We found that there was a strong positive and significant correlation between self-reported COVID-like illness (CLI) and the subsequent week’s number of new COVID-19 reported cases, highlighting that the rise in CLI may precede official statistics by at least one week. This demonstrates the utility of syndromic surveillance in predicting near-future disease activity. Furthermore, ONM trends in fever and cough correlated with FluWatchers, another Canadian participatory surveillance system. These findings suggest there is the potential for participatory surveillance data to be useful for understanding and even forecasting COVID-19 trends in Ontario.

These findings build on previous reports of digital crowd-sourced surveillance techniques reported in the literature. Yoneoka et al. and Nomura et al. reported analyses of syndromic data collected through a large-scale (over 350,000 participants) digital surveillance system in Tokyo, Japan for one week. They observed a strong spatial correlation between symptoms and confirmed COVID-19 cases in Tokyo, Japan^12–14. Here we show similar findings of self-reported symptoms tracking with COVID-19 confirmed cases for long-term data collected over several months in an entire Canadian province.

When comparing demographic characteristics between survey respondents and those who received COVID-19 testing, we found significant differences in age and gender. ONM respondents were more likely to be female and aged 40 – 59 years than those being tested for SARS-CoV2 in Ontario. Others have similarly reported that middle-aged females were the group most engaged with influenza participatory surveillance tools⁷. The ability of this survey to collect symptom data from groups other than those being tested allows for more comprehensive population coverage. Early limitations on testing to only the most severe cases caused uncertainties in true reported case counts⁶. Eligibility criteria for testing have changed over the course of the pandemic - most recently, testing was again restricted to those with symptoms, and only available by pre-scheduled appointment, effectively restricting the number of tests available¹⁹. Survey data that differs in demographic characteristics from those that received a test can provide valuable insights and fill knowledge gaps about groups not receiving testing. The addition of symptom surveillance information ensures that the most representative and timely data is used to inform policy decisions.

Participatory surveillance data also demonstrated that an increasing proportion of those reporting CLI, and of those testing positive for SARS-COV-2, were ages <20 and 20-39 years. In April those <40 years made up ∼25% of positive COVID-19 tests. As of October 2020, approximately 50% of positive COVID-19 tests in Ontario were individuals <40 years old. Similarly, approximately 50% of those with a COVID-like illness were <40 years of age at the end of our study observation period. These findings demonstrate the utility of ONM to detect the evolving demographic distribution of COVID-19.

We found that 0.12% – 0.66% of ONM respondents reported being tested for COVID-19. This was similar to the weekly Ontario population testing rate which ranged from 0.04% - 0.28% of the population. ONM respondents consistently reported a higher rate of COVID-19 testing than the Ontario general population. This may relate to ONM respondents taking a more active role in their health. We also found that a large proportion of respondents with COVID-like illness reported not being tested – with a decreasing trend since week 25. Among those who reported CLI, 16.3% reported being tested in the last 2 weeks, which dropped to 7.3% in week 37. In the first wave of the pandemic, we estimated that only 2-9% of those with symptoms had been tested²⁰. Given growing awareness of the transmissibility and health impacts of COVID-19, one can only speculate as to why the testing rate among symptomatic people might be falling in the second wave. It is possible that due to a lack of paid sick leave, individuals may be less inclined to undergo testing due to the implications of a positive test on the inability to engage in employment.

There are several limitations to this study. Rapidly rising COVID-19 case counts are often broadcasted in the media and this can influence both healthcare-seeking behaviour and symptom self-awareness²¹. In the United Kingdom, significant national and local media coverage of COVID-19 cases resulted in observed increases in syndromic signals (both nationally and locally), particularly those associated with the symptoms of COVID-19 reported in the news media²¹. However, some participatory surveillance systems are more sensitive to the impact of media reporting than others – hence, a multi-source surveillance approach can be beneficial to reliably track population symptomatology. Next, the ONM participatory surveillance method relies on access to the internet, which may exclude individuals who are underhoused or experiencing homelessness, those with poor internet or computer access, or limited English literacy. These characteristics are more common among the racialized and marginalized groups who are disproportionately affected by COVID-19²². In addition, an estimated 17.9 – 30.8% of COVID-19 infected individuals are asymptomatic or minimally symptomatic^23,24. The ability of syndromic surveillance methods to accurately predict case numbers will be confounded by changes in the rate of asymptomatic testing. This is because asymptomatic cases contribute to case counts but cannot be detected through syndromic surveillance methods. This study was also limited to Ontario. The generalizability of findings will depend on local uptake of syndromic surveillance tools, as well as on access to timely testing. In locations where testing is less accessible or characterized by long delays for test results, syndromic surveillance data may anticipate testing data by 1-2 weeks. Where testing is broadly accessible, with short delays, we expect present week symptom and case data to be more tightly correlated.

There have been few previous reports of an association between participatory surveillance signals and confirmed case counts for COVID-19^13,14. This is the first such study to be reported in Canada. We found that online volunteer symptom surveys signalled subsequent increases in COVID-19 case counts in Ontario, Canada. Digital surveillance systems such as ONM are low-cost tools that may be helpful in determining the burden of COVID-19 in a community, particularly if there is under-detection of cases through conventional laboratory diagnostic testing. This additional information can be used as a cost-effective resource to guide a healthcare response and guide policy decision-making.