A risk prediction algorithm for venous ulceration: A prospective case-control study

Introduction:Previous research has focused on the treatment of leg ulcers, in order to investigate treatments for the prevention of venous ulceration (VLU) it is essential to identify the population at risk. We performed a prospective case-controlled study to identify risk factors for VLU and to develop a risk prediction score.Methods:Health questionnaires were sent to 441 patients with leg ulcers (cases) and friends of the same sex and age with no history of leg ulcer (controls). A 70% random sample from the 441 subjects was taken to form the training set (the subgroup from which the scoring system was derived). The remaining 30% formed the test set (the subgroup used to assess its predictive ability). Multivariate logistic regression was used to select predictors which had a significant independent relationship with VLU. This multivariate model was used to derive a diagnostic score, using the regression coefficients from the fitted model. This scoring system was then applied to the test set and sensitivity and specificity for different cut-off values on the score were calculated. The estimated risk of developing a venous ulcer was derived for different categories of score, corrected for the prevalence of 1.5%, and was then separated in tertiles. The expected risk of VLU for a person with a score in each tertile was calculated using Bayes theorem on the distribution of risk prediction scores in the training set.Results:Data was obtained from 441 subjects (231 cases and 210 matched controls); the two groups were well matched. Risk factors significantly associated (Odd ratio ± 95% CI) with VLU on multivariate analysis included a history of DVT (11.72; 1.43-95.8), phlebitis (10.28; 2.19-48.4), hip replacement (5.28; 1.01-27.8), poor mobility (4.00; 1.88-8.50), weight>100Kg (3.45; 1.42-8.36), varicose veins (VV) (2.49; 1.18-4.99), family history of VV (1.89; 1.01-3.56) and weight >75