Impact of integrated Medicare-Medicaid plans for dual eligibles with serious mental illness: a mixed-methods study

• In Event: Building and Assessing the Credibility of Causal Inference in Policy Evaluation

Abstract

Introduction: Fragmentation between Medicare and Medicaid exacerbates access barriers, care delays, and poor financial and health outcomes among the over four million dual eligibles with serious mental illness (SMI). Integrated dual eligible special needs plans (D-SNPs), including highly integrated D-SNPs (HIDE-SNPs) and fully integrated D-SNPs (HIDE-SNPs), are Medicare Advantage (MA) plans that financially integrate Medicare and Medicaid benefits and may address the significant behavioral health, medical, and long-term services and supports needs in this population.

Aims: To study the impact of enrollment in integrated D-SNPs on care quality and outcomes among dual eligibles with SMI using a mixed-methods approach integrating target trial emulation and qualitative interviews with implementation leaders. 

Methods: Using novel linked data capturing the entirety of Medicare and Medicaid services for 20% of dual eligibles with SMI from 2018-2023, we use a concurrent-embedded mixed-methods design integrating a quantitative trial emulation approach with qualitative study of D-SNP implementation. Qualitative data collection includes focus groups (completed at time of abstract submission) and individual interviews (ongoing at time of abstract submission) with ~60 D-SNP implementation leaders including representatives of state Medicaid agencies, health insurance companies offering D-SNP plans, professional organizations representing providers & D-SNPs (e.g., D-SNP Alliance), and state and national mental health and Medicare/Medicaid advocacy organizations. 

Results: The study cohort included 719,542 dual eligibles with a diagnosis of SMI. The proportion of dual eligibles with SMI enrolled in integrated HIDE- or FIDE-SNPs increased from 2.4% in 2018 to 17.2% in 2023, with a corresponding decline in traditional Medicare enrollment from 58.6% to 42.6%. After covariate adjustment, HIDE-SNP enrollment was associated with lower likelihood of all-cause hospitalization relative to traditional Medicare (-6.8 percentage points [pp]; 95% CI, -8.2 to -5.4), and standard MA (-2.1 pp; 95% CI, -3.2 to -1.0) as well as lower likelihood of psychiatric hospitalization relative to traditional Medicare  (-2.5 pp; 95% CI, -3.3 to -1.7) and standard MA (-1.7 pp; 95% CI, -2.9 to -0.4). Both HIDE- and FIDE-SNP enrollment were associated with lower likelihood of 30-day readmission compared to traditional Medicare. Qualitative findings emphasizing the importance of state Medicaid program design for D-SNP implementation were used to refine covariate adjustment for causal inference. In addition, qualitative findings suggest that integrated D-SNPs conduct case management focused on reducing preventable acute care use and proactively coordinate Medicaid-covered behavioral health services for dual eligibles with SMI; these implementation strategies may help to explain quantitative results. 

Conclusions: Integrated D-SNPs, particularly HIDE-SNPs, were associated with lower overall and preventable acute care use, suggesting that Medicare-Medicaid integration may reduce costly acute care use in this population. Qualitative findings helped to refine the quantitative methods to improve causal inference and to interpret results.

Author

• Beth McGinty, Cornell University
  Presenting Author