Skip to main content
Figure 1 | BMC Medical Informatics and Decision Making

Figure 1

From: Evaluating predictive modeling algorithms to assess patient eligibility for clinical trials from routine data

Figure 1

Evaluation of the prototype’s performance for models based on age, gender and the 20 codes most associated with trial eligibility: model fits were compared by area under the receiver operating characteristic (ROC) curve complemented with 95% confidence limits (y-axis). Models were trained with differently sized training sets of patients (x-axis). All models were tested against one third of the original patient set (hold-out method), thus the maximum size of the training set was two thirds of all patients. The experiment was done for the original codes as documented, after aggregation to category level and after aggregation to block level.

Back to article page