Diabetes Diagnosis Patterns in Medicaid

OBJECTIVE: To describe patterns of type 2 diabetes diagnosis prevalence in Medicaid administrative data by individual and contextual factors. METHODS: Diagnosed prevalence was estimated using the 2016-2021 Transformed Medicaid Statistical Information System (T-MSIS) and a Bayesian multilevel regression framework. The first level included patient characteristics (age, sex, and race/ethnicity), the Social Vulnerability Index (SVI), plan-level characteristics, payment models, and federal level of poverty (FLP) eligibility thresholds. The second level incorporates partial pooling over the state to model state fixed effects. Adults ages 21-64 years with 24+ months consecutive enrollment were included. States with a high degree of missing data were excluded. RESULTS: Among 4.41 million eligible Medicaid beneficiaries in N=11 states with high-quality T-MSIS data, 10.3% had a type 2 diabetes diagnosis record. State-level diagnosed prevalence based on administrative data ranged from 6.5%-13.0%, and increased with age (4.1% for 21-35 vs. 23.9% for 55-64 y), was higher among "Other" race (14.3%), Hispanic (12.0%), and Black (10.9%) enrollees versus White (9.1%). Higher county-level obesity was associated with increased odds (third vs. first tertile OR 1.26, 95% CI: 1.25-1.27), while living in urban areas (OR 0.91, 95% CI: 0.90-0.91) and in the highest SVI quartile (OR 0.74, 95% CI: 0.73-0.75) with lower odds. Race interactions were most pronounced for contextual socioeconomic modifiers (SVI, FPL, and plan type). CONCLUSIONS: Evidence generated confirms that odds of having a recorded type 2 diabetes diagnosis are shaped by age, comorbidities, race/ethnicity, and social vulnerability. Notable variations by state and care plans were identified. Documentation and data improvements are needed.