Medicaid Managed Care: CMS’s Accuracy Check Process For State-Submitted MLR Data Has Limitations
Learn how the AI-generated research projects were createdOverall Conclusion
CMS’s oversight of State-submitted Medicaid MLR data has limitations due to reliance on optional data elements and lack of internal communication; while CMS has taken steps such as releasing an MLR Toolkit, implementing a web-based submission portal, and making MLR data public, the 2019 testing identified potential inaccuracies and missed follow-ups, indicating a need for enhanced procedures and internal sharing to improve data accuracy for capitation rate development.
Source Document
Audit Scope
Scope: This audit evaluated CMS's oversight activities to verify State-submitted MLR data accuracy, focusing on 2019 MLR reporting year with 595 final MLR summary reports from 44 States, and follow-up practices; data sources included CMS's five accuracy checks and CMS follow-up on missing data, with data requests spanning December 2023.
Key Findings Summary
CMS’s accuracy checks of the four required MLR data elements were limited because states often did not submit optional MLR data elements, preventing CMS from conducting the checks for more than half of the final MLR summary reports.
CMS did not follow up with states that submitted potentially inaccurate MLR data.
Results of CMS’s accuracy checks were not shared across CMS staff responsible for reviewing capitation rates, limiting use of findings to inform rate development.
View the Findings tab to see all 6 findings
AI-Assisted
AI Scope Summary
This audit assesses CMS’s oversight activities to verify the accuracy of State-submitted Medicaid MLR data, focusing on the 2019 MLR reporting year and CMS’s five accuracy checks; it concludes that CMS conducted limited accuracy checks, did not follow up on identified potential inaccuracies, and did not share findings across relevant CMS staff, informing future audits to emphasize cross-team data sharing, internal procedures for material differences, and automated front-end checks to improve MLR data quality for capitation rate development.
AI-Generated Insight
The report highlights a data quality oversight gap: accuracy checks are incomplete because optional data elements are not consistently submitted, and findings are not shared across CMS teams, reducing their utility for rate setting. There are clear opportunities to introduce front-end automated checks and strengthened internal communication to improve Medicaid MLR data quality.
Audit Objectives
To determine the oversight activities, if any, CMS conducted to verify that States submitted accurate medical loss ratio (MLR) data to CMS.
Audit Findings (6)
CMS’s accuracy checks of the four required MLR data elements were limited because states often did not submit optional MLR data elements, preventing CMS from conducting the checks for more than half of the final MLR summary reports.
CMS did not follow up with states that submitted potentially inaccurate MLR data.
Results of CMS’s accuracy checks were not shared across CMS staff responsible for reviewing capitation rates, limiting use of findings to inform rate development.
For the 2019 MLR reporting year, CMS identified 57 MLR summary reports with potential inaccuracies submitted by 16 states.
Potential inaccuracies identified included 23 MLR numerators, 15 MLR denominators (Check 1), 18 MLR denominators (Check 2), and 1 adjusted MLR.
About 25% of the 595 MLR summary reports could not be matched to the Enrollment Report, so CMS could not perform the Member Months Check; among the matched reports, 25 had potentially inaccurate member months.
Recommendations (3)
Establish procedures that outline the steps to address potentially inaccurate Medicaid MLR data submitted by States and follow up with States, as needed.
Ensure that insights from CMS’s reviews of State-submitted Medicaid MLR data are communicated to CMS staff responsible for reviewing capitation rates.
Improve the efficiency and effectiveness of CMS’s oversight of the accuracy of State-submitted Medicaid MLR data.