Benchmarking the Transformed Medicaid Statistical Information System Analytic Files for analysis of opioid use disorder treatment
本研究评估了医疗补助理赔数据在阿片类药物使用障碍治疗方面的一致性和质量，将转型后医疗补助统计信息系统分析文件与11个州直接获取的数据进行对比，发现TAF数据在患者计数上略低但整体指标一致，适合用于相关研究。
OBJECTIVE: To evaluate the data consistency and quality of Medicaid claims data on opioid use disorder (OUD) treatment, benchmarking the Transformed Medicaid Statistical Information System Analytic Files (TAF) against metrics reported by the Medicaid Outcomes Distributed Research Network (MODRN), which analyzed data obtained directly from 11 state Medicaid agencies. DATA SOURCES AND STUDY SETTING: The primary data source was TAF claims for the years 2017 and 2018, limited to non-dual, full-benefit Medicaid beneficiaries aged 12-64 in one of the 11 states in our study sample. STUDY DESIGN: Using TAF data, we replicated performance on the following five OUD quality metrics reported by MODRN: number of enrollees receiving any OUD medication, receipt of behavioral health counseling, completion of urine drug screens, receipt of any opioid analgesic fill, and receipt of any benzodiazepine fill. DATA COLLECTION/EXTRACTION METHODS: Access to TAF data was facilitated through the Chronic Conditions Warehouse. PRINCIPAL FINDINGS: There were 11% fewer Medicaid enrollees with OUD in TAF compared to MODRN (912,478 vs. 1,034,412). Patient characteristics were largely similar across the two datasets, with the exception of more missing race information in TAF (20.9% vs. 7.1%). Across the 11 states, performance on the six quality measures was similar. For example, the rate of use of any OUD medication in 2018 was 57.1% in MODRN and 58.6% in TAF. However, there were important discrepancies in the TAF data in individual states for single years. CONCLUSIONS: TAF data may undercount patients with OUD, but otherwise exhibited consistency with MODRN benchmarks, suggesting suitability of TAF for research on OUD treatment. Our results highlight several data quality issues with TAF that researchers who use these data should be aware of, including reporting of race and ethnicity.