Development and Evaluation of a New Predictive Nomogram for Predicting Risk of Herpes Zoster Infection in a Chinese...

To identify potential risk factors for herpes zoster infection in type 2 diabetes mellitus in southeast Chinese population.

We built a model involving 266 herpes zoster patients collecting data from January 2018 to December 2019. The least absolute shrinkage and selection operator (Lasso) predictive model was used to test herpes zoster virus risk using the patient data. Multivariate regression was conducted to decide which variable would be the strongest to decrease the Lasso penalty. The predictive model was tested using the C-index, a calibration plot, and decision curve study. External validity was verified by bootstrapping by counting probabilities.

In the prediction nomogram, the prediction variables included age, sex, weight, length of hospital stay, infection, and blood pressure. The C-index of 0.844 (0.798-0.896) indicated substantial variability and thus the model was adjusted appropriately. A score of 0.825 was achieved somewhere in the above interval. Examination of the decision curve estimated that herpes zoster nomogram was useful when the intervention was determined at the 16 percent of the herpes zoster infection potential threshold.

The herpes zoster nomogram combines age, weight, position of the rash, 2-hour plasma glucose, glycosuria, serum creatinine, length of the hospital stay, and hypertension. This calculator can be used to assess the individual herpes zoster risks in patients diagnosed with type 2 diabetes mellitus.