The following is a summary of “Deep learning predicts postoperative opioids refills in a multi-institutional cohort of Surgical Patients,” published in the May 2024 issue of Surgery by Salehinejad et al.

Efforts to address the opioid epidemic have led to strategies aimed at reducing unnecessary postoperative opioid prescriptions. However, this approach can result in inadequate pain management for some patients who genuinely need additional opioids. Deep learning models are promising to enhance healthcare delivery by focusing on patient-centered outcomes. This study explores the potential of deep learning models to predict which patients will require additional opioid prescription refills following elective surgery.

This retrospective study analyzed patients who underwent elective surgeries at the Mayo Clinic from 2013 to 2019. The study included adult patients aged 18 and older who spoke English. Several machine learning models, including deep learning, random forest, and eXtreme Gradient Boosting (XGBoost), were developed to predict the need for postoperative opioid refills upon discharge.

The study included 9,731 patients with a mean age of 62.1 years, of which 51.4% were female. Deep learning and random forest models demonstrated high accuracy in predicting the need for opioid refills, with accuracy rates of 0.79 ± 0.07 and 0.78 ± 0.08, respectively. The primary predictors for requiring opioid refills post-discharge were the type of procedure performed, the highest pain score recorded during hospitalization, and the total oral morphine milligram equivalents prescribed at discharge.

Deep learning models are effective in predicting which patients will need additional opioid prescriptions after surgery, achieving high accuracy. Moreover, other machine learning models, such as random forest, perform comparably to deep learning, underscoring the broad applicability of machine learning techniques in addressing the opioid crisis. These findings suggest that integrating such predictive models into clinical practice could help tailor opioid prescribing practices to individual patient needs, thereby improving pain management and reducing the risk of opioid misuse.

Source: sciencedirect.com/science/article/abs/pii/S0039606024002198