Table of Contents
Provider directories are essential tools to facilitate access to care, coordinate referrals, evaluate network adequacy and support interoperability. Yet, the data behind them is widely recognized as inadequate and inaccurate.
Government sources like the National Plan and Provider Enumeration System (NPPES) and Care Compare aim to centralize and standardize provider information. However, these systems are slow to update, rely on self-reporting and lack enforcement. According to the Centers for Medicare and Medicaid Services (CMS), provider directories "often contain inaccurate information, rarely support interoperable data exchange or public health reporting, and are costly to the health care industry."
This fragmented, outdated ecosystem that fails to reveal the most essential information about the U.S. healthcare system: Who provides it, and where.
Challenge | NPPES & Care Compare | Our Approach |
Unreliable updates and data sources | Self-Reported Updates NPPES updates are self-reported and infrequent, typically limited annual schedules or manual batch processes, if they occur at all. |
Monthly Updates We leverage more than 100 billion rows of claims with monthly refreshes. |
Inactive Providers Listed | Self-Reported Activity NPPES encourages providers to deactivate their NPI when they retire, but it is not mandated. |
Active Provider Flag We flag providers as inactive when their clinical activity stops. |
Misclassified Specialties | Static Taxonomies NPPES is often outdated and represents taxonomies at varied levels of granularity. |
Practicing Specialties We determine specialty from recent claims, improving accuracy, especially for subspecialists. |
Incorrect Provider-Organization Relationships | Self-Reported Provider-Organized Relationships Care Compare lists outdated affiliations, making it difficult to determine which relationships are relevant. |
Behavior-Based Provider-Organization Relationships We derive relationships from observed claims behavior to offer a dynamic view of how providers are connected to organizations. |
Missing Organization Hierarchies | No Organization Relationships NPPES does not link organization NPIs, and a single hospital can have dozens – making it hard to understand which facilities belong to which system. |
Parent-Child Organization Hierarchies We build parent-child relationships to group related NPIs, reflecting how health systems are actually organized. |
Incorrect and Incomplete Practice Locations | Single Address Attribution NPPES only assigns a provider to a single administrative or billing location. |
Multiple Practice Locations We assign up to five practice locations based on recent care delivery locations. |
Trilliant Health’s dynamic, claims-driven provider directory analyzes more than 100 billion rows of claims monthly to determine how, where and for whom each provider practices. Our unique approach has five guiding principles:
We distinguish between providers who are listed in a directory and those who are currently delivering care – an essential distinction for operational, clinical and strategic decision-making.
Instead of assuming a provider is active because they appear in a registry, we look at the past 12 months of claims. If none are present, the provider is flagged as inactive. This binary classification – active or inactive – is updated monthly to reflect the current clinical workforce.
By grounding activity status in claims recency, we remove the guesswork and reduce the risk of acting on outdated information.
Here is a blinded market example that highlights how a claims-based view of provider activity can reveal meaningful differences from registry-based directories:
Taxonomy | Active Providers per NPPES |
Active Providers per Trilliant Health |
Internal Medicine | 2,728 | 1,361 |
Cardiovascular Disease | 459 | 339 |
Vascular Surgery | 112 | 91 |
Orthopaedic Surgery | 556 | 378 |
Public directories often misclassify specialists as generalists. This error is most common in general classifications like Internal Medicine, Family Medicine or Student in an Organized Health Care Education/Training Program. Why? Many providers register for their NPI during medical school – and never update their taxonomy codes after entering residency or fellowship.
To classify specialties more accurately, we apply a combination of heuristics and machine learning techniques to evaluate patterns in claims data. This approach corrects outdated or generic taxonomy codes with the provider’s observed practicing specialty.
Here is a sample of providers whose NPPES taxonomies do not reflect their actual practicing specialties, as determined by claims-based classification:
Provider Name and NPI | NPPES Taxonomy | Practicing Specialty |
Stephen H Gamboa, MD MPH; 1003006149 | Family Medicine | Emergency Medicine |
Jyotsna Mareedu, MD; 1003006586 | Internal Medicine | Hospitalist |
Saroj Neupane MD; 1801184296 | Internal Medicine | Cardiovascular Disease |
Christina P Hitchcock MD; 1265524847 | Family Medicine | Obstetrics & Gynecology |
Instead of relying on rosters, we examine billing patterns to capture a provider’s relationships with organizations.
Physician employment is inherently complex. Some providers are directly employed by hospitals or corporate entities, while others work through independent contractor groups. These relationships shift frequently and are not publicly disclosed. Critically, no comprehensive and reliable dataset exists to track physician employment.
NPPES lists practice addresses, but not employers. We solve this data gap by analyzing billing patterns to identify provider-organization relationships in near-real time. This model aims to associate providers with a medical group, rather than the facility where care is rendered. We analyze a provider’s claims activity to generate the most accurate and up-to-date view of their organizational relationships. This gives a more accurate view of network alignment and competitive dynamics.
We connect disjointed billing entities to show how organizations relate within broader networks.
Public datasets like NPPES list each organization under its own NPI with no clear linkage to a parent system or brand. A single hospital may appear under dozens of separate NPIs, obscuring the true scope of a health system’s footprint or provider alignment.
We group related NPIs under a unified system structure by normalizing entity names and applying probabilistic models to build parent-child relationships. This approach enables users to analyze networks more accurately, supporting competitive analysis, physician outreach and care coordination.
Here is a partial view of a health system, illustrating how parent–child relationships clarify connections and uncover the true structure of the organization:
Parent Organization | Organizations |
Vanderbilt Health | Vanderbilt University Medical Center |
Vanderbilt Medical Group | |
Vanderbilt Wilson County Hospital | |
Vanderbilt Imaging Services Hillsboro | |
Vanderbilt Pediatric Associates |
Most directories assign providers to a single administrative or billing address, which often creates confusion about the provider’s location and misrepresents geographic access to care.
We use claims-derived sites of service to identify all practice locations and attribute volumes by day of the week. Locations are refreshed monthly with near real-time utilization data, capturing provider movement and location-specific activity over time.
By dynamically tracking how providers distribute their practice across facilities, our approach supports more efficient provider outreach and network planning. For a deeper look at how this data informs market-level planning and specialty coverage analysis, explore our methodology for provider needs assessments.
Trilliant Health’s provider directory has the most reliable data for initiatives ranging from strategic growth to day-to-day provider engagement.
Need help applying these insights to improve your health system’s provider needs assessments? Schedule a consultation with our analytics team.