Abstract
Whereas the nature of the post-COVID condition following mild acute COVID-19 is increasingly well described in the literature, knowledge of its risk factors, and whether it can be predicted, remains limited. This study, conducted in Norway, uses individual-level register data from 214,667 SARS-CoV-2 infected individuals covering a range of demographic, socioeconomic factors, as well as cause-specific healthcare utilization in the years prior to infection to assess the risk of post-COVID complaints ≥3 months after testing positive. We find that the risk of post-COVID was higher among individuals who prior to infection had been diagnosed with psychological (OR = 2.12, 95% CI 1.84-2.44), respiratory (OR = 2.03, 95% CI 1.78-2.32), or general and unspecified health problems (OR = 1.78, 95% CI 1.52-2.09). To assess the predictability of post-COVID after mild initial disease, we use machine learning methods and find that pre-infection characteristics, combined with information on the SARS-CoV-2 virus type and vaccine status, to a considerable extent (AUC = 0.79, 95% CI 0.75-0.81) could predict the occurrence of post-COVID complaints in our sample.