The Reliability Crisis of Provider Directories in Healthcare Data: An Audit of CMS Public Data
The Provider Directory Lie - by FHIR Data guy
Summary
The author audited the US CMS National Provider Directory (NDH) and revealed significant flaws in its data quality and structure. Key issues include inconsistencies in organizational records, endpoint references, and contradictions between code systems, highlighting serious problems with foundational industry data.
Details
This article presents a detailed audit of the CMS-published National Provider Directory (NDH), which contains over 27.2 million records. The author argues that the reliability of this dataset poses a systemic risk to the entire healthcare information ecosystem and details multiple specific flaws. A primary issue is that Organization NPIs do not map to single real-world entities; 70.5% of unique NPIs map to more than one Organization resource. Furthermore, there is structural disconnection: most Endpoint resources lack a reference back to the managing organization, and conversely, most Organizations have zero Endpoint references. Technically, 18.4% of declared FHIR-REST hosts lack valid SMART discovery documents, suggesting poor network adequacy. Crucially, the `meta.lastUpdated` field is found to be merely an export timestamp, not a per-resource freshness signal, making it impossible to measure actual update frequency. A more severe issue is the existence of 'ghost records' and 'zombie records' due to synchronization delays between the identifier minting agency and the directory publisher. Additionally, taxonomy codes suffer from conflicting systems (NUCC vs. CMS Medicare crosswalk), leading to discrepancies up to 14% of the time. These defects are not merely technical curiosities; they undermine every process that relies on provider data—including claims adjudication, credentialing, and network searches.
Original content copyright by respective publishers