Table 3.

Discrimination of the reference ML model, clinical variables, NP only, troponin only, and combined biomarkers model for predicting 10-year risk of developing HF in the derivation cohort among Black and White adults.

	Top-20 (reference)	Clinical variables	Clinical + laboratory variables	NP only	Troponin only	Complete data
Black adults	0.88 (0.85–0.90)	0.75 (0.73–0.78)	0.75 (0.73–0.78)	0.70 (0.66–0.74)	0.70 (0.66–0.72)	0.88 (0.85–0.90)
White adults	0.88 (0.85–0.90)	0.76 (0.71–0.80)	0.76 (0.71–0.81)	0.64 (0.61–0.66)	0.69 (0.66–0.71)	0.88 (0.85–0.90)

Data presented are C-index (95% confidence interval).

The following models were analyzed: race specific oblique random survival forests (oRSF) with the top 20 most important variables (oRSF top-20) [reference], oRSF with only clinical variables (age, sex, systolic blood pressure, diastolic blood pressure, heart rate, body mass index, smoking history, history cardiovascular disease, history of diabetes); oRSF with clinical and laboratory variables (clinical variables + hemoglobin A1c, total cholesterol, high-density lipoprotein-cholesterol), oRSF with natriuretic peptide (NP) levels only, oRSF with troponin levels only, and oRSF with complete data (i.e, no missing data). Of the 11,999 participants in the derivation cohort, 10,053 (83.8%) had no missing data. Confidence intervals are the 95% ranges among 1000 bootstrapped replicates.