Skip to main content
. 2016 Sep 7;354:i4450. doi: 10.1136/bmj.i4450

Table 1.

Prediction measures examining the effect of adding high density lipoprotein (HDL) cholesterol to cardiovascular disease risk models in the Women’s Health Study

Prediction measure Models derived using full population Models derived using intermediate risk group
Ln HDL coefficient (SE), P value −1.0 (0.14), <0.001 −0.52 (0.34), 0.13
Measures for changes in prediction for intermediate risk group comparing model adding HDL to model without HDL
Original Expected under the null
Probability of a case moving up (95% CI)* 0.27 (0.18 to 0.37) 0.22 0.45 (0.35 to 0.56)
Probability of a non-case moving up (95% CI)* 0.20 (0.18 to 0.22) 0.17 0.28 (0.27 to 0.31)
Probability of a case moving down (95% CI)* 0.14 (0.09 to 0.23) 0.19 0.16 (0.10 to 0.26)
Probability of a non-case moving down (95% CI)* 0.27 (0.25 to 0.29) 0.29 0.25 (0.23 to 0.27)
Intermediate NRI (95% CI)† 0.20 (0.05 to 0.33) 0.15 0.25 (0.10 to 0.41)
Bias corrected 0.05 (−0.10 to 0.18) NA
Reclassification calibration for model without HDL (P value) 6.7 (0.04) 6.8 (0.03)
Reclassification calibration for model with HDL (P value) 1.6 (0.5) 0.7 (0.7)

NA=not applicable; NRI=net reclassification improvement.

*Confidence intervals calculated using Agresti-Coull method.

†Confidence intervals calculated using bootstrap.