Objective:

To develop prediction models for identifying older adults at risk of opioid use disorder (OUD) or overdose (OD) using patient-reported and claims-based measures.

Approach:

• Study Design: A prognostic study using the Health and Retirement Study (HRS) with linked Medicare claims data from 2006 to 2021.
• Data Sources: Utilized biennial HRS surveys for patient-reported factors and Medicare data for capturing OUD and OD events and prescription opioid use.
• Study Sample: Included HRS-Medicare participants aged ≥ 65 years with chronic pain and opioid prescriptions, excluding those with prior OUD or OD.
• Outcomes: Primary outcomes were incident encounters of OD or OUD defined by specific ICD codes and a conservative algorithm for OUD.

Key Findings:

• The age-adjusted rate of deaths involving synthetic opioids increased 14-fold among older adults from 2000 to 2020.
• High-risk opioid use occurs in less than 5% of older adults before OUD or OD onset.
• Prior predictive models primarily focus on younger populations and may not apply to older adults due to differing risk factors.

Interpretation:

Existing predictive models for OUD or OD may not adequately account for the unique risk factors and time-varying features relevant to older adults.

Limitations:

• Previous models relied heavily on administrative claims data, lacking comprehensive patient assessments.
• No prior studies utilized survival models or accounted for changing predictors over time.

Conclusion:

This study aims to fill the gap in predictive modeling for OUD and OD in older adults by utilizing a comprehensive dataset and focusing on relevant risk factors.