Supporting Rising-Risk Medicaid Patients Through Early Intervention

Rising-risk patients receiving Medicaid experience worsening medical and behavioral health conditions and increased acute care (ED and hospital) utilization, while typically being disconnected from primary care. Randomized controlled studies show that community-based early interventions outside brick-and-mortar clinical settings improve outcomes for these patients before they become high utilizers of acute care services. Yet limited data and infrastructure support, inconsistent grant-based funding, and high administrative burdens on staff have limited the accessibility and reproducibility of successful early community-based interventions for patients receiving Medicaid. The authors report on the launch of operational infrastructure and enabling technology across two health plans and multiple providers in the states of Washington and Virginia to introduce community-based teams providing early intervention services to rising-risk patients receiving Medicaid. The teams of community health workers, care coordinators, social workers, and pharmacists provided 1,652 patients receiving Medicaid with interventions included closing gaps in care, assisting patients in following health care provider instructions, providing cognitive behavioral therapy, assisting with prescription medication access and adherence, and arranging and accompanying patients to social service and health care appointments. Using data synthesis efforts to assist teams with streamlined outreach and communications, and generative AI tools - such as prior authorization writers and social service matching models - reduced administrative intake time for intervention teams by 34.2% (from 38 to 25 minutes per intake), facilitated improvement of Health Effectiveness Data and Information Set (HEDIS) quality metrics by an average of 11.8 absolute percentage points (to 43.8% from 32.0% of care gaps closed in year one of implementation), and enabled patients to achieve 63.3% of their health care and social goals. Using novel rising-risk machine learning models to optimize patient selection and outreach timing, the community-based early intervention teams achieved a 22.9% reduction (95% confidence interval [CI]: 13.4%-31.4%; P<0.0001) in all-cause acute events (to 169.9 from 220.4 per 1,000 patient-months), including a 48.3% reduction in ambulatory care-sensitive hospitalizations (95% CI: 26.5%-63.7%; P<0.0001), to 10.0 from 19.4 per 1,000 patient-months, and 20.4% in ambulatory care-sensitive ED visits (95% CI: 6.4%-32.4%; P=0.006), to 61.0 from 76.6 per 1,000 patient-months, among patients receiving the intervention, versus a matched control group of nonintervened patients studied in a difference-in-differences analysis to control for secular trends and unmeasured differences between the groups. Addressing clinical and social service gaps between primary care practices and health plans provided critical but narrow windows of opportunity to improve patient outcomes.