0.1;

p < 0.05;

p < 0.01.

Sources: Results of regression analysis conducted by the authors. Data for the analysis come from a data set compiled by the authors using data from Ontario (2008, 2020a), Ontario Health Coalition (2019), Health Quality Ontario (2020), and MOHLTC (2018b), as noted in the “Data Collection and Variables” section.

Notes: This figure includes means with no other variables controlled for. Differences for all three are statistically significant using at least the 95% confidence level (statistical significance was tested using a t-test for differences in means). COVID-19 = coronavirus disease 2019.

Source: Percentages calculated using data from Ontario (2020a).

This evidence suggests that, once an outbreak has occurred, government-run LTCHs have been better able to prevent deaths than other home types—particularly compared with for-profit LTCHs—but there is also a statistically significant difference between government-run and non-profit LTCHs. The relative strength of non-profits is less clear. Although they perform worse than government-run LTCHs, their stronger performance compared with for-profits is not robust to controls.

Analysis Combining Outbreaks and Deaths

In a final set of analyses, we examine the combined ability of different homes to prevent outbreaks and limit deaths. Figure 4 shows the average number of deaths as a percentage of beds across each LTCH type, including homes that both had and did not have outbreaks. Government-run homes have the lowest percentage of deaths, at 1.3 percent. By contrast, the percentage of deaths is 3.4 percent in non-profit LTCHs and 5.6 percent in for-profit LTCHs.

Figure 4:

Percentage of COVID-19 Deaths in Homes (Including Those without Outbreaks)

Table 4 shows the results of zero-inflated binomial models that capture the performance of different LTCH types, accounting for the health region that they are in, their size, and the controls included in the previous analyses. The bottom half of the table shows the effects for the zero-inflated portion of the model: the effect of home type on the probability that a home has no deaths. The top half of the table shows the effect of home type on the likelihood of additional deaths in a home, accounting for the different probabilities for each LTCH type of having no deaths. Consistent with the analysis in the preceding subsections of this article, most of the difference between LTCH types comes through in the top half of the table, looking at additional deaths. There is little evidence of a difference between provider types when one looks at their ability to prevent any deaths from occurring at all.

Table 4:

Effect of Home Type on COVID-19 Deaths as a Percentage of Beds in Homes with Outbreaks

	Model
Variable	1	2	3
Likelihood of additional deaths
Home type
Government run	−1.148***	−1.136***	−1.128***
	[−1.411,0.885]	[−1.397,−0.874]	[−1.446,−0.810]
Non-profit	−0.517***	−0.515***	−0.496***
	[−0.804,−0.230]	[−0.795,−0.236]	[−0.752,−0.240]
Beds (10)	0.034***	0.033***	0.026***
	[0.019,0.049]	[0.018,0.047]	[0.009,0.043]
Anti-psychotics			0.002
			[−0.017,0.020]
Pressure ulcers			−0.066
			[−0.193,0.061]
Falls			0.005
			[−0.025,0.034]
Restraints			−0.011
			[−0.063,0.041]
Depression			−0.006
			[−0.024,0.012]
Pain			−0.017
			[−0.042,0.008]
Orders			0.014
			[−0.022,0.041]
Violations			0.010
			[−0.022,0.041]
Constant	2.339	2.359	2.686
Non-profits (compared with government)	0.631***	0.620***	0.632***
	[0.301,0.961]	[0.290,0.951]	[0.254,1.010]
Likelihood of no deaths
Government run	0.572	0.731**	0.534
	[−0.115,1.259]	[0.012,1.450]	[−0.291,1.360]
Non-profit	0.314	0.255	0.227
	[−0.081,0.708]	[−0.258,0.767]	[−0.227,0.681]
Beds (10)	−0.101***	−0.112***	−0.1 11***
	[−0.135,−0.068]	[−0.156,−0.069]	[−0.161,−0.061]
Plurality
New beds		0.323	0.118
		[−1.483,2.128]	[−1.815,2.052]
A standard		−0.831	−1.078
		[−2.974,1.313]	[−3.307,1.150]
B standard		0.427	0.108
		[−1.385,2.238]	[−1.828,2.043]
C standard		−0.288	−0.508
		[−2.192,1.617]	[−2.566,1.550]
ELDCAP		15.252***	16.953***
		[12.672,17.831]	[14.356,19.551]
Anti-psychotics			−0.008
			[−0.045,0.028]
Pressure ulcers			−.118
			[−0.372,0.137]
Falls			0.051**
			[0.004,0.097]
Restraints			0.026
			[−0.036,0.089]
Depression			0.004
			[−0.025,0.033]
Pain			0.042
			[−0.015,0.099]
Orders			−0.147*
			[−0.313,0.019]
Violations			0.019
			[−0.045,0.083]
Health region fixed effects	Yes	Yes	Yes
Constant	−0.026	0.115	−0.286
Non-profits (compared with government)	−0.258	−0.476	−0.308
	[−1.056,0.539]	[−1.253,0.300]	[−1.152,0.536]
No. of observations	607	602	587

Notes: 95% confidence intervals are in brackets. ELDCAP beds serves as the base category for bed standards. Central serves as the base category for health regions. COVID-19 = coronavirus disease 2019; ELDCAP = Elderly Capital Assistance Program.

p < 0.1;

p < 0.05;

p < 0.01.

Sources: Results of regression analysis conducted by the authors. Data for the analysis come from a data set compiled by the authors using data from Ontario (2008, 2020a), Ontario Health Coalition (2019), Health Quality Ontario (2020), and MOHLTC (2018b), as noted in the “Data Collection and Variables” section.

Notes: This figure includes means with no other variables controlled for. Differences for all three are statistically significant using at least the 95% confidence level (statistical significance was tested using a t-test for differences in means). COVID-19 = coronavirus disease 2019.

Source: Percentages calculated using data from Ontario (2020a).

The zero-inflated binomial model shows a clear order among government-run, non-profit, and for-profit LTCHs with respect to their ability to prevent additional deaths. Government-run homes perform best. The likelihood of an additional death in a government-run LTCH is lower than the likelihood of an additional death in either a non-profit or a for-profit LTCH. For-profits are the worst performing, and non-profits are in the middle. The likelihood of an additional death in a non-profit LTCH is higher than in a government-run home but lower than in a for-profit home. These results are statistically significant regardless of which controls are included in the model. Notably, the clear difference between government-run and non-profit homes develops only after the second wave of the pandemic. Online Appendix C, which looks only at deaths that occurred before 18 August 2020, does not show a significant difference between government-run and non-profit LTCHs.

Predictive margins for the number of COVID-19 deaths are calculated and presented in Figure 5 because the substantive effects associated with the coefficients of zero-inflated binomial models can be difficult to interpret. Government-run LTCHs have the fewest predicted number of deaths at just over two deaths per home. The model indicates that for-profits have the highest number of deaths, at an average of just under eight per home, and non-profits fall in between at 4.5.

Figure 5:

Predicted Average Number of COVID-19 Deaths

Taking all of these analyses together we find substantial support for Hypothesis 1—which expects that the profit motive should lead for-profit LTCHs to perform worst—and Hypothesis 2—that the ability of governments and, to a limited extent, non-profits to cross-subsidize should lead us to expect them to outperform for-profit LTCHs. The findings are inconsistent with Hypothesis 3, which expects for-profits to perform best, and non-profits worst, based on capacity to invest. This is summarized in Table 5. More important, these differences come through largely in the ability of homes to stop outbreaks from leading to deaths as opposed to preventing outbreaks from occurring at all.

Table 5:

Summary of Hypotheses and Results

	Findings
Hypothesis	Management	Prevention
1: Profit-seeking	Supported	Inconclusive
2: Cross-subsidization	Supported	Inconclusive
3: Investment in the future	Not supported	Inconclusive

Conclusions

This article demonstrates that government-run LTCHs outperformed for-profit LTCHs in their management of the COVID-19 pandemic. We began with two groupings of theoretical arguments that establish expectations connecting provider type to service quality. The first grouping of arguments—innovative competition and erosive competition—is rooted in the institution of competitive contracting. We find that neither argument is suitable for the case of managed competition in Ontario long-term care and thus dispense with them. A second set of arguments is premised on the organizational characteristics of for-profit, non-profit, and government provider types: profit seeking (Hypothesis 1), cross-subsidization (Hypothesis 2), and capacity to make future investments (Hypothesis 3). Qualitative research analysing the motives and priorities of service providers is needed to more directly test the causal mechanisms underlying these explanations. Nevertheless, we find evidence of outcomes that are consistent with Hypotheses 1 and 2 and inconsistent with Hypothesis 3.

Notes: This figure shows predictive margins based on model 3 in Table 4. Differences between each home type are statistically significant at the 95% confidence level. COVID-19 = coronavirus disease 2019.

Predictive margins calculated using a data set compiled by the authors using data from Ontario (2008, 2020a), Ontario Health Coalition (2019), Health Quality Ontario (2020), and MOHLTC (2018b) as noted in the “Data Collection and Variables” section.

Our findings indicate that, as of 11 February 2021, government-run LTCHs in Ontario were better at managing the COVID-19 pandemic than private service providers. For-profit LTCHs were least able to manage outbreaks, and non-profits performed in the middle. We find that although no type of home was particularly effective at preventing outbreaks, government-run LTCHs were the most successful at limiting the number of deaths once an outbreak had occurred. This comes through in both the analysis of only those homes in which outbreaks occurred and the analysis of the combined ability of homes to prevent outbreaks and prevent deaths once an outbreak has occurred. Non-profits appear to occupy a middle ground, performing better than for-profit but worse than government-run homes with respect to limiting deaths. The clearest difference between non-profit and for-profit LTCHs occurs when we run the model that includes every home and accounts for the chance that there were no deaths, as well as the chance of additional deaths once one has occurred. This article confirms a finding from Stall et al. (2020), who use data from May 2020, that for-profits underperformed government-run and non-profit LTCHs in outbreak management. Whereas in Stall et al. design standard was a key driver of divergent outcomes, we now find that provider type is a more important variable.

These outcomes are consistent with the hypothesis that expects profit-seeking organizations (for-profits) to have organizational features that lead to lower quality care in these LTCHs, relative to government-run and non-profit homes. Profit seeking, it appears, may have decreased the quality of care offered, leading to higher death rates in for-profit LTCHs than in government-run homes. However, another possible explanation is cross-subsidization. Government-run and non-profit LTCHs both have external sources of funding above and beyond what for-profits can access (tax revenue and philanthropic donations, respectively). However, as we note in the “Three Arguments Linking Organization Type to Service Quality” section of this article, the capacity to cross-subsidize is likely higher for government-run than non-profit homes. This is an interesting variable worth further exploration, especially in areas of social care in which surge events occur periodically—and especially as governments consider policy changes to improve the efficacy of essential services in responding to infection outbreaks. Future research should also examine the distribution of the Government of Ontario’s emergency COVID-19 funding for long-term care, and especially which provider types have benefited the most.

From a public policy perspective, the results of this study provide evidence for the claim that governments should consider curtailing the for-profit provision of long-term care. The middling performance of non-profits points to the possibility that all contracted care should be replaced with direct government provision of care. However, further research should be done to understand why non-profit long-term care outperformed for-profit provision and underperformed government provision in outbreak management. It is possible that the variation is explained by cross-subsidization, if tax revenue enables municipal government LTCHs to supplement the service more than philanthropic revenue allows for non-profits to do so. Capacity to make future investments is another possibility, because non-profits experience particularly difficult limitations in accessing finance. There may also be other explanations that this article does not contemplate. Interview research could fruitfully tease out the advantages and drawbacks of non-profit long-term care.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.