Analysis and development of risk prediction models for chronic opioid use after surgery: a cohort study using the nationwide database

Abstract

Background Chronic opioid use has become a socioeconomic as well as a medical problem. This study aimed to identify risk factors and develop prediction models for postoperative chronic opioid use (PCOU). Methods This retrospective cohort study used data from the Korean National Health Insurance Service (NHIS) between January 2008 and December 2018. Of 2 077 825 patients aged seven years or older who underwent surgery, survived at least one year, and had no additional surgeries, 1 108 119 were randomly selected. Logistic regression (LR) and machine learning models were developed to identify risk factors for PCOU. PCOU was defined as having filled 10 or more prescriptions or receiving more than 120 days’ supply between postoperative days 91 and 365. Age, sex, medical comorbidities (systemic diseases, psychological disorders, and substance use disorders), preoperative medications (antidepressants, antipsychotics, anticonvulsants, benzodiazepines, opioids, and nonopioid analgesics), and type of surgery were assessed as potential risk factors. Results PCOU occurred in 9308 patients (0.84%). Older age, preoperative history of opioid use, and high in-hospital opioid doses were the three most important predictors. Among the 28 most commonly performed surgical procedures in Korea, lung surgery, general spinal surgery, and total knee arthroplasty were most strongly associated with chronic opioid use. Conclusions According to the best-performing gradient boosting model, older age, longer hospital stay, high in-hospital opioid consumption, and preoperative opioid use were the most important risk factors for PCOU.