Table 2. Performance of machine learning models for predicting recurrence and survival.

Post-cystectomy outcome	Model	Year	Test set performance metrics
			Sensitivity	Specificity	Precision	F1
Recurrence	Meta-classifier	1	0.739	0.714	0.388	0.508
	pT stage TNM 5th Edition	1	0.761	0.653	0.349	0.478
	pathologic stage subgroup	1	0.826	0.593	0.332	0.473
	Meta-classifier	3	0.720	0.708	0.535	0.613
	pathologic stage subgroup	3	0.774	0.631	0.493	0.602
	pT stage TNM 5th Edition	3	0.670	0.694	0.503	0.574
	Meta-classifier	5	0.700	0.702	0.588	0.636
	pathologic stage subgroup	5	0.744	0.611	0.537	0.623
	pT stage TNM 5th Edition	5	0.619	0.698	0.553	0.583
Survival	Meta-classifier	1	0.741	0.770	0.473	0.577
	pT stage TNM 5th Edition	1	0.739	0.672	0.387	0.506
	pathologic stage subgroup	1	0.805	0.602	0.362	0.499
	Meta-classifier	3	0.722	0.788	0.700	0.711
	pathologic stage subgroup	3	0.762	0.691	0.628	0.688
	pT stage TNM 5th Edition	3	0.696	0.739	0.646	0.670
	Meta-classifier	5	0.741	0.768	0.780	0.760
	pathologic stage subgroup	5	0.730	0.717	0.742	0.735
	pT stage TNM 5th Edition	5	0.664	0.766	0.759	0.708

Single predictor (pT stage and pathologic stage subgroup classifiers) and multiple predictor (Meta-classifier) models for predicting 1-, 3-, 5-year recurrence and survival after cystectomy. The performance of all models for a given year is ranked per F1 scores (2*precision*recall/(precision+recall)) as well as mean sensitivity, specificity, and precision on test sets from a 10-fold cross validation.