Table 3 –

Discrimination of the best-performing gradient boosting model on the external validation data.

Model	c-statistic	Average Precision	Sensitivity	Specificity	Positive Predictive Value	% Positive
Baseline	0.908 (0.899–0.916)	0.511 (0.481–0.539)	0.957 (0.946–0.969)	0.580 (0.573–0.587)	0.144 (0.137–0.151)	45.7% (45.0–46.4)
Gradient Boosting Machine	0.939 (0.933–0.944)	0.583 (0.554–0.610)	0.959 (0.949–0.969)	0.738 (0.732–0.744)	0.213 (0.203–0.223)	31.0% (30.4–31.6)

The best-performing gradient boosting model was evaluated on its ability to predict intraoperative transfusion for the external validation data using procedure-specific transfusion rates observed at this institution in 2019. Only procedures that occurred more than 50 times at this institution in 2019 are included, as transfusion rates are too uncertain below this threshold (Figure S3). c-statistic – area under the receiver operating characteristic curve. Average precision – area under the precision recall curve, indicative of model discrimination for the positive class. % Positive indicates percent of cases in the cohort for whom the model made a positive prediction, i.e., recommended a type and screen. All models were tuned with decision thresholds to achieve 96% sensitivity. 95% confidence intervals are shown in parentheses.