Abstract
Background The adult social care sector is being increasingly outsourced to for-profit providers, but the impacts of privatisation on service quality and resident outcomes are unclear. During the COVID-19 pandemic, for-profit providers have been accused of failing their residents by prioritising profits over care, prevention, and caution, which has been reported to result in a higher prevalence of COVID-19 infections and deaths in for-profit care homes. Although many of these reports are anecdotal or based on news reports, there is a growing body of academic research investigating ownership variation across COVID-19 outcomes, which has not been systematically appraised and synthesised.
Objectives To identify, appraise, and synthesise the available research on ownership variation in COVID-19 outcomes (outbreaks, infections, deaths, shortage of personal protective equipment (PPE) and staff) across for-profit, public, and non-profit care homes for older people, and to update our findings as new research becomes available.
Design Living systematic review.
Methods This review was prospectively registered with Prospero (CRD42020218673). We searched 17 databases and performed forward and backward citation tracking of all included studies. Search results were screened and reviewed in duplicate. Risk of bias (RoB) was assessed in duplicate according to the COSMOS-E guidance. The results were synthesised according to RoB, model adjustment, and country context and visualised using harvest plots.
Results Twenty-eight studies across five countries were included, with 75% of included studies conducted in the Unites States. For-profit ownership was not consistently associated with a higher probability of COVID-19 outbreaks across included studies. However, there was compelling evidence of worse COVID-19 outcomes following an outbreak; with for-profit care homes having higher rates of accumulative infections and deaths. For-profit providers were also associated with shortages in PPE, which may have contributed to the higher incidence of infections and deaths.
Chain affiliation was often found to be correlated with higher risk of outbreaks, but not consistently associated with an elevated number of deaths and infections. Private equity ownership was not consistently associated with worse COVID-19 outcomes.
Conclusion For-profit status was a consistent risk factor for higher cumulative COVID-19 infections and deaths. Thus, ownership among providers may be a key modifiable factor which can be regulated to improve health outcomes in vulnerable populations and reduce health disparities. This review will be updated as new research becomes published, which may change the conclusion of our synthesis.
Introduction
The COVID-19 pandemic has disproportionately affected people living in residential care, who are estimated to account for approximately half of all COVID-19 deaths 1–3. This disproportionate impact can be understood, in part, in terms of the vulnerability of people residing in such facilities and a lack of early intervention and support 4. However, the structural and institutional risk factors are not yet well understood. In many countries, adult social care services are delivered by a combination of for-profit, non-profit, and public providers 5. While it is well documented that quasi-market provision has increased the market share of private, and in particular, for-profit care providers 6,7, it is unclear if this provision design has improved service quality. The question of whether ownership influences quality of care across different provider types is well researched. For example, a 2009 systematic review of 82 studies on ownership variation among nursing homes showed that non-profit providers typically delivered higher quality services compared to for-profit providers8; a finding that has since been replicated 6,7,9.
During the COVID-19 pandemic, for-profit providers have been accused of failing their residents by prioritising profits over care, prevention, and caution 10, resulting in reports of higher rates of COVID-19 infections and deaths in for-profit care homes 11,12. Although many of these reports are not peer-reviewed – located, for example, in newspaper publications – there is a growing body of academic research that investigates ownership variation across COVID-19 outcomes, such as outbreaks, infection rates, and mortality 13,14.
The COVID-19 pandemic has tested the capability of not only individual social care providers, but also the commissioning system in which they operate. It is known that funding for adult social care has been characterised by years of austerity 15 at the expense of staffing, quality, and support, which may have been detrimental to the ability of care homes to cope with the pandemic 16. As such, the experiences of the adult social care sector during the pandemic must not be forgotten, and the growing body of research on ownership variation across COVID-19 outcomes offers an important opportunity to revisit the impact of outsourcing social care services to FP providers.
Research Question/objectives
The aim of this living systematic review is to identify, appraise, and synthesise the available research on ownership variation in outbreaks and infections across for-profit, public, and non-profit care homes for older people, and to update our findings as new research becomes available. A review protocol was registered prospectively on OSF 17 and on Prospero (CRD42020218673).
Methods
COVID-19 related research is published at a high frequency and the time between submission and publication is substantially shorter than for non-COVID-19 research18, which makes it particularly important that this rapidly growing body of evidence is critically appraised and systematically synthesised regularly. To ensure that this review will represent a recent and relevant synthesis of the available evidence on the topic, it will be conducted as a living systematic review, which is defined as “a systematic review that is continually updated, incorporating relevant new evidence as it becomes available” 19. Specifically, we plan to update our results every 3 months for the first year of publication and twice in the second year. An update will entail (1) running the below search string for the time period, and (2) forward and backward citation tracking of included studies. Any studies added throughout this process will be incorporated in the review and the synthesis, results, and interpretation will be adapted accordingly.
Inclusion criteria
We assessed study eligibility based on three criteria. First, studies must investigate variation in COVID-19 outbreaks, infection rates, and/or excess or COVID-19 related mortalities or outcomes related to personal protective equipment (PPE) use and availability, staff shortage, preparedness, and infection and mortality among staff and visitors. Second, studies must investigate variation in any of the above outcomes across ownership categories, which is conventionally understood as for-profit, non-profit, and public providers. However, ‘ownership’ is often understood and interpreted in different ways and in the below section we specify how we define this concept. Third, studies must employ an observational research design, including, but not limited to, cross-sectional studies, cohort studies, and secondary analyses of registry data. Both published and unpublished manuscripts (e.g., preprints and reports) were eligible for inclusion.
A note on ownership terminology
The literature on ownership variation is characterised by inconsistent and variable terminology, and terms such as ‘for-profit’, ‘non-profit’, ‘private’, ‘chain affiliated’, and ‘public’ are rarely clearly defined. In consequence, original ownership archetypes such as for-profit and non-profit are often merged as ‘private’ providers, while public (e.g. government or local authority providers) and non-profit (registered private not-for-profit) providers are often categorised as ‘non-profits’. Table 1 displays our definition of different ownership terms and outlines how these overlap.
Ownership categorisation and overlap.
Important nuances are lost by not defining ownership consistently, which is a challenge in this review (as it has been in previous reviews 8). Yet, since the objective of this review is to appraise and synthesise research on ownership variation, we consider any definition of ownership. However, all categorisations are scrutinised, and our results incorporate different conceptualisations where applicable.
Searches and screening process
Before developing our search strategy, we hand-searched Google Scholar and preprint repositories for relevant articles and performed citation searches on all relevant articles. This preliminary sample of includable studies were used to design our search strategy and to test its specificity. The full search strategy can be found in the supplementary material. Our search string was implemented in the following databases: ABI/INFORM Global, Coronavirus Research Database, Criminology Collection International (criminal justice database and NCJRS abstracts database), International Bibliography of the Social Sciences (IBSS), Politics Collection (PAIS index, policy file index, political science database, and worldwide political science abstracts), Social Science Database, Sociology Collection (applied social science index and abstracts, sociological abstracts, and sociology database) via ProQuest. We searched Embase, Global Health, Medline, and PsycINFO via Ovid. Two authors (ABM and MDE) double-screened all search results using Rayyan and subsequently reviewed the full text versions of potentially eligible studies. We undertook a citation search of all studies included as a result of the screening and full-text reviewing.
Data extraction, (selection and coding)
Descriptive information on the citation details (author, title, journal) and study characteristics (research design, analysis, sample details) were retrieved from all included articles. We also extracted detailed information on the data (e.g., country, source, and period of data coverage) and the outcome and exposure variables (e.g., definition, operationalisation, and cut-offs). The results related to ownership variation across COVID-19 related outcomes and accompanying interpretations were extracted for all studies. The results were extracted by ABM with feedback from BV and MDE.
Risk of bias assessment
Risk of bias (RoB) was assessed using the COSMOS-E guidance 20. We employed this guidance rather than, for example, ROBINS-I for non-randomised intervention studies, as it is specifically designed for systematically reviewing observational and correlational research. Specifically, we evaluated the following bias domains: confounding, selection bias, and information bias. In line with the consistent recommendation to avoid quantitative scoring of risk domains20,21, all RoB assessments were based on the qualitative subjective assessment of the reviewers, which were decided through discussion and in consensus. All assessments were conducted with the focus of our review (variation in COVID-19 outcomes across ownership) in mind, and the assessments may thus not represent the risk of bias across other investigated associations and outcomes. The overall RoB assessment for each study was based on the lowest assessment in any bias domain. We did not exclude studies based on the RoB assessments.
Data synthesis
Due to a high degree of heterogeneity among the included studies in terms of differing model specifications, operationalisation of outcomes, inconsistent ownership categorisation of the reference group, and overlapping data, we did not perform a statistical meta-analysis of the included results. This decision was made with attention to the pitfalls of employing statistical methods and assumptions designed for the analysis of very homogenous data or randomised controlled trials of interventions to observational and correlational research 22. If appropriate, we may, however, meta-analyse the results in later versions of this living review. In this version, our synthesis can be described as follows.
First, we narratively summarised key characteristics of the included studies, such as publication type, sample details, ownership categorisation, and data sources. Second, we assessed the risk of bias across all included studies. Third, we developed an overview of the outcomes and model adjustments of all included studies. The model adjustment categories were developed according to aspects known to influence the relationship between ownership and care homes. Fourth, we constructed a harvest plot to graphically illustrate the direction of the results across different outcomes and ownership categories with attention to the model specifications and risk of bias of the studies supporting a particular direction of effect 23. Harvest plots serve as a way to synthesise and describe a heterogenous body of evidence, which cannot be meaningfully synthesised by means of meta-analysis. Last, we thematically analysed the text of all studies that discuss and interpret results relating to ownership.
Results
Search results
Our search strategy yielded 3,845 results, of which 3,514 were left after removal of duplicates. These were double-screened by ABM and MDE, after which 81 were assessed in full-text. Of these, 19 studies were deemed eligible for inclusion. A citation search was then conducted on this sample, in which an additional 9 eligible studies (primarily preprints and government reports) were identified. Twenty-eight studies were thus included in the first version of this review.
Description of studies
The descriptive information on all included research can be found in Table 2. Most of the included studies were peer-reviewed publications (22/28), with 2 government reports 24,25 and 4 preprints or working papers. The unit of analysis across all included studies was care homes. For-profit providers were the largest ownership group across all studies that provided detailed sample information. All except for two studies 26,27 were published in July and after. The majority of studies were conducted in the United States (21/28), followed by Canada (3/28), England (2/28), Scotland (1/28), and France (1/28). Most of the studies were cross-sectional in nature, as only three studies included more than one time-point in their analysis 28–30.
Study and outcome characteristics
The most common way ownership was categorised was to compare for-profit (FP) and non-profit (NP) providers (16/28), usually with FPs as the reference category. Ten studies compared FPs and NPs (11/28), in which NPs included both private NPs and public providers, although this was not always explicitly described. Two studies focused on private equity (PE) providers in their categorisation of ownership 31,32. Eighteen studies also included chain affiliated (CA) providers as a separate covariate to investigate COVID-19 related outcomes.
The most commonly investigated outcome was COVID-19 outbreaks (15/28), followed by COVID-19 related mortality (13/28), and incidence of COVID-19 infections (12/28). Five studies investigated staff access to PPE and/or shortage of PPE. Three studies included investigated COVID-19 related outcomes among care home staff. Most studies investigated multiple COVID-19 related outcomes.
Data sources
Most of the included research merged multiple data sources across COVID-19 outcomes, information on the care facilities, and area characteristics to construct their dataset. Less than 15% of included studies (4/28) collected primary data on the investigated COVID-19 outcomes. The majority of studies used data routinely reported by care homes to public health departments and other government-collected data. Almost a third of the included studies (9/28) used data from the Centers for Medicare & Medicaid Services (CMS), who require American nursing homes to report COVID-19 related data, including confirmed and suspected infections and deaths among residents and staff 50 from May 24. Providers were encouraged, but not required, to retrospectively self-report COVID-19 outcomes before this date. Table A1 in the supplementary material provides an overview of the data sources and the time period of the dependent variables across all included studies.
Risk of bias assessments
The RoB assessments are detailed in Table A2 in the supplementary material. The main concerns related to systematic missing data and selection bias in the included studies. For example, studies that investigated the characteristics of care homes excluded from the CMS dataset (due to missing or incomplete data) found that such facilities were more likely to be FPs and also correlated to many risk factors, such as the ethnicity and socio-economic status of residents (discussed in detail below) 29,41,43,45. This is a potentially serious limitation of the studies using this data (for the purpose of this review), as it suggests that poorly performing FPs may be systematically underrepresented in the sample, which may underestimate the observed effect of ownership on COVID-19 outcomes.
Because of this limitation, all studies using this dataset were downgraded to (at least) moderate risk. For all studies using public and government data, we assumed the risk of information bias to be low.
All assessments of the risk of bias of confounding were based on consideration of factors known to exacerbate the effect of COVID-19 51,52 and the performance of care homes for older people 6,7.
Almost all studies adjusted for the size of the care homes (24/28), and characteristics on quality and staffing were also commonly included. Nineteen studies adjusted for the ethnic composition and 17/28 included information about the socio-economic status of residents. Rurality and/or population density was included in 11/28 of the studies. See Table A3 in the supplementary material for details on the direction of effect and model adjustment of all included studies.
Direction of effects
Figure 2 displays the direction of effect for all included studies across different ownership and outcome categories. Bar height indicates overall risk of bias (with taller bars indicating lower RoB) and colour denotes study context. See Figure A1 in the supplementary material for a harvest plot on the direction of effect across different data sources and Table A4 for details on the confidence of each finding.
Prisma flow diagram.
Harvest plot on the direction of effect across ownership, risk of bias, and study context. Bar height indicates overall risk of bias and colour denotes study context. Positive and negative effects are understood as associations in either directions that are statistically significant at the 5% level. Note that positive effects refer to elevated COVID-19 outcome values. Null effect is understood as differences that are not statistically significant at the 5% level. Outbreaks usually refer to the presence of any COVID-19 infections, except for one study that defined outbreaks as at least two cases 24. COVID-19 incidence usually refers to the cumulative number of COVID-19 infections, but also includes binary outcomes on large outbreaks (e.g. 44,49). COVID-19 mortality usually refers to cumulative cases, except for one study which used a binary indicator of at least 6 deaths 26, and another which analysed a dichotomised outcome of any number of deaths vs no deaths 30.
In most studies investigating outbreaks (11/14), FP status was not associated with a higher risk of outbreak, suggesting that FP providers were not more or less likely to have at least one infected resident. The evidence on COVID-19 incidence was mixed – though most of the studies found FP providers to be associated with a higher number of cumulative COVID-19 infections. This direction of effect (i.e., positive effect) was consistent across multiple contexts, including the US, England and Canada. Null effects were only reported in US studies.
Higher rates of COVID-19 related deaths were consistently associated with FP status in Canadian and US contexts. Five studies using state data in the US and in England did not find a positive effect. However, the study conducted in England only investigated variation in the probability of having at least 1 COVID-19 related death and did not analyse variation in the cumulative numbers of deaths 30. Studies using CMS data usually reported a statistically significant relationship between FP and mortality, which was less frequently identified in studies using data from individual states. Studies conducted in Canada generally found FP status to be associated with a higher incidence of infections and deaths among residents, whereas the direction of effect was mixed in the English studies 27,36,49.
All studies that investigated PPE outcomes found FP status to be positively associated with insufficient access to or shortage of PPE 29,32,35,45. FP status was not consistently associated with staffing shortages.
Chain status was often correlated with a higher likelihood of COVID-19 outbreaks, but not with a higher incidence of infections. Five studies from Canada, England, and the US found chain affiliation to be associated with a higher incidence of COVID-19 deaths, but most studies that included this variable did not identify any statistically significant variation. The one Canadian study that investigated this category found CA to be associated with higher incidence and mortality 36. The two studies conducted in England did not identify any variation in outbreaks and incidence across chain status 49, but one found CA to be associated with higher risk of at least one COVID-19 deaths among residents 30.
Care homes owned by private equity (PE) firms were not found to be consistently associated with worse outcomes than other ownership categories, and one study even found PE providers to be less likely to report PPE shortages and confirmed COVID-19 outbreaks 32.
Evidence on the relationship between ownership variation and risk of infection among staff was scarce, with only three studies investigating this population. In England, FP status was associated with a higher risk of infection among staff 49, but there was no difference for CA providers. The two studies conducted in the US did not identify any statistically significant variation related to FP status, but one study found CA status to be correlated with a higher risk of infection and deaths among staff 32.
Indirect ownership effects
In most studies, ownership was treated as a covariate for model adjustment, and the specific results relating to this variable were not often directly discussed and interpreted. In the studies which discussed the implications of ownership (16/28), the most common theme involved reflections about mediating pathways through which for-profit status indirectly influenced COVID-19 outcomes.
Importantly, FPs were more likely to report PPE shortages 29,31,45, which was interpreted by some to mean that this provider type was less willing to challenge the financial bottom-line by making additional investments, even during a pandemic 13,29. Yet private equity did not report PPE shortages 31,32, raising questions about whether the profit motive is the main driving factor. FP providers were more likely to be affiliated with corporate chains, which was reported by some to mediate the effect of FP status on COVID-19 outcomes. For example, one Canadian study found the effect of FP status to dissipate when controlling for chain status, which was interpreted by the authors to mean that CA mediates the effect of FP status on COVID-19 36. The only study that controlled for union membership among care homes did not find FP status to influence COVID-19 deaths.
FP facilities were also more likely to be larger (in terms of number of served clients)13, to serve minority groups and people on Medicaid 41, to have lower quality ratings and a history of regulatory deficiencies 24,40, and to be more crowded 27. This multicollinearity between ownership and important mediating characteristics may have biased the observed correlations, in that most of these factors are known to influence COVID-19 outcomes. However, a substantial body of studies with low to moderate risk of bias identified a significant difference across ownership, even when controlling for these characteristics 47, indicating a direct effect of ownership independent of mediating variables.
Importantly, this direction of effect is consistent with previous research on care quality and FP ownership 8,53. See Figures A2-A3 in the supplementary material for directed acyclic graphs (DAGs) of the hypothesised relationships between ownership, risk factors, and COVID-19 outcomes.
Discussion
Summary of findings
Our synthesis and critical appraisal of 28 studies suggests that FP ownership is not consistently associated with a higher probability of COVID-19 outbreaks. Yet, there is compelling evidence suggesting that the consequences of outbreaks, in terms of cumulative infections and deaths, may be exacerbated by FP ownership. The finding that FP providers are consistently associated with PPE shortages, may help to explain why these care homes suffer from higher rates of infections and deaths following a COVID-19 outbreak – because these care homes were often ill-prepared and underequipped.
However, there are other factors that may instead be mediating this relationship, as FP status is associated with other risk factors, such as resident ethnicity, socio-economic status, low quality ratings, and facility size.
Chain affiliation was often found to be correlated with higher risk of outbreaks, but was not consistently associated with elevated numbers of infections. There was some evidence demonstrating a higher incidence of COVID-19 deaths among chain affiliated providers. PE ownership was not consistently associated with inferior COVID-19 outcomes.
Implications
The findings of this review highlight the importance of ownership in accounting for poor COVID-19 outcomes across care homes. It is known that the adult social care sector found itself exceedingly exposed in the beginning of the pandemic 54, in large part due to delayed government support and intervention, but also as a result of many years of political and financial neglect 10. With this review, we do not suggest that the challenges faced by care homes during the pandemic can (or should) be understood through the lens of ownership alone (as illustrated in the Figures A2-A3 in the supplementary material). It is clear that providers of all types have faced severe challenges, which cannot be reduced to ownership status. However, outsourcing to for-profit providers has become the status-quo in many care markets, often based on the rationale that open market competition will optimise the functioning of these facilities. This claim has been extensively criticised and is not supported by empirical work 6–8,53,35. This review adds to this evidence base by systematically appraising and synthesising the available research on how the consequences of the COVID-19 pandemic in care homes has varied by their ownership status. Although our results represent multiple national settings, most of the included research was conducted the US due to the availability of the national CMS dataset. Efforts are currently being made in the UK to create a similar type of systematic, live, and linked dataset on care homes 56, which is an important endeavour if the consequences of this pandemic are to be understood and addressed going forward.
Limitations
Our findings should be interpreted in light of certain caveats, most of which relate to the characteristics of the included studies. First, most studies were conducted in the US and Canada and relied on CMS data, which means that (1) the results primarily relate to North America, and (2) there may be overlap in the analysed data across certain studies. Second, the body of included research was too heterogenous to be meaningfully meta-analysed and this version of the review thus represents a critical appraisal and narrative synthesis conducted in line with best practice standards 20,57. Last, throughout our risk of bias assessments, we assumed that the reporting of COVID-19 outcomes was not systematically related to ownership. However, there is some suggestive evidence of a longer turnaround period for resident test results among for-profit providers 58, which if generally true, may bias the effect of FP status towards the null due to underreporting.
Conclusion and future research
This review constitutes the first version of a living appraisal and synthesis of evidence on ownership variation across COVID-19 outcomes. It will be updated as new research becomes available, which may change the conclusion of our synthesis. Based on our synthesis of the available research, we find FP status to be a consistent and credible risk factor of higher cumulative COVID-19 infections and deaths. Therefore, ownership among care home providers may be a key regulatable factor that can be addressed to improve health outcomes in vulnerable populations and reduce health disparities. Going forward, we hope future research will incorporate other national contexts and clearly define their ownership categories of interest.
Data Availability
Most of the data represented in available in the manuscript or in the supplementary material. Additional data is available upon request. Our protocol is available online on Prospero (CRD42020218673) and on OSF (osf.io/c8dq9/). The authors affirm that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
Contributors
ABM conceived the idea for the manuscript and designed the study protocol with feedback from BV and MDE. ABM and MDE double-screened and selected the included studies. ABM extracted and analysed all data with feedback from MDE and BV. All authors contributed to quality assessing the included research in duplicate. MDE and ABM developed the visualisations for the paper. ABM wrote the manuscript with feedback from MDE and BV. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Funding
Anders Bach-Mortensen is supported by a research fellowship from the Carlsberg Foundation. The funder was not involved in the research process.
Competing interests
All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare: no competing interests.
Ethical approval
Not applicable.
Data sharing
Most of the data represented in available in the manuscript or in the supplementary material. Additional data is available upon request. Our protocol is available online on Prospero (CRD42020218673) and on OSF (osf.io/c8dq9/). The authors affirm that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
Dissemination to participants and related patient and public communities
Our protocol is available online on Prospero (CRD42020218673) and on OSF (osf.io/c8dq9/).
References
