Risk Factors Associated With Mortality Among Residents With Coronavirus Disease 2019 (COVID-19) in Long-term Care Facilities in Ontario, Canada

IMPORTANCE The coronavirus disease 2019 (COVID-19) pandemic has been particularly severe among individuals residing in long-term care (LTC) facilities. As of April 10, 2020, half of Canada’s COVID-19 deaths had occurred in LTC facilities. OBJECTIVE To better understand trends and risk factors associated with COVID-19 death in LTC facilities in Ontario, Canada. DESIGN, SETTING, AND PARTICIPANTS This cohort study of 627 LTC facilities included 269 total individuals who died of COVID-19 in Ontario to April 11, 2020, and 83 individuals who died of COVID-19 in Ontario LTC facilities to April 7, 2020. Because population denominators were not available for LTC residents, they were approximated as the total number of LTC facility beds in Ontario (79 498), assuming complete occupancy. EXPOSURES Confirmed or suspected COVID-19 outbreaks; confirmed COVID-19 infection among residents and staff, diagnosed by real-time polymerase chain reaction testing. MAIN OUTCOMES AND MEASURES COVID-19–specific mortality incidence rate ratios (IRRs) for LTC residents were calculated with community-living Ontarians older than 69 years as the comparator group. Count-based regression methods were used to model temporal trends and to identify associations of infection risk among staff and residents with subsequent LTC resident death. Model-derived IRRs for COVID-19–specific mortality were generated through bootstrap resampling (1000 replicates) to generate median and 95% credible intervals for IRR over time. RESULTS Of 627 LTC facilities, 272 (43.4%) reported COVID-19 infection in residents or staff. Of 1 731 315 total individuals older than 69 years living in Ontario during the study period, 229 (<0.1%) died; of 79 498 potential residents in LTC facilities, 83 (0.1%) died. The IRR for COVID-19–related death in LTC residents was 13.1 (95% CI, 9.9-17.3) compared with community-living adults older than 69 years. The IRR increased sharply over time and was 87.3 (95% credible interval, 6.4-769.8) by April 11, 2020. Infection among LTC staff was associated with death among residents with a 6-day lag (eg, adjusted IRR for death per infected staff member, 1.17; 95% CI, 1.11-1.26). CONCLUSIONS AND RELEVANCE In this cohort study of COVID-19–related deaths during the pandemic in Ontario, Canada, mortality risk was concentrated in LTC residents and increased during a short period. Early identification of risk requires a focus on testing, providing personal protective equipment to staff, and restructuring the LTC workforce to prevent the movement of COVID-19 between facilities. JAMA Network Open. 2020;3(7):e2015957. doi:10.1001/jamanetworkopen.2020.15957 Key Points Question How does the risk of death from coronavirus disease 2019 (COVID-19) among residents of longterm care (LTC) homes compare with that among the general population? Findings In this cohort study of 627 LTC facilities, the incidence rate ratio for COVID-19–related death among LTC residents was 13 times higher than that among community-living adults older than 69 years. Meaning In this study, the risk of COVID-19–related death was elevated among LTC residents, highlighting the need for improved infection control, widespread testing, access to personal protective equipment, and other supports to protect this vulnerable population. Author affiliations and article information are listed at the end of this article. Open Access. This is an open access article distributed under the terms of the CC-BY License. JAMA Network Open. 2020;3(7):e2015957. doi:10.1001/jamanetworkopen.2020.15957 (Reprinted) July 22, 2020 1/7


Introduction
Communicable diseases do not respect boundaries, but they also do not affect all members of our society equally. Since the initial recognition of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in January 2020, it has become starkly apparent that the virus disproportionately affects older individuals and those with more comorbidities, among whom mortality is highest. 1 In Canada, we had an early indication of the particular susceptibility of those living in long-term care (LTC) facilities. The first confirmed death in the country, which occurred on March 9, 2020, was a resident of a North Vancouver LTC home. 2 Since then, outbreaks in LTC facilities have been identified across Canada. As of April 9, 2020, nearly 50% (198 of 401 [49.4%]) of all deaths attributable to COVID-19 in the country have occurred among LTC residents. 3 In addition to the age and comorbidity profiles of residents, other factors make LTC facilities especially susceptible to infectious disease outbreaks. 4 These include a lack of access to testing and personal protective equipment, the close quarters of residents, the difficulty of maintaining physical distancing among mobile patients with dementia, and a precariously employed workforce that can transmit the virus across LTC sites. [5][6][7] Recognizing and quantifying the scope of the issue and identifying which resident and worker factors may contribute to sudden increases in deaths is an important first step toward ensuring that comprehensive policies and response measures are in place to protect residents and the health care workers who provide essential services to them. We sought to evaluate the risk of death among residents of LTC facilities and to identify risk factors associated with mortality using data from the province of Ontario.

Methods
Data for this study were obtained from the Ontario Ministry of Health and Long-term Care as part of the province's emergency modeling table. The study was approved by the Research Ethics Board of the University of Toronto. The data were deidentified, and a waiver of informed consent was granted because the data were collected for public health surveillance purposes. The data included the following: (1) age and date of death of Ontario residents who tested positive for COVID-19 and (2) cumulative death and positive COVID-19 case counts by date among LTC residents and staff.
Confirmed cases were diagnosed by real-time polymerase chain reaction testing. 8 The available time series for LTC facilities covered the 10-day period from March 29 to April 7, 2020, inclusive; however, deaths before March 29 were included as part of the cumulative counts. The provincial death data included the period from March 3 to April 11, 2020. Daily deaths and cases were estimated by taking the difference of cumulative cases and deaths on sequential days.
Population denominators for deaths among non-LTC residents were derived from Statistics Canada estimates for the appropriate age groups. Population denominators were not available for LTC facilities and were approximated as the total number of facility beds in Ontario (ie, 79 498) assuming complete occupancy. Age data for deaths reported in LTC facilities were not available, but data show that approximately 93% of residents in LTC facilities are aged 65 years or older. 9 Although the combination of aggregate and individual-level data made our cohort atypical, we have adhered as closely as possible to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. 10

Statistical Analysis
We compared differences in characteristics between LTC facilities that reported cases and those not reporting cases. Differences in proportions were tested by χ 2 tests, and differences in bed numbers were tested with the Wilcoxon rank sum test (α = .05).
In our main analysis, we estimated incidence rate ratios (IRRs) for COVID-19 deaths in the LTC resident population compared with deaths in community-living adults in Ontario older than 69 years.

JAMA Network Open | Infectious Diseases
Risk Factors and Mortality Among Residents With COVID-19 in Long-term Care Facilities In sensitivity analyses, we compared outcomes in LTC residents with COVID-19 deaths in the entire Ontario population, the population older than 59 years, and the population older than 79 years.
Cumulative deaths at the end of the respective time series were considered, with risk in both LTC residents and non-LTC residents considered to have begun on March 3, 2020. However, person-time denominators were adjusted for the fact that the LTC facility database ended 4 days before the provincial death database (ie, 36 days in the LTC population and 40 days in the general Ontario population). Statistical significance was evaluated with a χ 2 test (α = .05).
We also created a negative binomial regression model to evaluate the change in IRR among LTC residents compared with the non-LTC population from March 29 to April 7, 2020. Negative binomial regression was performed as a deviance statistic test and suggested that the data were not Poisson distributed. Non-LTC residents older than 69 years were used as the comparator population, and model offsets were the population denominators described previously. The model included 3 covariates, as follows: LTC residence, time (centered on April 3, 2020), and an interaction term of LTC residence (ie, yes/no) multiplied by time. Model estimations were generated both forward and backward to generate graphical representations of how risk in the LTC population has changed over time. Model-derived IRRs for COVID-19-specific mortality were generated through bootstrap resampling (1000 replicates) to generate medians and 95% credible intervals (CrIs) for IRRs over time.
Lastly, we created Poisson regression models that evaluated risk of death within LTC facilities as a function of the number of residents with laboratory-confirmed infection as well as staff with confirmed infection at lags of 0 to 7 days, adjusting for date. Given that our model could not accommodate multiple simultaneous lags, we constructed 8 models, each of which included cumulative infections among resident and staff at identical lags (0-7 days). Deviance statistics indicated that counts were Poisson distributed. We used the number of LTC beds as model offsets, and confidence intervals were adjusted for clustering by LTC facility. All analyses were conducted in Stata version 14 (StataCorp). The figures were made using ggplot in R version 1.2.1335 (R Project for Statistical Computing). Statistical significance was evaluated with 2-sided z scores set at α = .05.

Results
A total of 627 LTC facilities were included in the provincial data set; of these, 272 (43.4%) were identified as having either confirmed or suspected COVID-19 infection in residents or staff. No significant differences between LTC facilities with and without confirmed COVID-19 infections were seen in number of licensed beds, operator type (eg, for profit vs nonprofit), or geographic location in Ontario ( Table 1).
Projected trends in risk of COVID-19-related death among LTC residents were generated using model estimations that incorporated time × LTC resident status as an interaction term (Figure 1) In analyses focused on risk of death within LTC facilities, we found that lagged infections among staff were associated with death among residents (Figure 2) and were significant at all lags (0 to 7 days) after adjustment for date and numbers of residents with infection. The strongest associations were seen with staff with infection at a 2-day lag (relative increase in risk of death per staff member with infection, 1.20; 95% CI, 1.14-1.26) and a 6-day lag (relative increase in risk of death per staff member with infection, 1.17; 95% CI, 1.11-1.26). In contrast, the association between infection in residents and subsequent death was variable and far weaker than the associations seen for staff. It was statistically significant only at a 0-day lag (increased risk of death per infected resident, 1.08; 95% CI, 1.01-1.15).

Discussion
In this analysis, we documented the rapid spread of COVID-19 through Ontario's LTC system, with a marked increase in risk of death among older residents with frailty during a brief period from late March to early April 2020. Issues such as crowding, use of communal space, low staffing ratios, and high care needs (with resultant high density of physical contact between residents and staff) have   long been recognized as key drivers of susceptibility to outbreaks in the LTC facility setting. 4,7 In the context of COVID-19, this susceptibility has proven particularly deadly, with (as we demonstrate here) an incidence of mortality more than 13 times greater than that seen in community-living adults older than 69 years during a similar period, with relative risk of death rising sharply over time.
We also found that documented infection in facility staff, as opposed to residents, is a strong identifiable risk factor for mortality in residents, with temporality suggesting that residents are infected by staff and not vice versa. Although it might be argued that the limited nature of testing and the tendency to test staff and residents for COVID-19 after a resident dies might lead to spurious associations between identified infections and deaths, such spurious associations might result in equivalent effect sizes between residents and staff, rather than the divergent effect sizes observed here. The greater mobility and connectedness of staff, compared with residents, lends biological plausibility to this association. 11 Transmission of infection is not the only mechanism by which infection in staff could result in increased mortality in an older population.

Limitations
Like any observational study, this study has limitations, including possible incompleteness of data collected rapidly during an outbreak, inconsistency in testing across Ontario, and absence of individual-level data on LTC facility infections and deaths. We have been unable to explicitly structure autocorrelation in our time series owing to effects on standard errors resulting from the limited size of our data set. We regard our outcome of interest, ie, death from COVID-19 among residents of LTC facilities, to be less likely misclassified than nonfatal infection among staff and residents. If misclassification of infection status in these individuals occurs at random, that would likely mean the associations reported here are lower-bound effect sizes. If underidentification of both fatal and nonfatal infections among residents and staff are clustered by home, that would result in association estimates that are biased upward. The temporality in the associations we observed provides a degree of reassurance in this regard.

Conclusions
This study documented that the rapid movement of COVID-19 through Ontario's LTC facility system has resulted in a marked surge in mortality in that population relative to community-living older adults. We found evidence that associates mortality with infection among LTC staff, highlighting the urgent need for improved infection control, more widespread testing, access to personal protective equipment, and economic protections and support for those who do this important work.