Treatment-Effect-Based Versus Risk-Based Targeting of Care Management Outreach in Medicaid: A Retrospective Cohort...

Area of Science:

• Health Services Research
• Causal Inference
• Health Economics

Background:

• Medicaid care management often targets high-risk beneficiaries for outreach.
• This strategy assumes risk and responsiveness are aligned and stable.
• Limited outreach capacity necessitates efficient targeting methods.

Purpose of the Study:

• To compare the effectiveness of targeting outreach by predicted individualized treatment effect (CATE) versus risk-based targeting.
• To evaluate if monthly re-estimation of CATE improves acute event prevention in Medicaid programs.
• To assess the impact of targeting strategies on acute events and disparities.

Main Methods:

• Analysis of 164,063 adult Medicaid beneficiaries in Washington and Virginia (Jan 2023-Dec 2025).
• Causal forest with cross-fitted nuisance functions and within-person fixed effects to estimate CATE.
• Comparison of risk-based vs. effect-based targeting rules at 10% capacity using off-policy evaluation.

Main Results:

• Effect-based targeting prevented 13.3 acute events per 2000 members/month, compared to 2.5 for risk-based targeting (a 5.3-fold improvement).
• Within-person variance explained 63.6% of the total CATE variance, highlighting individual responsiveness.
• Gains in event prevention were consistent across racial/ethnic groups and states.

Conclusions:

• Monthly allocation of Medicaid care-management outreach by predicted treatment effect significantly increases acute events prevented.
• This effect-based targeting strategy is more efficient than risk-based targeting at fixed capacity.
• The approach substantially improved outcomes without widening between-group disparities.