BMC Urology (Aug 2025)

Artificial intelligence-based prediction of treatment failure and medication non-adherence in overactive bladder management

  • Mert Başaranoğlu,
  • İsa Kamil Taşdemir,
  • Erdem Akbay,
  • Hasan Erdal Doruk

DOI
https://doi.org/10.1186/s12894-025-01911-7
Journal volume & issue
Vol. 25, no. 1
pp. 1 – 17

Abstract

Read online

Abstract Background Overactive bladder management presents significant challenges, with treatment failures and medication non-adherence posing substantial barriers to patient outcomes. Early prediction of these challenges could enable timely interventions and treatment modifications. Objectives To develop and validate an artificial intelligence-based prediction model for early identification of treatment failure and medication non-adherence in overactive bladder patients, with specific focus on different pathological subgroups including diabetic neuropathy. Methods In this single-center retrospective study (January 2018–April 2025), we analyzed data from 285 patients with overactive bladder. We developed and validated artificial intelligence models using comprehensive clinical parameters, including ultrasonography findings, uroflowmetry results, standardized voiding diaries, and disease-specific questionnaire outcomes. Primary outcome measures were treatment failure and medication non-adherence at three months. Results The gradient boosting model achieved an accuracy of 87.3% (95% CI: 84.1–90.5%) for predicting treatment failure and 85.1% (95% CI: 81.8–88.4%) for predicting medication non-adherence. Key predictive factors included early changes in bladder wall thickness (OR: 3.82, 95% CI: 2.14–6.81), diabetes duration > 7 years (OR: 2.73, 95% CI: 1.58–4.72), and urgency improvement < 25% (OR: 2.94, 95% CI: 1.76–4.92). Treatment failure rates varied significantly among pathological subgroups, with highest rates in diabetic neuropathy (42.8%) and lowest in idiopathic OAB (28.6%, p = 0.024). Among treatment failure patients, 68.4% proceeded to advanced therapies, with differential success patterns across subgroups. Conclusions Our artificial intelligence model effectively identifies patients at risk of treatment failure and medication non-adherence in overactive bladder management. This approach enables early identification of high-risk patients, potentially improving treatment outcomes and healthcare resource utilization through timely intervention and treatment modification.

Keywords