Study population and data sources

We conducted a population-based retrospective cohort study of individuals who had a positive test result of the novel coronavirus (SARS-CoV-2) reported through the Ontario Laboratories Information System (OLIS) from January 15 to December 31, 2020. Individuals who were not eligible for the Ontario Health Insurance Plan (OHIP) and those who were not residents of Ontario at the beginning of the study period were not included in the database. The OLIS collects testing SARS-COV-2 results based on the polymerase chain reaction (PCR) test processed at all provincial public health laboratories, hospital laboratories, and commercial laboratories in Ontario. We treated the first date of sample collection with a following positive SARS-COV-2 result diagnosed in 2020 as the diagnosing date. The study cohort derived from OLIS was linked to population-based provincial health administrative data to ascertain baseline information on socio-demographic characteristics and chronic conditions, as well as outcomes of interest. These datasets were linked using unique encoded identifiers and analyzed at ICES. ICES is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement.

Ethics consideration

The study has received ethical approval from the Office of Research Ethics at the University of Toronto (#39138). The consent form was not obtained as the data is from ICES, which is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement.

Definitions of outcomes and risk factors

The primary outcome of our study is mortality, which is defined as all-cause death within 30 days after the first positive SARS-COV-2 test. Data on individuals’ mortality was retrieved from the Canadian Institute for Health Information’s Discharge Abstract Database and the Registered Persons Database until January 30, 2021.

We adopted a composite severity outcome, which consists of all-cause deaths within 30 days after the first SARS-COV-2 positive test and hospitalization happened between the period of 14 days prior to and 30 days after the first SARS-COV-2 positive test, as our secondary outcome. Data on hospital admission and discharge was retrieved from the National Ambulatory Care Reporting System, and the Ontario Health Insurance Plan databases until January 30, 2021.

We extracted individuals’ demographic and residential information, including age, sex, residence, income quintile, and whether the individual was resident of a long-term care (LTC) facility 90 days prior to the testing date from the database. Age was grouped as below 50, 50–59, 60–69, 70–79, and above 80 years old. Residence was distinguished between rural and urban. The income variable was neighborhood-based and determined using methods developed by Statistics Canada, where income was adjusted for household size and cost of living across the province so that each dissemination area would have 20% of its population in each income quintile. Quintile five indicated the highest income group while quintile one indicated the lowest. Each individual was assigned with the neighborhood income quintile of the dissemination area which was matched with his/her postal code [20]. LTC residence was included as risk factor because high mortality in LTC facilities was reported in several countries, including Canada [21–23]. Individuals’ pre-existing comorbidities (S1 Appendix) were identified using pre-existing ICES chronic disease cohorts and data from the Discharged Abstract Database, National Ambulatory Care Reporting System, Ontario Health Insurance Plan, and the Ontario Drug Benefit Claims Database. ICES chronic disease cohorts have been derived for several chronic conditions using administrative data algorithms. The majority of these have been validated through chart review and demonstrate high sensitivity and specificity [24, 25]. The comorbidities include asthma, chronic pulmonary obstructive disease (COPD), dementias (including Alzheimer’s and delirium), human immunodeficiency virus (HIV), hypertension, diabetes, chronic kidney disease, cancer, CVD (including cardiac ischemic disease, congestive heart failure, acute ischemic stroke, and hemorrhagic stroke), rheumatoid arthritis, inflammatory bowel disease, liver disease, severe mental illness (including other hospitalized mental illness excluding dementia), and solid organ transplant. A list of the case definition of the comorbidities can be found in S1 Appendix.

Statistical analysis

We first conducted descriptive statistical analyses of risk factors, by calculating proportions, means with standard deviation (SD), and medians with inter-quartile ranges (IQRs) of the variables. To compare the differences in risk factors between the deceased and alive groups, we employed statistical tests such as student-t test, Wilcoxon rank-sum test, Chi-Square test, Fisher’s exact test, or Kruskal-Wallis test when appropriate.

To explore risk factors for mortality, we employed Cox proportional hazards regression model to estimate the hazard ratios (HR) and corresponding 95% confidence intervals (CIs). We calculated the follow-up time for each individual from the date of COVID-19 diagnosis until death, 30 days if alive, or the end of the follow-up date of study (30 January 2021). Since many cases were hospitalized before COVID-19 diagnosis in the first wave, we used logistic regression to estimate the risks of the composite severity outcome, which consists of COVID-19 death and hospitalization, in odds ratio (ORs) and 95%CI. In multiple regression analyses employing either Cox or logistic regression models, we controlled demographic, socio-economic variables, LTC contacts, and all comorbidities. We conducted separate analyses for three comorbidity-related variables (whether having any comorbidity, the number of comorbidities, and types of comorbidity) by fitting them in three models to avoid collinearity. To understand the influence of comorbidities under different age groups, we conducted stratified multivariable Cox regression and logistic regression analyses to examine the associations between comorbidities and the two outcomes in five age strata: individuals below 50, between 50–59, between 60–69, between 70–79, and above 80 years old. We also applied the Kaplan-Meier product limit method to graphically examine the relation between comorbidities and time to mortality. Log-rank test was used to compare the differences in survival functions between levels of comorbidities.

All statistical analyses were performed using R version 4.0.3. All the variables had no missing variables except two, residence type and income quintile, having 0.28% missing values, which were randomly imputed based on the distributions of non-missing values of the variables. For all the statistical tests, a p-value less than 0.05 was considered statistically significant. The study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S2 Appendix). Analyses were planned prior to accessing the complete data prepared by ICES, but no prospective analysis plan was published.

Age stratified analysis

We conducted subgroup analyses using the multivariable Cox proportional hazard regression models to explore the effect of comorbidities on mortality in each age group. As shown in Table 2, the number of comorbidities profoundly elevated mortality risks among individuals in lower age groups. On the other hand, the influence of comorbidities on mortality reduced as age increases. For example, compared with individuals of the same age group without comorbidities, the mortality risk among those having five or more comorbidities was 395.44 (95%CI, 57.93–2699.44, p<0.001), 35.87 (95%CI, 18.42–69.85, p<0.001), and 12.30 (95%CI 2.94–51.46, p<0.001) times in the age groups of below 50, between 50–59, and between 60–69 years respectively. However, the associations were not significant among individuals in their 70s and 80s. Additionally, among individuals under 50 years old, elevated mortality risks were observed, including comorbidities such as HIV (HR = 13.07, 95%CI, 3.19–53.58, p<0.001), severe mental illness (HR = 13.33, 95%CI, 5.00–35.52, p<0.001), rheumatoid arthritis (HR = 6.96, 95%CI, 1.63–29.68, p<0.01), CVD (HR = 6.82, 95%CI, 2.46–18.9, p<0.001), cancer (HR = 6.82, 95%CI, 2.61–17.82, p<0.001), liver disease (HR = 6.05, 95%CI, 1.83–20.00, p<0.001), and diabetes (HR = 2.48, 95%CI, 1.25–4.93, p<0.01). Among all age groups, diabetes and cancer remained a significant risk factor for mortality.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Age groups based analyses regarding associations of demo-socio-economic characters and comorbidities with all-cause 30-day mortality by Cox proportional hazards regressions.

https://doi.org/10.1371/journal.pone.0258154.t002

Similar patterns were identified in the subgroup analysis for the composite severity outcome that the number of comorbidities were associated with elevated risk to disease severity among younger age groups, while it became non-significant to disease severity among individuals of 80 years or older (Table 3). Among individuals under 50 years old, preexisting conditions such as severe mental illness (OR = 8.00, 95%CI 6.33–10.13, p<0.001), solid organ transplant (OR = 6.69, 95%CI 3.43–13.02, p<0.001), chronic kidney disease (OR = 2.83, 95%I 2.15–3.72, p<0.001), liver disease (OR = 2.59, 95%CI, 1.65–4.06, p<0.001), CVD (OR = 2.51, 95%CI, 1.76–3.57, p<0.001), diabetes (OR = 2.08, 95%CI, 1.81–2.38, p<0.001), cancer (OR = 1.82, 95%CI, 1.36–2.44, p<0.001), hypertension (OR = 1.45, 95%CI 1.26–1.66, p<0.001), and asthma (HR = 1.31, 95%CI, 1.16–1.47, p<0.001) were significant predicators for severe outcomes. COPD, dementia and cancer appeared significantly associated with severe outcomes in most age groups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Age groups based analyses regarding associations of demo-socio-economic characters and comorbidities with the composite severity outcome that includes all-cause 30-day mortality and hospitalization by logistic regression models.

https://doi.org/10.1371/journal.pone.0258154.t003

Strengths and limitations

To our knowledge, it is one of the first studies employed near complete population-based data in examining risk factors for COVID-19 mortality and severity. Because the Government of Ontario is the single payer for all physician consultations and hospitalization under the universal health coverage, we were able to link all individuals diagnosed with COVID-19 with their previous medical history and encounters. As our database recorded exact date of mortality, we were able to report hazard ratio through Cox models, which is a more accurate estimate of risks than odds ratios [28]. As shown in S1 Appendix, the ICES chronic disease databases employed multi-database to catch the chronic disease records of all Ontarians. The majority of the chronic disease were have been validated from previous studies demonstrating high sensitivity and specificity [24, 25], so misclassification of comorbidities is minimal.

Our findings should be considered in light of a number of limitations. First, the study was designed to identify risk factors associated with mortality and severity, and the estimates reported here do not reflect any causal effects. Second, our data was collected until December 31 2020 when the new SARS-COV-2 variants of concern (VOC) including B.1.1.7 were only sporadically reported in Ontario, but the B.1.1.7 variant has become predominant in the United Kingdom and Canada at the time of writing [29]. Variant B.1.1.7 has higher transmissibility and the younger population is more likely to be infected [30]; hence, our knowledge base regarding the risk factors for COVID-19 mortality and severity needs to be updated in light of the prevalence of the new VOCs. Third, the ICES database did not record socio-demographic data regarding ethnicity, education, and individual level income, which were found to be associated with COVID-19 outcomes in several studies. Researchers have identified that people in ethnic minority groups or with low social economic status had high risks of COVID-19 infection and death [10, 31]. Variables of ethnicity and other socio-economic status such as education may also interact with each other [32, 33]. Fourth, since the reasons of death were not available in our database, we used the all-cause 30-day mortality as our COVID-19 mortality outcome, which was also used in the previous studies [27, 34]. However, we may underestimate the risk factors as we excluded a small number of COVID-19 individuals who died more than 30 days following their diagnoses but we may also overestimate as this may include death unrelated with COVID-19. Fifth, we defined disease severity as death or hospitalization. In Ontario, Canada, all COVID-19 patients would reach a physician/ nurse practitioner if symptoms persist and would be hospitalized if the case is severe. All Ontarian residents are covered by a tax-funded universal health coverage that removes any financial barriers to access medical care. We understand this may not be the case in other settings as patients who are hospitalized have to pay or be covered by insurance, so hospitalization may not be an appropriate indicator for severity. Sixth, most of the comorbidity information were updated until March 31 2020 except three comorbidities, cardiac heart failure, inflammatory bowel disease and rheumatoid arthritis were updated until March 31, 2019. We may miss cases diagnosed after the date. The ICES chronic disease databases do not contain further details of specific comorbidities. Seventh, although the study population is close to the entire population in Ontario it excluded a very small number of people who lived in Ontario but were not eligible for OHIP. Last but not the least, though OLIS covered almost all laboratories in Ontario, the system may slightly under report COVID-19 cases, especially at the beginning of the pandemic when laboratory testing for SARS-COV-2 was still ramping up.

In conclusion, by analysing data from a large population-based cohort that includes all COVID-19 individuals identified in 2020 in Ontario Canada, we found that solid organ transplant, severe mental illness, dementia, chronic kidney disease, CVD, diabetes, hypertension, cancer and COPD were predictors of mortality and severity. Our study highlights that the number of comorbidities was a strong risk factor for deaths and severe outcomes among the younger COVID-19 individuals. We also found that the impact of comorbidities was more substantial among individuals under 50 years old. Findings of our study suggests that in addition of prioritizing by age, vaccination priority groups should include younger population with multiple comorbidities.