Predictors of shorter- and longer-term mortality after COVID-19 presentation among dialysis patients: parallel use of machine learning models in Latin and North American countries

Table 8 Performance of machine learning models in predicting death after COVID-19 presentation

		LatAm *Time after COVID-19*
Metric	Dataset	Any time	0–14 days	15–30 days	> 30 days	Any time	0–14 days	15–30 days	> 30 days
*Area under the curve*	Training	0.999	0.959	0.997	0.996	0.935	0.986	0.979	0.954
	Validation	0.783	0.755	0.755	0.739	0.779	0.812	0.790	0.802
	Testing	0.763	0.746	0.728	0.809	0.790	0.828	0.788	0.813
*Balanced accuracy*	Training	96.5	81.9	95.1	91.9	80.9	89.6	86.9	84.1
	Validation	70.2	66.7	73.3	70.0	68.1	72.4	70.2	70.0
	Testing	70.8	66.1	66.0	74.6	69.8	73.9	71.2	71.2
Area under precision-recall curve	Testing	0.210	0.375	0.062	0.040	0.515	0.298	0.233	0.356

The area under the curve (AUC) and area under precision-recall curve (AUPRC) are presented on a scale from 0 (lowest) to 1 (highest). Chance equals a value of 0.5 for AUC. Chance equals the fraction of positive cases in each regional group for each model for AUPRC (i.e., the number of patients who died in each group divided by the total number of patients in each group). Balanced accuracy is presented as a percentage

ISSN: 1471-2369