Objectives: Prior studies show that long COVID has a heterogeneous presentation. Whether specific risk factors are related to subclusters of long COVID remains unknown. This study aimed to determine pre-pandemic predictors of long COVID and symptom clustering.
Methods: A total of 3,022 participants of a panel representative of the Dutch population completed an online survey about long COVID symptoms. Data was merged into 2018/2019 panel data covering sociodemographic, medical, and psychosocial predictors. A total of 415 participants were classified as having long COVID. K-means clustering was used to identify patient clusters. Multivariate and lasso regression was used to identify relevant predictors compared to a COVID-19 positive control group.
Results: Predictors of long-term COVID included older age, Western ethnicity, BMI, chronic disease, COVID-19 reinfections, severity, and symptoms, lower self-esteem, and higher positive affect (AUC = 0.79, 95%CI 0.73-0.86). Four clusters were identified: a low and a high symptom severity cluster, a smell-taste and respiratory symptoms cluster, and a neuro-cognitive, psychosocial, and inflammatory symptom cluster. Predictors for the different clusters included regular health complaints, healthcare use, fear of COVID-19, anxiety, depressive symptoms, and neuroticism.
Conclusions: A combination of sociodemographic, medical, and psychosocial factors predicted long COVID. Heterogenous symptom clusters suggest that there are different phenotypes of long COVID-19 presentation.
Keywords: COVID-19; Clustering; Long COVID; Prediction model; Symptoms; post-COVID-19.
Copyright © 2024 The Author(s). Published by Elsevier Ltd.. All rights reserved.