Skip to main content
. 2019 Jul 19;14(7):e0219964. doi: 10.1371/journal.pone.0219964

Table 2. ROCs and integrated discrimination improvement (IDI) for prediction of diabetes at years 1, 3, and 5.

Models Year 1 Year 3 Year 5
ROC AUC (95th% CI) Model 1: Demographic factors a 0.701
(0.699, 0.704)
0.703
(0.701, 0.705)
0.702
(0.700, 0.704)
Model 2: Demographic + CVD factors b 0.708
(0.705, 0.710)
0.705
(0.703, 0.707)
0.703
(0.701, 0.705)
Model 3: RPG (maximum) 0.856
(0.855, 0.858)
0.822
(0.820, 0.823)
0.791
(0.790, 0.793)
Model 4: RPG (average) 0.877
(0.876, 0.879)
0.836
(0.834, 0.837)
0.799
(0.798, 0.801)
Model 5: RPG (≥2 measures at/above threshold) 0.878
(0.876, 0.880)
0.839
(0.837, 0.840)
0.803
(0.802, 0.805)
Model 6: Demographic + CVD + RPG (maximum) 0.882
(0.881, 0.884)
0.855
(0.854, 0.857)
0.832
(0.831, 0.833)
Model 7: Demographic + CVD + RPG (average) 0.899
(0.898, 0.901)
0.869
(0.868, 0.871)
0.843
(0.842, 0.845)
Model 8: Demographic + CVD + RPG (≥2 at/above threshold) 0.900
(0.899, 0.902)
0.872
(0.870, 0.873)
0.847
(0.846, 0.848)
IDI c Model 1: Demographic factors a reference reference reference
Model 2: Demographic + CVD factors b 0.0008 0.0011 0.0014
Model 3: Demographic + CVD + RPG (≥2 at/above threshold) 0.1872 0.2456 0.2615

n = 941,561, including only subjects with all CVD risk factors

a Demographic factors (age, sex, BMI, race, ethnicity)

b CVD factors (systolic blood pressure, non-HDL-cholesterol, smoking)

c p<0.0001 for all IDI analyses comparing prediction performances using model 1 as reference within each year