Knowledge Discovery of the Delays Experienced in Reporting COVID-19 Confirmed Positive Cases Using Time to Event Models

Survival analysis techniques model the time to an event where the event of interest traditionally is recovery or death from a disease. The distribution of survival data is generally highly skewed in nature and characteristically can include patients in the study who never experience the event of interest. Such censored patients can be accommodated in survival analysis approaches. During the COVID-19 pandemic, the rapid reporting of positive cases is critical in providing insight to understand the level of infection while also informing policy. In this research, we introduce the very novel application of survival models to the time that suspected COVID-19 patients wait to receive their positive diagnosis. In fact, this paper not only considers the application of survival techniques for the time period from symptom onset to notification of the positive result but also demonstrates the application of survival analysis for multiple time points in the diagnosis pathway. The approach is illustrated using publicly available data for Ontario, Canada for one year of the pandemic beginning in March 2020.