Table 2.

Patient-level aggregation results: from the highest scoring predictions, a match is counted if the predicted value matches a value in the adjudicated reference data (top-N match).

Set	Model	Top-N match predictions	Top-N match reference	Relative match
Training	Top-1	75/120 (0.625)	83/120 (0.692)	0.904
	Top-3	87/120 (0.725)	92/120 (0.767)	0.946
	Top-5	92/120 (0.767)	94/120 (0.783)	0.988
Test	Top-1	19/31 (0.613)	25/31 (0.806)	0.760
	Top-3	22/31 (0.710)	26/31 (0.839)	0.846
	Top-5	22/31 (0.710)	26/31 (0.839)	0.846