Table 5. Detailed prediction accuracies of random forest model for different categories of surgeries and aggregate opioid requirements.

Prediction accuracies prior to and after surgery are shown.

Surgical Specialty	Mean opioid requirement (MME)	Accuracy
		Beginning of surgery	End of surgery
General (N = 603)	12.1 ± 17.8	67%	70%
Gynecology (N = 289)	11.0 ± 17.2	71%	72%
Neuro (N = 132)	12.6 ± 25.1	70%	74%
Oral (N = 94)	7.7 ± 17.1	87%	86%
Orthopedic (N = 219)	19.7 ± 21.2	74%	70%
Otolaryngology (N = 495)	11.3 ± 25.4	68%	67%
Plastic (N = 317)	20.0 ± 23.4	67%	66%
Thoracic (N = 96)	2.9 ± 8.4	95%	94%
Urology (N = 393)	6.8 ± 14.4	80%	80%
Vascular (N = 27)	16.2 ± 25.2	70%	59%
Overall	12.2 ± 20.7	72%	72%