Table 5:

Evaluation metrics of CKD classifications based on predicted eGFR. Precision, Recall, and F1-score for each class for each year was used to show performance. Additionally, macro- and micro-averaged metrics were used for each year to show overall performance, regardless of the categorization. CKD-1 was left out due to no methods predicting a CKD-1 result for any year for any patients.

		Per-class accuracy			Macro-averaged metrics			Micro-averaged metrics
		Precisio n	Reca ll	F1	Macro precision	Macro recall	MacroF 1
Year 1	CKD- 2	0.980	1.000	0.99 0	0.966	0.876	0.917	0.979
	CKD- 3	0.983	0.918	0.95 0
	CKD- 4	0.901	0.787	0.84 0
	CKD- 5	1.000	0.801	0.88 9
Year 2	CKD- 2	0.978	1.000	0.98 9	0.969	0.882	0.922	0.979
	CKD- 3	0.983	0.923	0.95 2
	CKD- 4	0.913	0.770	0.83 5
	CKD- 5	1.000	0.835	0.91 0
Year 3	CKD- 2	0.980	1.000	0.99 0	0.964	0.882	0.919	0.980
	CKD- 3	0.985	0.933	0.95 9
	CKD- 4	0.892	0.814	0.85 1
	CKD- 5	1.000	0.782	0.87 8