Machine learning models to predict postoperative incontinence after endoscopic enucleation of the prostate for benign prostatic hyperplasia: An EAU-Endourology study

Background: Machine learning (ML) and artificial intelligence (AI) have demonstrated powerful functionality in the healthcare setting thus far. We aimed to construct an AI model to predict postoperative incontinence after enucleation surgery for benign prostatic hyperplasia (BPH). Methods: Data were taken from two BPH registries and split into training and validation datasets. The following characteristics were used as predictors of incontinence: age, prostate volume, preoperative IPSS, QoL score, Qmax and post-void residual; presence of preoperative indwelling catheter, early apical release (EAR), enucleation type (2-lobe, 3-lobe, or en-bloc), and laser energy type. Six types of ML models were constructed using the training dataset and applied to the validation dataset to assess their accuracy. Results: 3828 patients from both databases were analyzed. Median age was 68, median prostate volume was 85.5 cc. 5.4% had a preoperative indwelling catheter. The commonest enucleation type was 2-lobe, the commonest energy type was Thulium fiber laser, and EAR was performed in 34.0%. Of the six ML models tested, extreme gradient boosting with manual fine tuning was the best-performing with an accuracy of 86.2%, sensitivity of 96.8%, specificity of 23.7%, PPV of 88.2%, and NPV of 55.9%. Conclusions: We hereby present an ML model for incontinence prediction post-surgery for BPH. Its main strengths are high sensitivity and PPV, meaning that if a patient is predicted to be incontinent using this model, this is likely to reflect the eventual outcome. This allows clinicians to pay closer attention on follow-up to detect and manage postoperative incontinence expediently.