Table 6.
Comparison of models developed from baseline data with models developed from the combination of baseline and grouped data. The difference-in-difference (DiD) reflects the AUC and F1-score for surgical site infection. A positive DiD indicates the combination model resulted in a smaller drop in performance than the baseline model.
| Combination | Medications | Laboratory tests | Diagnosis codes | AUCa, DiD (95% CI) | F1-score, DiD (95% CI) | P value |
| Combination 1 | Grouped | Baseline | Baseline | 0.054 (0.052-0.057) | 0.072 (0.069-0.074) | <.001 |
| Combination 2 | Baseline | Grouped | Baseline | 0.012 (0.010-0.014) | 0.046 (0.043-0.049) | <.001 |
| Combination 3 | Baseline | Baseline | Grouped | 0.049 (0.047-0.051) | 0.134 (0.131-0.137) | <.001 |
aAUC: area under the receiver operating characteristic curve.