What predicts postoperative opioid prescribing and consumption? A statewide analysis of surgical quality registry data.

OBJECTIVE: Tailoring postoperative opioid recommendations to patient needs requires nuanced understanding of factors contributing to post-discharge opioid use. The study aims to identify key predictors of opioid prescribing and consumption while exploring the interplay between clinical factors that underlie these phenomena. DESIGN: We analyzed Michigan Surgical Quality Collaborative registry from 2017- 2019 to identify factors predicting sequential opioid-related outcomes following surgery: 1) prescription receipt, 2) likelihood of consumption, and 3) amount consumed. METHODS: To enhance predictive accuracy, we used a three-part model applying machine learning methods (random forests, support vector machines, extreme gradient boosting) and ranking predictive factors by variable importance scores. RESULTS: : Among 34,505 patients (57% female, mean age 56 years), 10,572 (31%) received no prescription, 6,069 (18%) received a prescription but reported no opioid consumption, and 17,864 (52%) received a prescription and reported some consumption. The most important factors predicting prescription receipt included younger age, procedure type, inpatient/outpatient location, urgent/emergent status, and higher body mass index (BMI). Top factors for likelihood to consume included younger age, prescription quantity, higher BMI, smoking, and procedure type. For amount consumed, prescription quantity was the most important factor, with lesser contributions from preoperative opioid prescriptions, younger age, surgery type, and smoking. CONCLUSIONS: : These findings suggest an overlapping set of key factors of age, procedure type, BMI, prescription quantity, and smoking influence post-discharge opioid use. These factors may help identify patients at higher risk for post-discharge use and inform targeted opioid stewardship strategies, including interventions focused on modifiable factors such as prescription quantity.