Comparing different chronic preoperative opioid use definitions on outcomes after spine surgery

Background Context No consensus exists for defining chronic preoperative opioid use. Most spine studies rely solely on opioid duration to stratify patients into preoperative risk categories. Purpose The purpose of this study is to compare established opioid definitions that contain both duration and dosage to opioid models that rely solely on duration, including the CDC Guideline for Prescribing Opioids for Chronic Pain, in patients undergoing spine surgery. Study Design This was a retrospective cohort study that used opioid data from the Tennessee Controlled Substance Monitoring Database and prospective clinical data from a single-center academic spine registry. Patient Sample The study cohort consisted of 2,373 patients who underwent elective spine surgery for degenerative conditions between January 2011 and February 2017 and who completed a follow-up assessment at 12 months after surgery. Outcome Measures Postoperative opioid use and patient-reported satisfaction (NASS Satisfaction Scale), disability (Oswestry/Neck Disability Index), and pain (Numeric Rating Scale) at 12 month follow-up. Methods Six different chronic preoperative opioid use variables were created based on the number of times a prescription was filled and/or daily morphine milligram equivalent for the one year before surgery. These variables defined chronic opioid use as 1) most days for > 3 months (CDC), 2) continuous use for ≥ 6 months (Schoenfeld), 3) >4,500 mg for at least 9 months (Svendsen wide), 4) >9,000 mg for 12 months (Svendsen intermediary), 5) >18,000 mg for 12 months (Svendsen strict), 6) low-dose chronic (1-36 mg for >91 days), medium-dose chronic (36-120 mg for >91 days), and high-dose chronic (>120 mg for >91 days) (Edlund). Multivariable regression models yielding C-index and R2 values were used to compare chronic preoperative opioid use definitions by postoperative outcomes, adjusting for type of surgery. Results Chronic preoperative opioid use was reported in 470 to 725 (19.8% to 30.6%) patients, depending on definition. The Edlund definition, accounting for duration and dosage, had the highest predictive ability for postoperative opioid use (77.5%), followed by Schoenfeld (75.7%), CDC (72.6%), and Svendsen (59.9% to 72.5%) definitions. A combined Edlund and Schoenfeld duration and dosage definition in post-hoc analysis, that included 3 and 6 month duration cut-offs, performed the best overall with a C-index of 78.4%. Both Edlund and Schoenfeld definitions explained similar amounts of variance in satisfaction, disability, and pain (4.2% to 8.5%). Svendsen and CDC definitions demonstrated poorer performance for patient-reported outcomes (1.4% to 7.2%). Conclusions The Edlund definition is recommended for identifying patients at highest risk for postoperative opioid use. When opioid dosage is unavailable, the Schoenfeld definition is a reasonable choice with similar predictive ability. For patient-reported outcomes, either the Edlund or Schoenfeld definition is recommended. Future work should consider combing dosage and duration, with 3 and 6 month cutoffs, into chronic opioid use definitions.