Back to 2026 ePosters
Pre-surgical Prediction of Post-operative Opioid Overuse in Hand Surgery Patients
Richard Hum, MS; Aviram M. Giladi, MD, MS
The Curtis National Hand Center, Baltimore, MD
IntroductionWe hypothesize that higher initial opioid prescription amounts predict prolonged opioid use at six weeks and three months post-operatively. Additionally, certain patient characteristics, surgery type, and pre-operative patient reported outcome scores (PROs) are also associated with long term opioid use and may inform unique predictive models to help identify high-risk patients before surgery.
Materials and MethodsThis study examined the relationship between initial opioid prescription and long-term opioid use. Data collected at point-of-care from our clinic in 2018-2022 were randomly split into training and test sets (80% and 20%). Quantile regression categorized patients into high or low initial opioid prescription groups based on total morphine milligram equivalents (MME) and adjusting for key demographics. Independent variables for our final models included procedure type (elective or trauma), procedure depth (soft tissue only vs. bone/joint), age, BMI, sex, and history of opioid use. Multivariable logistic regression was performed to predict post-operative opioid use at six weeks and three months, incorporating demographic, clinical, and patient reported outcome variables, including high vs. low initial prescription status from the quantile model. Stepwise logistic regression across 15 imputed datasets generated pooled odds ratios and 95% confidence intervals. Model performance was assessed using receiver operating (ROC) and precision-recall curves (PRC).
ResultsAmong 12,117 patients, those receiving high initial opioid doses - controlling for various factors known to be associated with opioid use - had significantly greater odds of continued opioid use at six weeks and three months post-op. Additional risk factors included higher Charlson Comorbidity Index (CCI), greater pre-operative pain, lower pre-operative PROMIS Global Physical Health scores (GPH), opioid/marijuana/medication history, and post-operative antibiotic use. Black and other race patients, and those with Medicare/Medicaid, had higher odds of prolonged opioid use than white patients and those with commercial insurance. Predictive models performed moderately well, with PRC AUCs of 0.196 (six weeks) and 0.135 (three months) in the training set, and 0.224 and 0.157, respectively, in the test set.
ConclusionsCombining prescription amount and patient-reported data with routinely captured electronic-health-record variables yields a predictive algorithm that accurately flags hand-surgery patients at risk for prolonged post-operative opioid use. If prospectively validated, embedding this tool into clinical workflows may enable targeted counselling, proactive multimodal analgesia, and safer, data-driven opioid stewardship across high-volume practices.
Back to 2026 ePosters
