Identifiability Analysis of a Mathematical Model for the First Wave of COVID-19 in Türkiye

Abstract

In this work, a structurally identifiable mathematical model is developed to capture the first peak of COVID-19 in Türkiye. The daily numbers of COVID-19 cases, deaths, prevalence in the ICU, and prevalence on ventilation, obtained from the open-access TURCOVID-19 database, during the first peak, are used as observations. Structural identifiability analysis is performed using the open-source software Julia. For parameter estimation, some parameters are fixed based on the literature while the remaining parameters are estimated using the Data2Dynamics software. Our results align well with the observations. Then, a practical identifiability analysis based on the profile likelihood method is conducted to investigate uncertainties in the parameter values. It reveals that three of the model parameters, namely the progression rate of symptomatically infectious individuals to hospital and the transmission rates associated with exposed and symptomatically infectious individuals, are not practically identifiable. This means that the implementation of intervention strategies via this model must be performed carefully.