Aim: to propose a model to estimate the risk of hospital admission due to COVID-19.

Study design: a cohort study of suspected and confirmed COVID-19 patients presenting in 14 general practices (GP) in and 8 out-of-hours services in the Netherlands, merged with hospital admission records. The development cohort spanned 01/03/2020 to 15/06/2020, the validation cohort 13/03/2021 to 19/06/2021. Methods: Adult patients with suspected or confirmed COVID-19 who contacted their GP or out-of-hours service for a face-to-face consultation were included. Data on patient characteristics and physical examination were recorded retrospectively in a case report form (CRF ). A logistic regression model was developed to predict hospital admission with COVID-19 within 14 days. Candidate predictors included demographic data, comorbidities, and symptoms. Model performance was evaluated using the C-statistic, calibration plots, and net benefit. Regional internal-external validation was used, in addition to a temporal external validation. 4806 patients were included in the development cohort and dataset. Their median age was 56 (IQR 40 to 70), 56.4% were female. The median duration of complaints at the first visit was 7 days (IQR 2.5 to 14 days). The majority of patients presented with dyspnoea (64.4%) or a cough (63.8%). Being male, auscultation abnormalities, confusion, cough, haemoptysis, stomach complaints, headache, chronic kidney disease, lower oxygen saturation and higher body temperature increase the probability of hospital admission.

Publication list will be added