This study utilized secondary data from administrative health databases accessed through the Ontario Health Database Platform (OHDP). The platform permitted access to patient-level linked population-based health administrative data held at the Institute for Clinical Evaluative Sciences (ICES). The ICES data repository consists of record-level, coded, and linkable health data sets that integrate health services data, registries, and surveys. ICES is an independent, non-profit research institute that houses routinely collected health data from Ontario’s publicly funded healthcare system.

All Ontario residents covered by the province’s universal health coverage were identified using the Registered Persons Database (RPDB), which provided age, sex, and demographic information. COVID-19 status was determined using C19INTGR, a comprehensive dataset of PCR COVID-19 diagnostic results in Ontario. Chronic conditions such as asthma, hypertension, diabetes, COPD, and congestive heart failure were identified using ICES-derived cohorts. These datasets were linked using unique encoded identifiers and analyzed at ICES (Supplementary Tables S3, S4). This study followed the Reporting of Studies Conducted Using Strengthening the Reporting of Observational studies in Epidemiology (STROBE) reporting guidelines [19]. The codes for identifying mental health visits are used for mental health system performance measurement in Ontario and elsewhere in Canada [20–24].

We constructed a retrospective cohort of people who were living in Ontario, Canada, had a valid health card, and for whom information on sex and age was available between the period of January 2020 to March 2021 (Figure 1). All those cases who were not Ontario residents or were less than 18 years of age, as well as those who had a standard COVID-19 lab-based PCR test result after the study end period, i.e., beyond 31 March 2021, were excluded. We utilized the National Ambulatory Care Reporting System (NACRS) and the Ontario Health Insurance Plan (OHIP) to identify physician, outpatient visits, and emergency department visits for mental health, and we used the Canadian Institute for Health Information’s Discharge Abstract Database (DAD) and Ontario Mental Health Reporting System (OMHRS) to capture psychiatric hospitalizations. Throughout the study period, COVID-19 testing guidelines were applicable to the general population in Ontario and were not limited to individuals considered at high risk. This helped to maintain a focus on the general population and ensure that our findings accurately reflect the broader impact of the pandemic on Ontario residents.

The exposure in this study was COVID-19 testing by standard lab-based PCR tests in Ontario between January 2020 and March 2021, irrespective of COVID-19 disease status, leading to three groups for comparison: (i) those PCR tested with a positive COVID result; (ii) those PCR tested with a negative COVID result, and (iii) those who were not PCR tested. The event of interest was the time in months until first recorded health services utilization related to anxiety and depression, using DSM-V criteria, ICD-10 codes, and OHIP diagnostic codes during the study period. Individuals who were not PCR tested were selected as the reference group to provide a population-level comparator reflecting mental health–related healthcare utilization among those who did not engage with COVID-19 testing services during the study period. This approach allows for the evaluation of differences associated with testing exposure within the broader population context. However, we acknowledge that untested individuals may differ systematically from those who were tested, particularly in terms of healthcare-seeking behavior, access to care, and underlying health concerns. To mitigate these differences, we applied propensity score matching to balance measured sociodemographic and clinical characteristics between groups. The participants were censored if there was no utilization of anxiety/depression-related health services by the end of the study follow-up period on 31 March 2021. Also, those who died during the study period were censored at the date of death, assuming the event of interest was not experienced first.

Descriptive statistics (e.g., mean, median, range, standard deviation) were calculated for the study sample (N = 6,175,114) (Figure 1). Frequencies for demographic and health service use variables were examined with cross-tabulations across both COVID-19 test-based groups using t-tests and chi-squared statistics (Table 1). Criteria for statistical significance were set at p < 0.05. We followed people with and without COVID-19 standard lab-based PCR tests from January 01, 2020, to March 31, 2021, to estimate survival time. Time was defined as the interval between the COVID-19 PCR test date and an event (i.e., first contact with health services for anxiety and depression during study period).

Cox proportional hazards regression models were used to estimate the association between COVID-19 testing and time to mental health–related healthcare utilization for anxiety and depression, with results reported as hazard ratios (HRs) and 95% confidence intervals (CIs) for both the overall cohort and the propensity score–matched sample. Kaplan–Meier methods were used to estimate time-to-event functions, and differences between exposure groups were assessed using the log-rank test.

To address potential confounding by indication, propensity scores were calculated using age and sex and applied through matching with replacement, resulting in a matched sample of 664,233 individuals. This approach matched each exposed individual to the nearest comparator based on propensity score, allowing reuse of controls to improve balance. Covariates were selected a priori based on clinical relevance and prior literature, including sociodemographic characteristics, comorbidities, and measures of healthcare utilization, and were incorporated into both regression and propensity score models to enhance comparability between groups. All analyses were conducted using SAS Enterprise Guide 7.1 (SAS Institute, Cary, NC).

This study was approved by the University of Waterloo research ethics board (protocol reference no. 43910), and data use was authorized under section 45 of Ontario’s Personal Health Information Protection Act. ICES is a prescribed entity under section 45 of Ontario’s Personal Health Information Protection Act, which authorizes ICES to collect personal health information, without consent, for the purpose of health system evaluation and improvement.

This study examined differences in first recorded mental health service utilization for anxiety and depression between individuals who did and did not undergo COVID-19 PCR testing in a large population-based cohort during the pandemic. We observed higher utilization among individuals who underwent testing, reflecting underlying differences between tested and untested populations rather than a direct effect of testing itself, with adjusted hazard ratios of 6.37 (95% CI 6.25–6.50) for PCR-positive individuals and 5.91 (95% CI 5.87–5.95) for PCR-negative individuals. Higher utilization was particularly evident among younger individuals, females, and those residing in rural areas, with additional socioeconomic gradients indicating greater use among more deprived populations. The magnitude and similarity of effect estimates across PCR-positive and PCR-negative groups suggest that testing functions primarily as a proxy for underlying vulnerability and healthcare-seeking behavior rather than a causal determinant of mental health outcomes.

The observed patterns of mental health–related healthcare utilization may be consistent with broader reports from the pandemic; however, these findings cannot distinguish between true increases in mental health burden and differences in healthcare-seeking behavior. The increased mental health–related healthcare utilization observed in this study is consistent with evidence on the psychological impact of the COVID-19 pandemic, driven by factors such as uncertainty, social isolation, and economic disruption [21, 22]. Higher risks among younger individuals and females align with prior studies reporting disproportionate mental health impacts in these groups [28–32].

Social determinants of health play a key role in shaping these outcomes. Individuals in lower-income groups experienced greater mental health challenges, consistent with evidence linking financial stress, job loss, and poverty to anxiety and depression [33–35]. Similarly, we observed higher risks among rural residents, which may reflect barriers to care, social isolation, and limited access to mental health services [34, 36].

Chronic conditions, including asthma, hypertension, and congestive heart failure, were also associated with increased mental health service utilization. These conditions, identified as risk factors for severe COVID-19 illness, may contribute to heightened psychological distress, further exacerbated by disruptions in routine care during the pandemic [37, 38].

Overall, these findings highlight the disproportionate burden of mental health service utilization among socioeconomically and clinically vulnerable populations during the pandemic [37, 39].

This study characterizes patterns of mental health–related healthcare utilization among individuals who engaged with COVID-19 testing during the pandemic rather than establishing a causal relationship between testing and mental health outcomes. The findings indicate higher utilization among tested individuals; however, the magnitude of these associations is unusually large for observational mental health research and is unlikely to represent causal effects. Instead, these patterns are most plausibly explained by substantial residual confounding, including underlying differences in healthcare-seeking behavior and psychosocial vulnerability between individuals who seek testing and those who do not. The similarity of effect estimates across PCR-positive and PCR-negative groups further supports the interpretation that testing functions primarily as a proxy for underlying vulnerability rather than a direct determinant of mental health outcomes. Accordingly, these results should be interpreted as descriptive of healthcare utilization patterns during the pandemic rather than evidence of a direct psychological impact of COVID-19 testing.

Individuals undergoing testing may differ systematically in ways that predispose them to higher mental health service use, including greater baseline health anxiety, increased psychological vulnerability, and a higher propensity to engage with healthcare services [40]. These factors may contribute to residual confounding, even after adjustment for measured variables. Surveillance bias is also a likely contributor, as individuals interacting with the healthcare system through testing may be more likely to have mental health conditions identified and recorded. The similar effect estimates observed in both PCR-positive and PCR-negative groups further support the interpretation that testing itself is not the primary driver, but rather a marker of shared underlying characteristics among those who access testing [41]. Reverse causality must also be considered, as individuals with pre-existing anxiety or heightened concern about COVID-19 may have been more likely to seek testing.

Taken together, these findings suggest that COVID-19 testing serves as an indicator of a population with elevated underlying risk of mental health service utilization, rather than a direct causal determinant of mental health outcomes.

The observed disparities in mental health–related healthcare utilization among socioeconomically disadvantaged populations point to important inequities in patterns of healthcare engagement during the pandemic. However, these findings should not be interpreted as evidence of greater underlying mental health need, as they may reflect differences in healthcare access, utilization, and help-seeking behavior rather than true differences in burden. Instead, the results are best understood as identifying populations with higher levels of interaction with the healthcare system during this period.

While causal inferences cannot be drawn, these patterns may still be useful for public health planning by highlighting groups that were more frequently represented in mental health service use during the pandemic. Monitoring trends in healthcare utilization can support resource planning and help ensure that services are responsive to populations engaging with care. However, caution is required in translating these findings into policy, as observed differences may not correspond to underlying need.

Future research should prioritize disentangling the roles of psychosocial factors, healthcare-seeking behavior, and structural inequalities in shaping observed patterns of mental health service utilization [42]. The findings highlight the importance of interpreting administrative healthcare data within the broader context of social and behavioral responses to the pandemic. Strengthening mental health systems and incorporating mental health considerations into pandemic preparedness remain important goals, but should be informed by evidence that distinguishes true need from patterns of healthcare engagement.

To our knowledge, this is one of the largest population-based studies examining the association between COVID-19 testing and mental health–related healthcare utilization in Ontario. The use of linked administrative health data within a universal healthcare system allowed for comprehensive capture of physician visits, emergency department encounters, and psychiatric hospitalizations, enhancing the generalizability of the findings. Additionally, the application of propensity score matching improved comparability between tested and untested individuals by balancing measured sociodemographic and clinical characteristics.

However, several important limitations should be considered when interpreting these findings. The outcome reflects the first recorded use of mental health–related services during the study period rather than incident diagnoses of anxiety or depression. Accordingly, the results represent patterns of healthcare utilization, which may include both new-onset and pre-existing but previously unrecorded conditions, rather than true incidence of mental illness.

Moreover, despite the use of propensity score matching, residual confounding remains likely, particularly due to unmeasured psychosocial factors. Variables such as employment status, income loss, social isolation, media exposure, baseline anxiety traits, and individual propensity for healthcare-seeking were not available in administrative data. These factors are strongly associated with both the likelihood of undergoing COVID-19 testing and the use of mental health services. Individuals experiencing greater psychological distress or socioeconomic disruption during the pandemic may have been both more likely to seek testing and more likely to access care. As a result, the observed associations may be inflated, reflecting underlying vulnerability rather than the effect of testing itself.

The similar magnitude of associations in both PCR-positive and PCR-negative groups further supports this interpretation, suggesting that testing is unlikely to be a direct causal driver of increased mental health service utilization. Instead, testing may act as a marker of heightened vulnerability, capturing individuals with greater health anxiety, perceived risk, or engagement with the healthcare system. Effect sizes of this magnitude are unlikely to be explained by measured confounders alone and likely reflect selection processes and unmeasured differences between tested and untested populations.

The administrative data captures only individuals who interact with the healthcare system, potentially excluding those with untreated or unrecognized mental health conditions. This may introduce detection (surveillance) bias, as individuals who seek testing are also more likely to have mental health concerns identified and recorded.

Additional limitations include the exclusion of individuals with missing demographic data, limited information on certain populations (e.g., immigrants and Indigenous communities), and the inability to assess severity or chronicity of mental health conditions. Finally, the study period did not extend into later phases of the pandemic, which may limit the assessment of longer-term mental health impacts.

Our study identifies higher mental health–related healthcare utilization among individuals who underwent COVID-19 PCR testing for anxiety and depression. These patterns reflect differences between tested and untested populations and should not be interpreted as evidence of a direct psychological impact of testing. The similar magnitude of associations observed in PCR-positive and PCR-negative individuals strongly suggests that testing functions primarily as a proxy for underlying vulnerability and healthcare-seeking behavior rather than a causal determinant of mental health outcomes. These findings pertain to patterns of healthcare utilization rather than new-onset mental illness. As such, the results are best understood as descriptive of healthcare engagement during the pandemic, highlighting the importance of interpreting administrative data within the context of underlying differences in access, utilization, and psychosocial factors. Incorporating mental health considerations into pandemic preparedness remains important but should be guided by evidence that distinguishes true need from patterns of healthcare use.