A.I. Machine Learning Model Reliably Predicts Success of Inguinal Hernia Repair to Resolve Chronic Pelvic and Groin Pain

Background The differential diagnoses of chronic pelvic pain (CPP) and groin pain are vast and may include inguinal hernia (IH) as well as orthopedic, urologic, and gynecologic disorders. Methods A 50-point Hernia Health Questionnaireⓒ (HHQ) was administered prospectively to all patients with CPP or groin pain who underwent inguinal hernia repair. Patients were followed for resolution of preoperative symptoms as the endpoint. The machine learning model randomly split patients into blinded training and testing sets followed by Random Forest modelling of postoperative symptom resolution. Recursive Feature Elimination identified optimal predictive variables from the HHQ. Bootstrapping determined confidence interval. Results Between 2008 and 2022, 494 patients underwent primary IH repair. Mean age was 52 years, mean BMI 26.2 kg/m2, and 202 (41%) were female. Postoperative resolution of preoperative pain was confirmed in 426 (86%). Random Forest model exhibited strong predictive performance, with high two-sided 95% confidence interval (0.72, 0.91, median 0.83). Significant predictors of IH included female sex (p < 0.001), location/radiation of pain (p < 0.001), painful intercourse (p = 0.02), pain medications preoperatively (p < 0.001), and history of hernia in a female family member (p = 0.02). Conclusion We present the first A.I. model to reliably predict whether repairing IH will resolve the patient’s CPP or groin pain. This can be valuable to surgeons and non-surgeons, with the aim of reducing delay in diagnosis and treatment of patients with enigmatic CPP or groin pain.