. Author manuscript; available in PMC: 2022 Aug 5.

Published in final edited form as: IEEE J Biomed Health Inform. 2021 Aug 5;25(8):3121–3129. doi: 10.1109/JBHI.2021.3063721

TABLE V.

COMPARISON OF PREDICTION PERFORMANCE FOR DIFFERENT MODELS

Prediction window	Two weeks		Three months		Six months		One year
Models	ROCAUC	PRAUC	ROCAUC	PRAUC	ROCAUC	PRAUC	ROCAUC	PRAUC
MiME* ICD-9+CPT+Medication	0.76 (0.01)	0.67 (0.02)	0.74 (0.01)	0.64 (0.02)	0.72 (0.02)	0.61 (0.01)	0.70 (0.01)	0.61 (0.01)
BEHRT ICD-9	0.77 (0.02)	0.68 (0.01)	0.75 (0.02)	0.65 (0.01)	0.71 (0.01)	0.61 (0.02)	0.69 (0.02)	0.60 (0.02)
Dipole ICD-9+CPT	0.78 (0.02)	0.70 (0.01)	0.76 (0.02)	0.67 (0.02)	0.75 (0.01)	0.65 (0.02)	0.74 (0.01)	0.64 (0.01)
HCET All	0.81 (0.01)	0.73 (0.01)	0.80 (0.01)	0.70 (0.02)	0.79 (0.01)	0.69 (0.01)	0.78 (0.01)	0.67 (0.01)
BRLTM All	0.85 ^† (0.02)	0.78 ^† (0.01)	0.84 ^† (0.01)	0.76 ^† (0.01)	0.83 ^† (0.01)	0.74 ^† (0.02)	0.81 ^† (0.01)	0.73 ^† (0.01)

Values in parenthesis stand for standard deviations across randomizations and bold values denotes the highest in each column.

^†

indicates the value significantly better than that from the best baseline model HCET (p<0.05). The words after each models denotes the input data modalities where all means all five in our dataset.