Answers are grounded in this paper and cite their sources.
BACKGROUND: We compared three strategies for predicting 1-year primary patency after first autogenous arteriovenous fistula (AVF) creation. METHODS: This retrospective cohort included 274 patients. Models A, B, and C were a 5-predictor logistic model, a 31-predictor random forest, and a 9-predictor random forest-assisted logistic model. The three models were compared in 251 patients with complete Model C predictors using 20 repeats of stratified 5-fold cross-validation. RESULTS: Mean AUCs (SD) were 0.9361 (0.0045), 0.8843 (0.0095), and 0.9301 (0.0089) for Models A, B, and C; corresponding Brier scores were 0.0886 (0.0033), 0.1246 (0.0026), and 0.0933 (0.0060). Model A showed slightly better average cross-validated discrimination and prediction error than Model C, whereas Model C performed better than Model B. CONCLUSIONS: Random forest-assisted modeling did not improve cross-validated predictive performance over the simpler logistic model. External validation is required before clinical implementation.
No takes yet. Share an insight, caveat, or question.
Han et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: