Comparing Approaches to Identify Mobile Crisis Services in Medicaid Administrative Claims

BACKGROUND: Mobile crisis services offer a community-based alternative to law enforcement and emergency departments for individuals experiencing behavioral health crises. Medicaid claims represent a promising but underused tool for national surveillance of these services. Despite growing federal investment and expansion of crisis response systems, there is no standardized method for identifying mobile crisis encounters in administrative claims, limiting the ability to monitor access, evaluate implementation, or compare delivery across states. OBJECTIVE: To develop and evaluate claims-based approaches for identifying mobile crisis service delivery in Medicaid administrative data. METHODS: We conducted a cross-sectional analysis of 2021 Medicaid claims from the Transformed Medicaid Statistical Information System Analytic Files. We used the 2022 KFF Behavioral Health Services Survey data as the most recent standardized benchmark for Medicaid mobile crisis coverage. We tested 3 claims-based identification approaches that combine national and state-specific procedure codes (eg, H2011) with mobile-relevant place-of-service (POS) codes: strict [POS 15 (mobile unit)], moderate [POS 15 (mobile unit), POS 12 (home), or POS 04 (homeless shelter)], and inclusive [POS 15, POS 12, POS 04, or POS 99 (other)]. We additionally constructed a claims-based score summarizing state-level use of mobile-relevant POS codes and examined its alignment with reported coverage in exploratory analyses. RESULTS: Among 44 states and the District of Columbia with complete survey data, 33 reported Medicaid coverage of mobile crisis services. We identified 97,088 claims under the strict approach, 395,141 under the moderate approach, and 1,152,653 under the inclusive approach. Across all approaches, states reporting coverage had higher median claim counts than noncoverage states, though substantial overlap remained. Under the strict approach, median claim counts were 0 in both coverage and noncoverage states. The moderate approach demonstrated the greatest separation (median, 1324 vs. 364), while the inclusive approach captured larger volumes overall but with greater overlap between groups. Predicted probabilities derived from a claims-based score increased monotonically across score values, indicating stronger directional alignment with reported coverage at higher score levels. CONCLUSION: A claims-based approach combining mobile unit (POS 15), home (POS 12), and homeless shelter (POS 04) codes with crisis intervention procedure codes provides a practical framework for identifying billed mobile crisis encounters in Medicaid data. This moderate approach showed the strongest directional alignment with state-reported coverage and may support cross-state monitoring of crisis service delivery. Further work is needed to validate these methods using provider-level or encounter-level data.