I read with interest the article by Schulze et al. (1). In this article, they provide a novel risk score algorithm for the risk of diabetes by allocating specific points from the regression coefficients in the Cox proportional hazards model. Developing a simple, reliable, and valid risk score method for predicting incident diabetes is useful for screening and management of diabetes risk (2–5). Schulze et al. constructed the score points from the Cox model and estimated the probability (P) of diabetes during the following 5 years based on the formula:
Unfortunately, the above formula was wrong because the sample mean point of the study population was not removed from the exponential term. The above formula should be modified into the following:
The sample mean point was a constant derived from the product of the regression coefficients and the means (or proportions) of the risk factors (6).
