Assessing COVID-19 in Austria with infection surveys

GSEM Professors, Stéphane Guerrier and Maria-Pia Victoria-Feser co-authored an article published in the Journal of American Statistical Association, a top tier publication, along with Christoph Kuzmics (Karl-Franzens-Universität Graz). It explores the combination of infection survey outcomes with case count data to estimate COVID-19 prevalence in Austria in 2020. By combining these two data sources, researchers achieve a significant efficiency gain in prevalence estimation. Notably, smaller infection survey samples can yield the same level of accuracy. Guerrier, Victoria-Feser, and Kuzmics propose specific estimation methods that account for measurement errors and nonrandom sample weighting in infection surveys. Additionally, they introduce an open-source R package named 'pempi' for implementing the proposed estimators and confidence intervals.

Stéphane Guerrier and Maria-Pia Victoria-Feser are financed by the Swiss National Science Foundation as well as Innosuisse.

ABSTRACT

Countries officially record the number of COVID-19 cases based on medical tests of a subset of the population. These case count data obviously suffer from participation bias, and for prevalence estimation, these data are typically discarded in favor of infection surveys, or possibly also completed with auxiliary information. One exception is the series of infection surveys recorded by the Statistics Austria Federal Institute to study the prevalence of COVID-19 in Austria in April, May, and November 2020. In these infection surveys, participants were additionally asked if they were simultaneously recorded as COVID-19 positive in the case count data. In this article, we analyze the benefits of properly combining the outcomes from the infection survey with the case count data, to analyze the prevalence of COVID-19 in Austria in 2020, from which the case ascertainment rate can be deduced. The results show that our approach leads to a significant efficiency gain. Indeed, considerably smaller infection survey samples suffice to obtain the same level of estimation accuracy. Our estimation method can also handle measurement errors due to the sensitivity and specificity of medical testing devices and to the nonrandom sample weighting scheme of the infection survey. The proposed estimators and associated confidence intervals are implemented in the companion open source R package pempi available on the Comprehensive R Archive Network (CRAN). Supplementary materials for this article are available online including a standardized description of the materials available for reproducing the work.

The study is available open access: Assessing COVID-19 Prevalence in Austria with Infection Surveys and Case Count Data as Auxiliary Information

> Click here to view the GSEM faculty’s publications in top-tier journals.