AI-Powered Detection of Functional Gastrointestinal Disorders in Adults Visiting Tertiary Hospitals

Background: Functional gastrointestinal disorders are common in tertiary gastroenterology practice and remain difficult to diagnose because they are defined largely by symptom patterns rather than structural or biochemical abnormalities. In busy referral settings, overlap among irritable bowel syndrome, functional dyspepsia, functional constipation, and non-functional gastrointestinal conditions can reduce diagnostic consistency. Artificial intelligence offers a potential solution by integrating structured clinical variables with free-text symptom narratives to support more standardized classification. Objective: To evaluate the diagnostic accuracy of an artificial intelligence–based system for detecting functional gastrointestinal disorders among adults attending tertiary hospitals. Methods: This multicenter cross-sectional diagnostic accuracy study was conducted over eight months in three tertiary hospitals in Lahore, Pakistan. A total of 420 adults aged 18–65 years presenting with gastrointestinal symptoms were enrolled through consecutive sampling. After exclusion of relevant organic pathology, Rome IV–based clinician diagnosis served as the reference standard. The artificial intelligence system analyzed structured clinical variables and unstructured consultation notes using supervised machine learning and natural language processing to classify irritable bowel syndrome, functional dyspepsia, functional constipation, and non-functional gastrointestinal disorder status. Diagnostic performance was assessed using sensitivity, specificity, predictive values, area under the receiver operating characteristic curve, and Cohen’s kappa. Results: The mean age was 37.4 ± 10.8 years, and 54.8% of participants were male. Clinically, 33.3% had irritable bowel syndrome, 23.8% functional dyspepsia, 14.3% functional constipation, and 28.6% non-functional gastrointestinal disorders. The artificial intelligence model achieved a sensitivity of 88.2%, specificity of 84.5%, positive predictive value of 86.9%, negative predictive value of 86.0%, overall accuracy of 87.1%, and an overall area under the curve of 0.912. Agreement with clinician diagnosis was strong (κ = 0.83, p < 0.001). Conclusion: The artificial intelligence system showed high diagnostic performance and strong concordance with Rome IV–based clinical diagnosis, supporting its potential role as an adjunctive decision-support tool for functional gastrointestinal disorder classification in tertiary care.