Reporting of candidate predictors.

Impact studies (n = 3) were excluded from this table as their aim is not to develop or validate a prediction model, but rather to quantify the effect or impact of using a prediction model on physicians' behaviour, patient outcome, or cost-effectiveness of care relative to not using the model or usual care. Hence, for this table total n = 68.aNot applicable for the three external validation studies. Hence, n = 65.bNot applicable in four studies, because one studied no continuous predictors, and the others were the three external validation studies. Hence, n = 64. Of these, handling was unclear in 19 studies, not described in two studies. The sum 43+12+30+21+2 is more than 43 because some studies handled continuous predictors in two ways (e.g., dichotomizing blood pressure and categorising body mass index into four categories).