Table III.

Performance assessment of all selected models on unforeseen test data.

Variable	Neural network	Gradient boosting	LASSO	Ridge	Elastic net	Random forest	Logistic regression	Pre-surgery PROM scores
Knee arthroplasty
EQ-5D-5L (n = 288) AUC (95% CI)	0.76 (0.7 to 0.81)	0.79 (0.74 to 0.84)	0.75 (0.69 to 0.8)	0.75 (0.69 to 0.81)	0.76 (0.7 to 0.81)	0.80 (0.74 to 0.85)*	0.74 (0.68 to 0.8)	0.76 (0.7 to 0.81)
EQ-VAS (n = 307), AUC (95% CI)	0.73 (0.67 to 0.78)	0.74 (0.69 to 0.8)	0.76 (0.71 to 0.82)	0.76 (0.7 to 0.81)	0.76 (0.71 to 0.82)*	0.73 (0.68 to 0.79)	0.76 (0.7 to 0.81)	0.75 (0.7 to 0.81)
KOOS-PS (n = 309), AUC (95% CI)	0.68 (0.62 to 0.75)	0.71 (0.64 to 0.77)	0.75 (0.69 to 0.81)	0.73 (0.67 to 0.79)	0.76 (0.7 to 0.82)*	0.69 (0.63 to 0.76)	0.76 (0.7 to 0.81)	0.74 (0.68 to 0.8)
Hip arthroplasty
EQ-5D-5L (n = 290), AUC (95% CI)	0.8 (0.75 to 0.86)	0.81 (0.76 to 0.86)*	0.81 (0.76 to 0.86)	0.8 (0.75 to 0.85)	0.81 (0.76 to 0.86)	0.81 (0.75 to 0.86)	0.81 (0.76 to 0.86)	0.79 (0.73 to 0.84)
EQ-VAS (n = 364), AUC (95% CI)	0.82 (0.78 to 0.86)	0.83 (0.79 to 0.87)	0.84 (0.8 to 0.88)*	0.84 (0.8 to 0.88)	0.84 (0.8 to 0.88)	0.84 (0.8 to 0.88)	0.84 (0.8 to 0.88)	0.8 (0.75 to 0.84)
HOOS-PS (n = 366), AUC (95% CI)	0.71 (0.65 to 0.76)	0.67 (0.62 to 0.72)	0.66 (0.61 to 0.72)	0.71 (0.66 to 0.76)*	0.71 (0.65 to 0.76)	0.64 (0.58 to 0.69)	0.67 (0.61 to 0.72)	0.58 (0.47 to 0.68)

^*Best-performing model (sometimes identified using further decimal digits than those shown in the table).