Table 5.
Difference-in-difference (DiD) metrics for the comparison between baseline and granular models and the comparison between baseline and grouped models. A positive DiD indicates the comparison model resulted in a reduced drop in performance compared to the baseline model.
| Metric, outcome | Granular comparison, DiD (95% CI) | Grouped comparison, DiD (95% CI) | |
| AUCa | |||
|
|
SSIb | 0.035 (0.033-0.037) | 0.072 (0.070-0.074) |
|
|
Pneumonia | 0.226 (0.223-0.229) | 0.250 (0.247-0.252) |
|
|
Sepsis | 0.015 (0.013-0.017) | 0.008 (0.007-0.010) |
|
|
UTIc | –0.049 (–0.052 to –0.045) | 0.006 (0.002-0.009) |
| F1-score | |||
|
|
SSI | 0.017 (0.014-0.020) | 0.100 (0.097-0.103) |
|
|
Pneumonia | 0.186 (0.179-0.193) | 0.212 (0.206-0.218) |
|
|
Sepsis | 0.026 (0.023-0.028) | 0.091 (0.089-0.093) |
|
|
UTI | 0.039 (0.035-0.043) | 0.073 (0.068-0.077) |
aAUC: area under the receiver operating characteristic curve.
bSSI: surgical site infection.
cUTI: urinary tract infection.