Changes in predicted opioid overdose risk over time in a state Medicaid program: a group-based trajectory modeling...

Area of Science:

• Public Health
• Data Science
• Pharmacology

Background:

• Healthcare data access delays hinder opioid overdose prediction implementation.
• Understanding longitudinal changes in individual overdose risk is limited.

Purpose of the Study:

• To identify longitudinal patterns of predicted opioid overdose risk.
• To analyze risk trajectories among Medicaid beneficiaries initiating opioid prescriptions.

Main Methods:

• Retrospective cohort study of Pennsylvania Medicaid beneficiaries (aged 18-64).
• Utilized a validated machine-learning algorithm to calculate 3-month interval overdose risk scores.
• Employed group-based trajectory modeling to identify risk patterns over time.

Main Results:

• 0.61% of beneficiaries experienced an opioid overdose within 15 months.
• Five distinct overdose risk trajectories were identified.
• 92% of beneficiaries exhibited consistent risk levels (low, medium, or high) over time.
• 8% showed significant risk changes, either decreasing from high-to-medium or increasing from medium-to-high risk.

Conclusions:

• Over 90% of beneficiaries demonstrated consistent opioid overdose risk over 15 months.
• Machine learning prediction algorithms are likely practical for most individuals.
• Longitudinal risk assessment can inform targeted overdose prevention strategies.