Predicting persistent opioid use after surgery using electronic health record and patient-reported data.

Area of Science:

• Medical Informatics
• Public Health
• Pain Management

Background:

• Over 90% of surgical patients receive opioid prescriptions, with a significant minority developing long-term use.
• This contributes to the ongoing national opioid epidemic, highlighting a critical public health concern.

Purpose of the Study:

• To develop and validate a predictive model for persistent opioid use following surgical procedures.
• To identify surgical patients at high risk for long-term opioid dependence.

Main Methods:

• A cohort of 24,040 surgical patients (≥18 years) was studied between 2015-2018.
• Electronic health records, state prescription drug monitoring data, and patient-reported measures were used to develop three prediction models (full, restricted, minimal).
• Persistent opioid use was defined as filling prescriptions between post-discharge days 4-180.

Main Results:

• The developed models demonstrated strong predictive performance, with C-statistics ranging from 0.85 to 0.87 in the validation cohort.
• Models performed better in patients with prior opioid use compared to opioid-naive patients.
• The restricted model showed a significant net benefit for identifying patients needing preoperative counseling.

Conclusions:

• Validated prediction models utilizing accessible data can effectively identify patients at risk for persistent postsurgical opioid use.
• These models offer improved performance over previous tools and can aid in targeted interventions to mitigate the opioid crisis.