Selection of long COVID symptoms influences prevalence estimates in a prospective cohort

Abstract

AbstractBackgroundStudies on long COVID differ in the selection of symptoms used to define the condition. We aimed to assess to what extent symptom selection impacts prevalence estimates of long COVID.MethodsIn a prospective cohort of patients who experienced mild to critical coronavirus disease 2019 (COVID-19), we used longitudinal data on the presence of 20 different symptoms to evaluate changes in the prevalence of long COVID over time when altering symptom selection.ResultsChanging symptom selection resulted in wide variation in long COVID prevalence, even within the same study population. Long COVID prevalence at 12 months since illness onset ranged from 39.6% (95%CI=33.4-46.2) when using a limited selection of symptoms to 80.6% (95%CI=74.8-85.4) when considering any reported symptom to be relevant.ConclusionsComparing the occurrence of long COVID is already complex due to heterogeneity in study design and population. Disparate symptom selection may further hamper comparison of long COVID estimates between populations. Harmonised data collection tools could be one means to achieve greater reproducibility and comparability of results.