How to develop a more accurate risk prediction model when there are few events

BMJ 2016; 353 doi: (Published 08 June 2016) Cite this as: BMJ 2016;353:i3235

In this Research Methods and Reporting paper (BMJ 2015;351:h3868, doi:10.1136/bmj.h3868), the section “Application of penalised regression” has a few text errors from the third paragraph onwards. A male patient should be considered rather than a female patient, and the risk score is −1.504 rather than −1.714. The text should therefore read as follows:

“Consider, for example, a male patient aged 20.5 years and with 1.7 m2 BSA, who had a 31 mm mitral valve manufactured after 1981 from a batch without fractured implants. Using the estimated coefficients from standard regression (table), the risk score for this patient is calculated by the following formula:

Risk score = −7.8 (intercept) + (−0.24×0(male sex)) + (−0.052×20.5(age; years)) + (1.98×1.7(BSA; m2)) + (2.62×1(mitral size 31 mm)) + (0.589×0(no fracture)) + (1.38×1(date of manufacture after 1981)) = −1.504.

Therefore, the predicted risk of mechanical failure is:

exp(−1.504) ÷ (1+exp(−1.504)) = 18% (average risk is 1.8%).

When the estimated coefficients from ridge and lasso are used instead, the predicted risks are less extreme: 12% and 15%, respectively.”

View Abstract