Table 3.

Comparison of prediction models for chronic postsurgical pain (CPSP) in the testing dataset - calibration metrics.

	Logistic regression	RF	GBDT	XGBoost
ICI	0.070 (0.050 to 0.146)	0.093 (0.053 to 0.168)	0.072 (0.049 to 0.156)	0.050 (0.038 to 0.122)
E50	0.070 (0.027 to 0.127)	0.109 (0.034 to 0.170)	0.071 (0.035 to 0.138)	0.046 (0.024 to 0.123)
E90	0.111 (0.095 to 0.307)	0.129 (0.107 to 0.303)	0.114 (0.083 to 0.286)	0.071 (0.064 to 0.257)
H-L P value	0.059	0.429	0.384	0.829

Each cell contains the appropriate calibration metric and its 95% confidence interval.

E50 and E90, the median and 90^th percentile of the absolute difference between observed and predicted probabilities respectively; GBDT, gradient boosting decision tree; H-L, Hosmer-Lemeshow test; ICI, integrated calibration index (a measure of calibration, which could be interpreted as weighted difference between observed and predicted probabilities); RF, random forest; XGBoost, extreme gradient boosting.