Predicting Chronic Opioid Use Among Patients With Osteoarthritis Using Electronic Health Record Data
Authors' conclusion
We found that EHR-based predictive models could estimate the risk of future COU among patients with OA to help inform care decisions.
Study context
- Population
- Patients with osteoarthritis (OA) with a new opioid prescription between March 1, 2017 and February 28, 2019 and no opioid use in the prior 6 months, electronic health record (EHR) data from 13 health care organizations
- Participants
- 33,894 patients
- Comparison
- 4 machine learning models vs simplest regression model
- Outcome
- Chronic opioid use (≥3 prescriptions ≥84 days, maximum gap ≤60 days); discrimination of prediction models (AUC)
- Follow-up
- 1 year
Methodological detailsmulticentre (13 health care organizations), machine learning, transportability to held-out organizations
- Multicentre (13 health care organizations)
- Machine learning
- Transportability to held-out organizations
Performance
As reported in the abstract
Discrimination for COU (AUC)
| Test or model | Value | AUC |
|---|---|---|
| Random forest | 0.728 (95% CI 0.711–0.745) | 0.728 |
| Simplest regression model | 0.717 (95% CI 0.699–0.734) | 0.717 |
Value
Random forest0.728 (95% CI 0.711–0.745)
Simplest regression model0.717 (95% CI 0.699–0.734)
AUC
Random forest0.728
Simplest regression model0.717
Further findings
Incidence of chronic opioid use (COU) within 1 year
- Developed COU
- 2,925 (8.6%)
Calibration: predicted risk deciles (well-calibrated models)
- 10th percentile predicted risk
- 2%
- 90th percentile predicted risk
- 18%
Discrimination across organizations (transportability)
- AUC range across organizations (highly variable)
- 0.571-0.842
Original publication
Predicting Chronic Opioid Use Among Patients With Osteoarthritis Using Electronic Health Record Data. Arthritis Care Res (Hoboken). Published 28 December 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →