CMS Releases Medicaid Spending Data: What Providers Need to Know
- The Centers for Medicare and Medicaid Services (CMS) is intensifying its efforts to combat fraud, waste, and abuse within vital healthcare programs like Medicaid.
- On February 14, 2026, CMS released a new dataset containing provider-level spending data, intended to help identify unusual billing patterns.
- The data encompass both fee-for-service claims paid directly by Medicaid and those processed by Medicaid managed care organizations on behalf of their enrollees, covering the period from 2018...
The Centers for Medicare and Medicaid Services (CMS) is intensifying its efforts to combat fraud, waste, and abuse within vital healthcare programs like Medicaid. These initiatives span a wide range of coverage types and provider services. In , CMS issued a letter to states outlining opportunities for collaborative action between federal and state governments. This isn’t a new focus; the Center for Program Integrity (CPI), established within CMS in , signaled a shift from a reactive “pay and chase” approach to a more proactive model leveraging data analytics for fraud detection and prevention. CPI has supported states through the Medicaid Integrity Institute, providing training and access to comprehensive data sets to bolster program integrity efforts.
On , CMS released a new dataset containing provider-level spending data, intended to help identify unusual billing patterns. While this data release represents a valuable step toward transparency and accountability, it’s crucial to understand both its scope and limitations. This article will detail what the dataset includes, what it excludes, and how potential misinterpretations could arise without careful consideration of these factors.
What do the data include and exclude?
The newly released dataset comprises seven key data points:
- The National Provider Identifier (NPI) for the billing provider.
- The NPI of the servicing provider (which can be an individual or an organization).
- The procedure code (HCPCS code).
- The month and year of service.
- The number of beneficiaries seen.
- The number of procedures delivered (claim count).
- The total amount paid for the services.
The data encompass both fee-for-service claims paid directly by Medicaid and those processed by Medicaid managed care organizations on behalf of their enrollees, covering the period from to .
However, significant portions of Medicaid spending are *not* included. Notably, the dataset excludes all institutional records – representing care delivered in hospitals and other facilities – and all prescription drug costs. These omissions are substantial, as hospital care alone accounts for 37% of total Medicaid spending, making it the single largest expenditure category. Beyond these broad exclusions, several other critical pieces of information are missing, hindering a comprehensive evaluation of service volume and spending:
- Enrollment: The number of individuals eligible for Medicaid services fluctuates based on state policies, economic conditions, and demographic shifts. Comparing service utilization or spending across time or locations without accounting for enrollment numbers and beneficiary health status can be misleading.
- Benefits and Coverage: The specific services offered by Medicaid and the eligibility criteria for those services vary by state and can change over time. These variations impact service utilization patterns.
- Payment Rates: Spending levels are directly influenced by the rates states pay for each service, which are affected by local cost of living and state-level decisions regarding provider reimbursement.
- Diagnoses: The dataset lacks information about the underlying medical conditions for which procedures are performed, making it difficult to assess the appropriateness of care.
- Place of Service and Modifiers: Information about where services were delivered (e.g., in-person, telehealth) and other modifiers that provide context about the services is also absent.
How might the data lead to mistaken conclusions?
Data analytics are a powerful tool for identifying potentially problematic patterns, but relying on data in isolation can lead to inaccurate conclusions. Several shortcomings of the new Medicaid data warrant careful consideration.
- Procedure Comparability: Not all procedures are created equal. Some HCPCS codes are narrowly defined, while others encompass a broad range of services. For example, CMS highlighted personal care as a significant spending category. However, the “procedure” code for personal care can represent services ranging from 15 minutes to a full day. In contrast, psychotherapy codes are differentiated by visit length (30, 45, 60 minutes, etc.). Similarly, emergency department and office visit codes vary based on the complexity of the case. Including institutional spending would likely shift the relative proportions of spending across these categories.
- Provider Comparability: The dataset includes a diverse range of providers, from individual practitioners to large group practices, clinics, and even state and local health departments. CMS’s own example showed that 10 of the 20 largest “providers” were government agencies responsible for both administering and delivering Medicaid benefits. States employ different approaches to benefit delivery, but many health departments directly provide services, particularly for behavioral health and individuals with developmental disabilities.
- Data Quality and Methodology: The dataset’s creation process and the quality of the underlying data are not fully transparent. The data originate from the Transformed Medicaid Statistical Information System (T-MSIS), a valuable resource, but one that can contain data quality issues in specific states and for certain topics. CMS maintains a data quality atlas to identify potential problems. It’s unclear how CMS addressed these issues when generating the public dataset. For instance, CMS reports that in the data, six states had unusable spending information, and another 16 had data of high concern. It remains unknown whether this unusable data was included in the public file or if a different version of T-MSIS was used.
Beyond these data-specific concerns, it’s essential to remember that Medicaid spending and care utilization patterns underwent significant changes between and . The COVID-19 pandemic, beginning in , led to increased Medicaid enrollment due to the continuous enrollment provision and heightened awareness of unmet needs for behavioral health and long-term care. As states expanded access to these services, utilization and associated spending naturally increased, reflecting changes in coverage, eligibility, and provider payment rates.
The release of this provider-level spending data is a positive step toward greater transparency in Medicaid. However, responsible interpretation requires a thorough understanding of the data’s limitations and the broader context of healthcare delivery and policy changes. A nuanced approach, considering these factors, is crucial to avoid drawing inaccurate conclusions and ensure that data-driven efforts to improve program integrity are effective and equitable.