Development and validation of the Gynecologic Oncology Predictor Of Postoperative opioid use (GO-POP) model

Abstract

<h3>Objectives:</h3> With increasing use of minimally invasive surgical techniques and same-day discharges, the need to predict pain medication use on the day of surgery is critical to post-operative pain control, patient satisfaction and public health concerns. Our goal was to develop and externally validate a statistical model for predicting postoperative opioid use following gynecologic surgery in an increasingly minimally invasive surgical population, incorporating patient-reported surgical concerns. <h3>Methods:</h3> We prospectively enrolled 382 women scheduled for abdominal surgeries (including laparotomies and minimally invasive approaches [MIS]) in an academic gynecologic-oncology practice to a study to develop and validate a predictor of outpatient opioid pill use. Negative binomial, zero-inflated negative binomial, and ordinal regression models were developed using an initial cohort of 216 women enrolled from February 2018-March 2019 and followed for 6 weeks postoperatively. A total of 39 candidate predictors were considered, including preoperative assessments of anxiety and anticipated pain as well as intra-operative factors. The primary model outcome was the number of opioid pills used postoperatively. After fitting models using best predictors, all models were internally validated with 1000 bootstrap samples to obtain bias-corrected accuracy using ordinal concordance index and calibration curves. All models' predictions were then externally validated using a second cohort (n=166) enrolled from May 2019-February 2020. The best model (GO-POP) was updated by combining cohorts and an online calculator and nomogram were created. GO-POP included 7 predictors: age, total operating time, patient's self-reported pre-operative anticipated need for pain medication, whether pregabalin was administered preoperatively, patient educational attainment, smoking history and patients' self-reported preoperative anxiety. <h3>Results:</h3> Most surgeries were performed using MIS approaches (276; 72.3%) while 52 (13.6%) patients underwent planned laparotomies and 14.1% had planned MIS approaches converted to laparotomies. Median total number of opioid pills was 7 in the combined cohort. Nearly 40% of subjects used 0 pills after hospital discharge. After external validation, an updated combined model had the best fit, with ordinal concordance of 0.646 (95% CI:0.615, 0.679). Over 1/3rd of patients were either not planning to dispose of unused opioids or had not yet decided how to handle unused pills. A nomogram was created to facilitate clinical use of the model. <h3>Conclusions:</h3> A prospectively constructed and validated model accurately estimates the number of opioid pills used by women after gynecologic surgery. Use of the GO-POP nomogram enables providers to tailor prescriptions using information available immediately following surgery. More work is needing to educate patients on the importance of disposal of unused opioids post-operatively to protect our communities.