Table 2.

Performance of optimized surgical duration prediction models at each time point: test set, external validation set, and prediction intervals.

		Time of Scheduling model (secondary model)	Time of Patient in ORa model (primary model)	Time of Surgical Incision model (secondary model)
Test set
	Mean absolute error (min), mean (SD)	34 (47)	34 (47)	34 (47)
	Root mean square error, min	59	59	59
	Overage, %	58	58	58
	Underage, %	42	42	42
	Prediction within 20% of actual duration, %	46	46	46
External validation set
	Mean absolute error (min), mean (SD)	38 (52)	38 (52)	38 (52)
Prediction intervals
	Actual duration within prediction interval, %	Error prediction model: 64; quantile loss function: 61	Error prediction model: 63; quantile loss function: 61	Error prediction model: 65; quantile loss function: 61

^aOR: operating room.