Negative pressure wound therapy (NPWT) shows variable outcomes in diabetic neuropathic foot ulcers. This study developed and internally validated a Random Forest prediction model for NPWT treatment success in 343 patients treated between 2018 and 2024. The model was developed using 70% training (n = 240) and 30% validation (n = 103) split. Treatment success (complete closure or >80% area reduction within 16 weeks) occurred in 240 patients (69.9%). The Random Forest model achieved validation set area under the curve of 0.89 (95% CI: 0.84-0.94), accuracy of 85.2%, sensitivity of 86.5%, and specificity of 82.8%. Exploratory analysis using complete wound closure alone yielded comparable discrimination (AUC 0.87). Permutation importance identified HbA1c (25.4%), ulcer size (20.8%), and estimated glomerular filtration rate (eGFR, 18.2%) as the strongest predictors. In multivariable logistic regression, HbA1c > 8.5% (OR 3.90), ulcer size >15 cm 2 (OR 3.02), and eGFR <60 mL/min/1.73 m 2 (OR 2.85) were independently associated with treatment failure. In this single-center cohort, the model showed good internal discrimination for the composite NPWT outcome; external validation is required before clinical use.
Nikolić et al. (Mon,) studied this question.