COVID-19 and the Perils of Inferring Epidemiological Parameters from Clinical Data

• Activity Number: 41 - Storytelling on COVID-19 Impact Using Experts' Prior Knowledge and Data from Social Media, Official Clinical Data, Digital Phenotype from Smartphones' Raw Sensor Data, and Emergency Departments
• Type: Invited
• Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
• Sponsor: International Society for Bayesian Analysis (ISBA)
• Author(s): Ernst C Wit*
• Companies: Universita della Svizzera italiana
• Keywords: Covid-19; Infection fatality rate; clinical data; convenience sampling; ascertainment bias; Bayesian inference

Abstract

Knowing the infection fatality ratio (IFR) is of crucial importance for evidence-based epidemic management. Against this background, in an impressive paper, Verity et al. (2020) rapidly assembled case data and used statistical modelling to infer the IFR for COVID-19.

Given the importance of the issues, the necessarily compromised nature of the data at such an early stage in the epidemic and the consequent heavy reliance on modelling assumptions, I present an in-depth statistical review of what has been done, conscious that the circumstances require setting aside the usual standards of statistical rigour. After having identified some of the weaknesses in the analysis, we proposed a crude alternative Bayesian model to estimate the IFR, resulting in lower values later borne out by the facts.

Nevertheless, we believe that it is not possible to model our way out of the deficiencies of clinical data in order to estimate crucial epidemiological parameters. At the early stage of such calamity, there is an urgent need to replace complex models of inadequate clinical data, with simpler models using adequate epidemiological prevalence data based on appropriately designed, random sampling.