Predictive Models for Time to First Opioid Use Disorder or Opioid Overdose Among Older Adults

Area of Science:

• Gerontology
• Public Health
• Data Science

Background:

• Limited data exists on predictive models for opioid use disorder (OUD) or opioid overdose (OD) in older adults.
• Identifying at-risk individuals is crucial for early intervention and prevention strategies.

Purpose of the Study:

• To develop and validate predictive models for time to OUD or OD in older adults.
• To identify key predictors of OUD or OD using patient-reported and claims-based data.

Main Methods:

• A prognostic study utilized Health and Retirement Study (HRS) data linked with Medicare claims (2006-2021).
• Included older adults (≥65 years) with chronic pain and prior opioid prescriptions.
• Four survival models (Cox, LASSO-penalized Cox, survival random forest) were employed to predict time to OUD or OD.

Main Results:

• 181 out of 4190 older adults developed OUD or OD during follow-up.
• All four models demonstrated comparable predictive performance (C-statistics ranging from 0.723 to 0.849).
• Top predictors identified were duration of opioid use, uncontrolled pain, and concurrent use of central nervous system medications.

Conclusions:

• Both traditional Cox and machine learning models effectively predicted OUD/OD risk in older adults.
• These models, using diverse data sources, can aid in monitoring and identifying high-risk individuals.
• Early identification facilitates timely interventions to mitigate OUD and OD.