Table 1.
Database summary.
Table 2.
Characteristics of the four claims datasets.
Table 3.
The data and internal model performance details.
Table 4.
The AUC values for each validation.
The brackets represent the confidence intervals—this is presented when the validation data is <100,000 people and ignored for larger data where the AUC values are more stable.
Table 5.
The CROUD model questions and points per question response.
A patient’s risk can be determined by calculating their overall score as the sum of their points and then using Table 6 to match the risk for that score.
Table 6.
Predicted risk in each score strata compared to population risk of 0.14% in Optum.
Table 7.
External validation discriminative performance of CROUD.