Table 4.

Summary of model performance

	C-statistic [95% CI] (0 to 1, higher is better)	Brier Score [95% CI] (0 to 1, lower is better)
Outcome: Level of Treatment
Proportional odds model
Overall	0.686 [0.657, 0.741]	NA a
Level 1–4 vs. 0	0.686 [0.657, 0.742]	0.154 [0.126, 0.175]
Level 2–4 vs. 0–1	0.685 [0.658, 0.741]	0.217 [0.193, 0.226]
Level 3–4 vs. 0–2	0.686 [0.658, 0.741]	0.162 [0.137, 0.180]
Level 4 vs. 0–3	0.684 [0.656, 0.741]	0.063 [0.040, 0.084]
Separate binary logistic regression models
Level 1–4 vs. 0	0.663 [0.653, 0.712]	0.158 [0.146, 0.173]
Level 2–4 vs. 0–1	0.702 [0.664, 0.779]	0.216 [0.191, 0.226]
Level 3–4 vs. 0–2	0.796 [0.762, 0.868]	0.159 [0.127, 0.173]
Level 4 vs. 0–3	0.673 [0.615, 0.863]	0.065 [0.037, 0.084]
Outcome: Time to OAB Medication Discontinuation
Cox proportional hazard model	0.671 [0.579, 0.766]	0.185 [0.151, 0.247] b
Outcome: Sling
Binary logistic regression	0.823 [0.829, 0.834]	0.091 [0.089, 0.091]

Note: Brier score captures both discrimination and calibration. It ranges between 0 and 1, with lower score indicating better model performance. C-statistic assess discrimination and ranges between 0 and 1, with higher score indicating better discrimination performance.

^aBrier score is inappropriate for ordinal outcome.

^bBrier score evaluated at 6-month since OAB medication initiation or baseline if medication was started before baseline. It is not corrected for bias.

CI, confidence interval; OAB, overactive bladder