A Risk Prediction Model for Long-term Prescription Opioid Use.

Area of Science:

• Clinical Informatics
• Pharmacovigilance
• Health Services Research

Background:

• Tools are needed to help clinicians predict patients' risk of long-term opioid use.
• Accurate risk prediction can inform opioid prescribing decisions.

Purpose of the Study:

• To develop and validate a predictive model for long-term opioid use in previously opioid-naive patients.
• To assess the model's performance in discrimination, calibration, and clinical utility.

Main Methods:

• Statewide, population-based prognostic study using data from the California Prescription Drug Monitoring Program (PDMP).
• Developed a multiple logistic regression model using 2016-2017 data for patients aged 12+ receiving opioid analgesics.
• Validated the model using 2018 data, evaluating discrimination (c-statistic), calibration, and clinical utility (decision curve analysis).

Main Results:

• The development and validation cohorts included over 7 million and 2.7 million opioid-naive patients, respectively.
• The model demonstrated high discrimination (c-statistic: 0.904 development, 0.913 validation) and was well-calibrated.
• Decision curve analysis indicated significant clinical utility across a range of probability thresholds.

Conclusions:

• A predictive model for transition to long-term opioid use was developed and validated with high accuracy.
• The model's strong predictive performance suggests potential for integration into PDMPs to support clinical decision-making.
• This tool can aid clinicians in making informed opioid prescribing decisions at the point of care.