Neural networks outperform expert humans in predicting patient impressions of symptomatic improvement following overactive bladder treatment.

INTRODUCTION AND HYPOTHESIS: The objective was to accurately predict patient-centered subjective outcomes following the overactive bladder (OAB) treatments OnabotulinumtoxinA (OBTX-A) injection and sacral neuromodulation (SNM) using a neural network-based machine-learning approach. In the context of treatments designed to improve quality of life, a patient's perception of improvement should be the gold standard outcome measure. METHODS: Cutting-edge neural network-based algorithms using reproducing kernel techniques were trained to predict patient-reported improvements in urinary leakage and bladder function as assessed by Patient Global Impression of Improvement score using the ROSETTA trial datasets. Blinded expert urologists provided with the same variables also predicted outcomes. Receiver operating characteristic curves and areas under the curve were generated for algorithm and human expert predictions in an out-of-sample holdout dataset. RESULTS: Algorithms demonstrated excellent accuracy in predicting patient subjective improvement in urinary leakage (OBTX-A: AUC 0.75; SNM: 0.80). Similarly, algorithms accurately predicted patient subjective improvement in bladder function (OBTX-A: AUC 0.86; SNM: 0.96). The top-performing algorithms outcompeted human experts across outcome measures. CONCLUSIONS: Novel neural network-based machine-learning algorithms accurately predicted OBTX-A and SNM patient subjective outcomes, and generally outcompeted expert humans. Subtle aspects of the physician-patient interaction remain uncomputable, and thus the machine-learning approach may serve as an aid, rather than as an alternative, to human interaction and clinical judgment.

Machine learning provides an accurate prognostication model for refractory overactive bladder treatment response and is noninferior to human experts.

OBJECTIVE: The increasing wealth of clinical data may become unmanageable for a physician to assimilate into optimal decision-making without assistance. Utilizing a novel machine learning (ML) approach, we sought to develop algorithms to predict patient outcomes following the overactive bladder treatments OnabotulinumtoxinA (OBTX-A) injection and sacral neuromodulation (SNM). MATERIALS AND METHODS: ROSETTA datasets for overactive bladder patients randomized to OBTX-A or SNM were obtained. Novel ML algorithms, using reproducing kernel techniques were developed and tasked to predict outcomes including treatment response and decrease in urge urinary incontinence episodes in both the OBTX-A and SNM cohorts, in validation and test sets. Blinded expert urologists also predicted outcomes. Receiver operating characteristic curves were generated and AUCs calculated for comparison to lines of ignorance and the expert urologists' predictions. RESULTS: Trained algorithms demonstrated outstanding accuracy in predicting treatment response (OBTX-A: AUC 0.95; SNM: 0.88). Algorithms accurately predicted mean decrease in urge urinary incontinence episodes (MSE < 0.15) in OBTX-A and SNM. Algorithms were superior to human experts in response prediction for OBTX-A, and noninferior to human experts in response prediction for SNM. CONCLUSIONS: Novel ML algorithms were accurate, superior to expert urologists in predicting OBTX-A outcomes, and noninferior to expert urologists in predicting SNM outcomes. Some aspects of the physician-patient interaction are subtle and uncomputable, and thus ML may complement, but not supplant, a physician's judgment.