The Association Between Preoperative Opioid Exposure and Prolonged Postoperative Use

Charles Katzman, BS; Emily C. Harker, BS; Rizwan Ahmed, BS; Charles A. Keilin, BS; Joceline V. Vu, MD; David C. Cron, MD, MS; Vidhya Gunaseelan, MS, MHA; Yen-Ling Lai, MS, MPH; Chad M. Brummett, MD; Michael J. Englesbe, MD; Jennifer F. Waljee, MD, MPH


Annals of Surgery. 2021;274(5):e410-e416. 

In This Article


Data Sources and Study Cohort

We analyzed claims from Optum's de-identifed Clinformatics Data Mart Database. This database captures data for members of a large national managed-care company containing patient-level data from administrative health claims such as prescription drug fills, medical services and procedures, inpatient and outpatient data, and other patient characteristics.

We identified adults between the ages of 18–64 who underwent either major or minor surgical procedures between January 1st, 2008 and March 31st, 2015. Major surgical procedures included ventral incisional hernia repair, colectomy, reflux surgery, bariatric surgery, and hysterectomy; and minor surgical procedures included were varicose vein removal, laparoscopic cholecystectomy, laparoscopic appendectomy, hemorrhoidectomy, thyroidectomy, transurethral prostate surgery, parathyroidectomy, and carpal tunnel release.[7] We identified patients undergoing surgery using Current Procedural Terminology (CPT) and International Statistical Classification of Diseases and Related Health Problems (ICD-9) procedure codes (Appendix 1, The preoperative period was defined as 365 days to 31 days before admission to remain consistent with previous studies.[7,16] Patients over the age of 64 were excluded to ensure homogeneity in the capture of prescription claims. We also excluded patients with a length of stay longer than 30 days, who had subsequent surgeries in the 180 days after discharge identified by claims related to anesthesia, and patients who were not discharged home after surgery.[7,16] We included only patients who had continuous medical and prescription drug coverage in the 12-month preoperative period and 6-month postoperative period to capture all study relevant patient information such as opioid use, comorbid conditions, and health care utilization. Additionally, we only included patients who filled an initial postoperative opioid prescription, defined as a prescription within the time interval from 30 days before surgery to 14 days after discharge from surgery. Because the study used de-identified patient information, it was deemed exempt from review by the University of Michigan Institutional Review Board.

Preoperative Opioid use

Preoperative opioid use was captured by insurance claims for opioid medications, and doses were converted to oral morphine equivalents (OMEs) based on standards defined by the Centers for Disease Control.[21] Patients who did not fill any opioid prescriptions during the preoperative period were considered opioid-naïve[7,17] and patients were described as preoperative opioid users if they had filled at least 1 prescription during the preoperative period. Fills within 1 month of the surgical procedure were assumed to be provided in anticipation for postoperative care, consistent with previous studies.[7]

To classify preoperative opioid use, we used cluster analysis as previously described to group patients based on 4 characteristics of opioid use in the year before surgery: dose, recency, duration, and continuity of fills.[16] Dose was defined as the total amount of OMEs filled. Recency was defined as the number of months since the last opioid prescription filled before surgery. Duration was defined as the total number months during which patients filled an opioid. Continuity was defined as the longest number of months in which the patient consistently filled an opioid prescription.[16] The 6 patterns of opioid use include: no use (opioid-naïve), minimal use, remote intermittent use, recent intermittent use, low chronic use, medium chronic use, and high chronic use. Individuals meeting criteria for chronic use were not included in this study to focus on individuals with nonchronic, periodic use. Opioid-naïve patients were included for comparison. Minimal use is defined as low dose short course of varying recency which analogous to a short course of opioids for an acute conditions before surgery. Intermittent use was further classified into 2 groups: remote intermittent use and recent intermittent use. Remote intermittent use is defined as low-dose, remote, short-course use, and recent intermittent is defined as low-dose, recent, medium-course use.[16] Specific parameters for each of the preoperative use groups in our study population can be found in the Result section, and Table 2.

Outcome and Covariates

Our primary outcome was postoperative persistent opioid use, defined as filling at least 1 opioid prescription between 91 and 180 days post-discharge. This definition is consistent with previous studies of persistent opioid use after surgery and was chosen based on the expected resolution of acute postoperative pain within 90 days of surgery.[16,22] The primary explanatory variable was preoperative opioid use group (opioid-naïve, minimal use, remote intermittent use, and recent intermittent use as described above). Secondary outcomes included the proportion of patients whose postoperative dose exceeded the preoperative dose, and the correlation between type of opioid and probability of persistent opioid use.

We included other patient-level covariates such as age, sex, race/ethnicity, education, and major versus minor surgery (binary variable). The medical comorbidities were determined using claims in the year before admission for surgery and categorized using the Charlson comorbidity index.[23] Tobacco use, comorbid conditions, and co-existing pain conditions were identified using the ICD-9 codes. Pain disorders were further categorized as arthritis and joint pain, back pain, neck pain, and other pain disorders (Appendix 2, Mental health disorders were subdivided into adjustment, anxiety, mood, suicide or self-harm, disruptive, personality, psychosis, alcohol or substance abuse disorders, or other mental disorders (Appendix 3, using the Clinical Classification System from the Agency of Healthcare Research and Quality. The total OMEs of the initial postoperative dose were stratified into less than 25th percentile (Q1), median (Q2), and greater than or equal to the 75th percentile (Q3).

Statistical Analysis

We used descriptive statistics to characterize both patient demographics and clinical characteristics of the study cohort. We then used a multivariable logistic regression model to determine the association of preoperative opioid use group with the likelihood of developing persistent use after surgery whereas controlling for patient factors. The clustering analysis to determine preoperative opioid use groups were performed using R version 3.5.1 (Vienna, Austria), and other analyses were conducted using Stata/SE version 15.1 (Stata Corp., College Station, TX) and SAS version 9.4 (SAS Institute, Cary, NC). P-values were 2-tailed and statistical significance was set at P < 0.05.