Our study estimated healthcare costs attributable to confirmed COVID-19 in Ontario, Canada, over a 360-day period post-index, stratified by SDH.

Costs followed a typical U-shaped pattern (24): high during the acute phase, declining during the post-acute phase, and rising again during terminal care. Acute phase costs were the highest, consistent with the intensive resource demands of moderate to severe disease requiring hospitalization, ICU admission, and in some cases, mechanical ventilation (25).

Terminal phase costs for individuals with confirmed COVID-19 who died within 60 days of index (early deaths) were substantially higher than for those who died after 60 days (late deaths). Higher costs for early deaths likely reflect the use of resource-intensive interventions (e.g., mechanical ventilation and intensive care) during the acute phase of moderate to severe COVID-19 (26). In contrast, individuals with late deaths (>60 days) had longer hospital stays (median: 239 days [IQR: 55–627]), with hospitalization costs approximately 56.5% of those for early deaths, but incurred higher attributable home care and complex care costs, reflecting a shift to less costly care settings.

Compared to previous Ontario studies, our attributable mean cost estimate for the acute phase $244 was substantially lower than the $ 4,491 reported by Tsui et al (7) but similar to the data reported by Sander et al. $240 (9). The discrepancies between our findings and Tsui et al. could be due to differences in the costing window and the study time period. Tsui et al. estimated short-term COVID-19 healthcare costs using a 0–120-day costing window around the index date, based on cases from the first pandemic wave (January–June 2020). In contrast, both our study and Sander et al. used a much longer 360-day follow-up period with detailed 10-day costing intervals, capturing costs across multiple pandemic waves and enabling the inclusion of long-COVID and longer-term health-system impacts (7,9). Similar to Sander et al.(9), we observed sustained healthcare costs up to 360 days post-diagnosis, indicating a significant long-term burden of COVID-19. This aligns with evidence of persistent health issues associated with COVID-19, commonly referred to as post-COVID-19 condition, which lasts for more than 12 weeks after the initial infection (27). However, unlike Sander et al., which focused on broader cost implications without stratifying by social determinants (9), our stratified analysis provides novel insights into how SDH were associated with variations in healthcare costs attributable to COVID-19. We found that sex, age, immigration status, income, and household crowding were associated with variation in phase-specific attributable costs. These findings align with Canadian epidemiological studies reporting higher rates of severe disease and mortality among these groups during the COVID-19 pandemic (13).

Multiple factors may explain the higher per-person healthcare costs observed among residents of low-income neighbourhoods. Individuals living in these areas often experience structural barriers to timely and effective healthcare, including longer travel times to healthcare facilities, fewer available healthcare providers, and limited access to preventive services (28). These barriers can potentially lead to delayed care with disease progressing to moderate or severe phase at the time of presentation, increasing reliance on acute-phase services such as emergency departments, hospitalizations, and ICU services (7,8). From our findings, hospitalization was the largest contributor to acute-phase costs among marginalized populations, including individuals living in low-income neighbourhoods. Consistent with our findings, a recent meta-analysis including over two million individuals reported higher odds of hospitalization among socioeconomically disadvantaged populations compared with advantaged groups during the COVID-19 pandemic (29).

Our study has some limitations. First, the C19INTGR dataset likely underestimates COVID-19 infections in Ontario due to incomplete testing, particularly during the early stages of the pandemic when testing was restricted to symptomatic individuals meeting specific criteria (30). As a result, the dataset potentially over-represents individuals with moderate to severe COVID-19 infections, who were more likely to be tested and incur higher healthcare costs, while those with asymptomatic or mild infections are underrepresented. Supporting this, a study conducted in Toronto, Ontario estimated that for each laboratory-confirmed symptomatic case between March and May 2020, there were approximately 18 undetected community infections (31), most of which were likely asymptomatic and required minimal to no healthcare services. Thus, our estimated attributable COVID-19 costs may be an overestimate. It is important to note that testing patterns varied by socioeconomic status: early in the pandemic, individuals in high-income neighbourhoods were more likely to be tested (4); however, testing later increased in low-income neighbourhoods due to higher infection rates and targeted public health efforts (4). This shift may have influenced our findings, as the initially higher testing rates in high-income neighbourhoods could have underestimated cost differences, while the subsequent increase in testing in low-income neighbourhoods could have contributed to greater healthcare use and costs observed in these populations. Second, our findings reflect the first pandemic year (2020) with limited preventive or treatment options and higher hospitalization rates, potentially overestimating costs compared to later waves characterized by vaccine and treatment availability, and milder variants. Furthermore, the area-level SDH variables (e.g., income, household crowding index, percentage of essential workers) were measured at the neighbourhood level using data at the DA level. While DA-level data serve as useful proxies in the absence of individual-level SDH data, they may not capture within-area and individual-level variability and could introduce bias. Therefore, our findings reflect area-level associations and should not be interpreted as individual-level effects. Lastly, while positive pre-diagnosis costs may reflect residual confounding bias, healthcare visits often preceded testing due to constrained testing capacity in 2020, aligning with early-pandemic healthcare-seeking patterns.

Our study has several strengths. We provide comprehensive estimates of the health and healthcare burden of confirmed COVID-19, stratified by key area-level SDH. Additionally, we matched exposed individuals to a historical unexposed cohort to account for healthcare delivery changes (e.g., surgery cancellations) in Ontario from May 2020 to February 2022, minimizing overestimation of COVID-19-attributable costs (14). We used a phase-of-care approach to mitigate the effect of censoring during the study follow-up period (32). Lastly, our phases of care periods and their associated costs can be mapped as health states in decision-analytic models to inform economic evaluations of COVID-19 interventions.

Our study highlights the substantial healthcare costs associated with COVID-19, particularly during acute and terminal phases, as well as the ongoing costs up to 360 days post-infection, indicating long-term sequelae. Our stratified analysis further supports the need for structural policy interventions that address the social and environmental conditions contributing to higher healthcare costs in marginalized communities. For example, enhancing healthcare infrastructure in low-income and crowded housing neighborhoods may reduce delays in care and prevent disease progression. Incorporating social risk indicators (e.g., income, housing density) into resource allocation frameworks can also help ensure that future health system planning and emergency responses are more equitable.