Applying decision curve analysis to evaluate a patient-centered opioid prescribing strategy among patients with laparoscopic cholecystectomy in the US.

Abstract

With evolving prescription policies, the clinical utility of patient-centered opioid prescribing remains unclear. We evaluated whether a patient-centered opioid prescribing approach improves outcomes compared to default strategies: "treat all" (emulating opioid over-prescribing practices in the late 1990s and 2000s) or "treat none" (reflecting prescribing restrictions in Centers for Disease Control and Prevention 2016 guideline). We conducted a decision curve analysis using electronic medical data from 1136 adult patients undergoing laparoscopic cholecystectomy for benign indications (October 2015 - December 2018) at an urban tertiary care center. Perioperative opioid needs were categorized as "none/low" or "medium/high" based on pain scores and opioid prescriptions/administrations from 30 days pre-operative to two weeks postoperative. Of 803 patients in the training dataset, 63.9 % required medium/high opioid levels. Key predictors included sex, preoperative medication use (opioids, acetaminophen, other nonsteroidal anti-inflammatory drug, antidepressants), emergent surgery, anesthesia type, and surgical indications. The prediction model's c-statistic was 0.65 (95 %CI: 0.58-0.71) and Brier score was 0.20 (95 % CI: 0.009-0.71) in the testing dataset. At an optimal threshold probability of 0.60, the prediction model correctly identified 16 more true positive cases per 100 patients needing medium/high opioid prescriptions than the "treat all" strategy, accounting for false positives. Across a threshold range of 0.18-0.91, the model-based strategy constantly outperformed both default strategies. Prediction-based, patient-centered opioid prescribing provides greater clinical utility across most thresholds.