This systematic review of prediction models for COVID-19 showed that while they are quickly entering the academic literature and are urgently needed, most are poorly reported, at high risk of bias, and their reported performance is probably optimistic.
Authors reviewed and critically appraise studies of prediction models for COVID-19 in patients with suspected infection, for prognosis of patients, and for detecting those at risk of being admitted to the hospital. They included studies that developed or validated a COVID-19 related prediction model.
They included 27 studies describing 31 prediction models. Three models were identified for predicting hospital admission from pneumonia and other events; 18 diagnostic models for detecting COVID-19; and 10 prognostic models for predicting mortality risk, progression to severe disease, or length of hospital stay. Only one study used patient data from outside of China.
The most commonly used predictors of COVID-19 in patients with suspected disease included age, temperature, and symptoms. The most commonly used predictors of severe prognosis included age, sex, CT scans, C reactive protein, lactic dehydrogenase, and lymphocyte count.
All studies were rated at high risk of bias. Reporting quality varied substantially. Most reports did not include an adequate description of the study population or how the models were intended to be used.