Introduction: Opioid-induced respiratory depression is a life-threatening complication of opioid overdose. This study aimed to develop a model for predicting the risk of respiratory depression following opioid overdose using ChatGPT-4o.
Methods: A retrospective cross-sectional study was conducted on 2,005 patients admitted following opioid overdose at Loghman Hakim Hospital, Tehran, Iran, from February 2021 to February 2024. Demographic data, clinical presentations, interventions, and outcomes of patients were extracted from electronic medical records and a predictive model was developed using a no-code methodology with the assistance of ChatGPT-4o.
Results: 2,005 patients with the mean age of 32.97 ± 14.86 (Range: 1-100) years were studied (74.5% male). Respiratory depression was observed in 18% of patients upon admission. Naloxone was administered to 37.6% of patients, with higher usage in those requiring intubation. Key predictors included low oxygen saturation (SpO₂), low respiratory rate (RR), and increased heart rate (HR). The predictive model achieved an accuracy of 94.4% (95% confidence interval (CI): 87.0-96.3), a recall of 81.0% (95% CI: 78.0-84.0) for respiratory depression, and an area under the curve (AUC) of 0.98 (95% CI: 0.95-0.99).
Conclusion: The study highlights critical clinical predictors of respiratory depression risk in opioid overdose patients and demonstrates the potential of machine learning models in enhancing early detection and intervention.
Keywords: Generative Artificial Intelligence; Machine learning; Opiate overdose; Precision Medicine; Respiratory insufficiency.