Answers are grounded in this paper and cite their sources.
Background Obstructed Defecation Syndrome (ODS) can be complicated by systemic diabetic neuropathy. The objective of this study is to create machine learning models to predict failure of Conservative Treatment in patients with comorbid ODS and Type 2 Diabetes Mellitus (T2DM) and to investigate the underlying mechanisms of the comorbidities. Methods 513 patients with a balanced cohort of comorbid patients were analyzed. Clinical, metabolic and pelvic dynamic features were used for the evaluation of 6 machine learning algorithms. The Random Forest model was interpreted using SHapley Additive exPlanations (SHAP) analysis to understand the feature importance, non-linear associations, and interactions among the variables. Results The Random Forest model achieved a comparable global discriminative capacity (AUC = 0.814) among the evaluated models. SHAP analysis indicated the most important factors to be HbA1c, duration of T2DM and pudendal nerve terminal motor latency (PNTML), followed by the Wexner score and maximum voluntary squeeze pressure (MVSP). There was a potential interactive association between poor glycemic control and localized nerve impairment estimated by interaction analysis which was associated with a higher estimated risk on the multiplicative scale for treatment failure. Conclusion Machine learning algorithms are effective in identifying high-risk ODS-T2DM patients. However, the predictive performance seems to be mostly associated with systemic metabolic-neuropathic decline and not solely related to anatomical changes. The identified model-derived exploratory thresholds provide quantitative, hypothesis-generating insights, though independent external validation is required before any application in individual clinical management.
No takes yet. Share an insight, caveat, or question.
Fu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: