Table 2.

P value of hypothesis testing using IDI, NRI, AIC and BIC.

(a)	EF				EFa				LVM
	EF ≥ 50	40 ≤ EF < 50	EF < 40	All patients	EF ≥ 50	40 ≤ EF < 50	EF < 40	All patients	EF ≥ 50	40 ≤ EF < 50	EF < 40	All patients
IDI	0.134	0.212	0.236	<0.001	0.003	0.072	0.055	<0.001	0.002	0.110	0.230	0.476
NRI	0.434	0.378	0.482	<0.001	0.006	0.133	0.238	<0.001	0.038	0.368	0.324	0.150
AIC, BIC	0.186	0.150	0.342	0.410	0.012	0.117	0.082	<0.001	0.058	0.002	0.086	0.002
(b)
IDI	0.266	0.270	0.286	<0.001	0.050	0.252	0.244	<0.001
NRI	0.288	0.207	0.107	<0.001	0.018	0.207	0.115	<0.001

The null hypothesis is that risk association does not increase when either EF or EFa or LVM is added to the (a) “Baseline” model and (b) “15BT” model, for readmission risks (due to HF) within 3 years. The “Baseline” model utilizes age, sex and blood creatinine as confounders while the “15BT” model also includes other blood tests listed in Table 1. For AIC and BIC, only the higher p value was stated. Green highlight: p < 0.05 of lower or equal IDI/NRI, and high or equal AIC and BIC. See Supplementary Table 1 for shorter survival times.