External validation and optimisation of a model for predicting foot ulcers in patients with diabetes

In 2006 a risk stratification model was developed by Boyko et al. to predict foot ulceration in patients with diabetes, using seven commonly available clinical variables. We sought to validate and optimise this clinical prediction rule in a different setting. A retrospective cohort study was conducted on all patients with diabetes attending the podiatry section of a diabetic foot clinic at a tertiary hospital in Portugal (n = 360). Assessment at baseline included variables evaluated in the previous study and other relevant variables. Type 2 diabetes was present in 98% of patients, 45% were men and (at baseline) the median age was 65 years. Median follow-up was 25 months (range 3-86), during which 94 patients (26%) developed a foot ulcer. Boyko's model had an area under the receiver operating curve of 0.83 (95% CI 0.78-0.88). The corresponding value for the optimised model, which included the footwear risk variable, was 0.88 (95% CI 0.84-0.91). Both models had high classification accuracy for prediction of foot ulceration. However, the optimised model tended to produce higher specificity and positive likelihood ratio values at all levels. This study confirmed that Boyko's proposed model has a high capacity to predict foot ulceration in diabetes patients of both sexes. Our results suggest that the inclusion of a further footwear variable could improve the model. Nevertheless, prospective validation in a larger population is still necessary.