(HealthDay News) — Machine learning can identify patients who are at risk for persistent opioid use after surgery, according to a study published in the September issue of Plastic and Reconstructive Surgery.
Natalie B. Baxter, from the University of Michigan Medical School in Ann Arbor, and colleagues trained two algorithms to predict persistent opioid use using a general surgery data set and a hand surgery data set, resulting in four trained models for adult surgery patients. Each model’s performance was tested using hand surgery data. The first algorithm was the Michigan Genomics Initiative model, which accommodates patient-reported data and includes patients with or without prior opioid use; the second algorithm was the claims model designed for insurance claims data from opioid-naive individuals.
Data were included for 889 hand surgery patients; 49% were opioid-naive and 21% developed persistent opioid use. The researchers found that most of the patients underwent soft-tissue procedures or fracture repair (55 and 20%, respectively). Areas under the receiver operating curve (AUCs) were 0.84 and 0.85 when trained on hand surgery data and the full cohort of surgery patients, respectively, for the Michigan Genomics Initiative model. AUCs were 0.69 and 0.52 when trained only on hand surgery data and opioid-naive surgery patients, respectively, for the claims model.
“Considering the challenge of providing patient-tailored pain control regimens after surgery, applied machine learning has the potential to decrease provider burden and enhance quality of care,” the authors write.
Financial ties to Sonex and Teva Pharmaceuticals were disclosed.
