Postoperative Opioid Use Prediction Calculator Performance in Patients Undergoing Gynecologic Surgery

Importance Postoperative opioid overprescription contributes to the opioid epidemic. Objective To expand and validate a gynecologic oncology–developed postoperative opioid use prediction calculator in a diverse surgical population and evaluate whether it could be modified for preoperative prescribing. Design, Setting, and Participants This prognostic study in a quaternary health system used patient preoperative surveys and postoperative assessments to measure opioid use and pain scores from November 1, 2023, to November 1, 2024, with follow-up up to 6 weeks. Patients aged 18 years or older who underwent gynecologic surgery were eligible; those undergoing only minor procedures were excluded. Data were analyzed from January 15, 2025, through July 27, 2026. Main Outcomes and Measures The original 7-predictor model was externally validated. The prespecified updated proportional odds model included anticipated pain medication use, restricted cubic splines for age (3 knots) and anticipated operating time (5 knots), education, smoking, and anxiety. A total of 71 imputation-specific models were fitted. The outcome variable was the number of pills used. Validation methods included model discrimination and calibration. Results Of 884 identified patients, 870 were eligible, 835 underwent surgery, and 616 had complete opioid use outcome data (median [IQR] age, 48 [39-65] years). The median number of pills used was 2 (IQR, 0-7 pills), and 226 patients (36.7%) used no pills. For the original model, threshold-specific C statistics were 0.70 (95% CI, 0.66-0.74), 0.72 (95% CI, 0.68-0.76), and 0.74 (95% CI, 0.70-0.78) for use of at least 2, 5, and 10 pills, respectively. Calibration intercepts were −0.11 (95% CI, −0.30 to 0.08), −0.35 (95% CI, −0.54 to −0.17), and −0.51 (95% CI, −0.71 to −0.31), respectively, indicating overpredicted use at the higher thresholds of pill use. For the updated model, optimism-corrected C statistics were 0.71 (95% CI, 0.66-0.74), 0.73 (95% CI, 0.68-0.76), and 0.73 (95% CI, 0.67-0.77); calibration intercepts were −0.05 (95% CI, −0.09 to −0.01), −0.04 (95% CI, −0.08 to 0.002), and −0.03 (95% CI, −0.07 to 0.01); and calibration slopes were 0.93 (95% CI, 0.80-1.02), 1.06 (95% CI, 0.89-1.19), and 1.15 (95% CI, 0.85-1.37) for at least 2, 5, and 10 pills, respectively. Conclusions and Relevance This prognostic study of patients who underwent major gynecologic surgery found that the original calculator showed moderate threshold-specific discrimination but overpredicted opioid use at the higher thresholds of pill use. The updated model showed moderate discrimination after internal validation, with calibration intercepts near 0 and slopes ranging from 0.93 to 1.15. External institutional validation is needed. Individualized estimates may support opioid stewardship.