An Evolutionary Computation Approach for Optimizing Multilevel Data to Predict Patient Outcomes

Figure 3

Bullseye performance for baseline models (with specific complaints only and complaint categories only, resp.) and hierarchical genetic algorithm for the academic hospital. Overall performance is indicated outside of the bullseye. Statistical significance for the difference in 5-fold cross-validated AUC (using DeLong’s method) between the flattened genetic algorithm approach and the corresponding baseline models is indicated by for , for , and for .

(a)

(b)

(c)