In May 2023, staff from the Senate Finance Committee called 120 mental health providers listed as in-network across twelve Medicare Advantage plans. More than 80% were described as “ghosts” which means they were unreachable, not accepting new patients, or not in-network. A third of the phone numbers were incorrect or did not result in returned calls. These results were presented in a public hearing alongside other reports by the American Psychiatric Association, multiple Health Affairs pieces, and CMS’s secret shopper audits from 2016-2019. These studies used different methodologies, but pointed in the same direction: payer directories were significantly and persistently inaccurate. Testimonies by patients, providers, and policy makers recounted cases where directory errors resulted in access-to-care issues and unnecessary administrative burden for everybody involved.
The hearing helped build the legislative case for the REAL Health Providers Act, enacted three months ago. For the first time, MA plans will be required to measure the accuracy of their provider directories and publish those scores. CMS is holding its first stakeholder meeting in June, slightly ahead of the statutory deadline.
While it took years of studies, industry engagement, and legislative work to establish the law, it leaves many important implementation details to CMS. The decisions CMS makes about how directory accuracy is defined, and the respective roles of payers and providers, will establish the model for directory improvement for years to come.
The industry does not have a shared definition for provider directory accuracy
How accuracy gets measured has never been clearly defined. CMS started ‘secret shopper’ telephone audits in 2016 and, after a brief hiatus during the pandemic, has continued issuing audit reports to payers. Multiple studies cited at the Senate Finance Committee hearing relied on telephonic audits but went further than simple directory verification, also testing whether a patient could book a timely appointment. NCQA gives payers more latitude, with some using provider attestation and others using telephonic outreach to satisfy the same accreditation standards.
The problem with a lack of a common methodology is that payers, vendors, and regulators can make divergent claims with no way to adjudicate between them. Some payers have publicly claimed to have 99% accurate provider directory records. Upon further investigation, results like that tend to reflect favorable sampling, a clever denominator definition, or field-level rather than record-level accuracy rates — meaning a single data element like a phone number was confirmed correct, rather than the entire provider record being verified as accurate. Providers will assert that roster data submitted to payers is accurate, but payers and regulators do not agree with the providers upon auditing that data (especially around over-association of providers to addresses).
Without a common definition, every claim of accuracy or error is unfalsifiable. You cannot measure what you cannot define, and you cannot improve what you cannot measure. Because the REAL Health Providers Act requires that MA plans assess and publish their directory accuracy, CMS will need to formally define accuracy in a clear, reproducible manner. This will provide a common language that can be used by regulators, payers, providers, and vendors.
Why phone audits aren’t the answer at scale
The implicit standard for measuring provider directory accuracy has been the telephonic secret shopper audit. CMS pioneered this approach — calling providers listed in MA directories to verify participation status, addresses, phone numbers, and ‘accepting new patients’. CMS even required MA payers to conduct quarterly outreach to providers on directory data. For its time, it was reasonable. In 2016, most appointments were booked by phone, structured provider data was sparse, and the tools to do something more sophisticated at scale didn’t exist.
The world has changed since 2016. Online booking has overtaken phone scheduling for the general population, with some surveys putting phone bookings as low as 22% of appointments. When phone calls are made, more of them are being answered in centralized call centers where staff may be hundreds of miles away and have never met the providers they’re booking appointments for. Over the past 2 years, some practices are relying on AI voice agents to manage inbound phone calls. When AI voice agents, online booking, and centralized call centers are handling appointment-related questions, you know that they are more likely tapping into a system of record, versus relying on institutional knowledge or relationships with providers.
Phone calls also result in an administrative burden problem. Providers are already drowning in attestation requests, credentialing portals, roster spreadsheets, and payer outreach. Every verification phone call is a tax on staff time. We should not solve the directory data quality problem by adding to that burden. The mid-2010s made this concrete: when CMS ran secret shopper audits at scale, some payers responded by standing up call centers to preemptively verify their own directories, which flooded provider phone lines and blocked patients from making their own appointments. We should avoid seeing that unintended consequence again.
Phone audits still have a role. Statistically significant sampling, targeted at specialties with known high error rates, remains valuable as a targeted validation tool. But it should not be the foundation of a modern measurement framework in 2028 and should never be deployed across a payer’s entire provider network.
Providers and payers both have a hand in the problem
Even when payers and providers agree that a directory error exists, they frequently disagree about who is responsible and who should fix the error. This plays out in provider relations calls and delegation contract disputes on directory errors between payers and provider groups every day.
A payer conducts an audit and discovers that Dr. Smith is not seeing patients at a listed address. The provider organization pushes back: the physician attested to that location, it is in the payer’s contract, and under the delegation agreement the payer is obligated to publish data as submitted. The payer responds that their audit and a check of the provider’s own public directory shows no patients seen at that location by Dr. Smith. The provider counters that Dr. Smith could see patients there, and they are not willing to risk a rejected claim if he does. Meanwhile, a patient calls to make an appointment with Dr. Smith at that address and is told he is not there — and concludes, reasonably, that the payer’s directory is wrong.
The dynamic runs the other way too. There are times when a provider submits accurate data to a payer, and the payer is unable to update its directory in a timely fashion, fails to do so at all, or does so incorrectly. Rosters are submitted, providers are enrolled, but those providers do not appear in the directory upon enrollment. Providers follow up with payer enrollment teams to confirm updates, only to find the directory unchanged weeks later.
Everyone in these scenarios has a defensible position. The provider attested in good faith, and believes the primary purpose of the data is for claims and payment. The payer ingested data from the provider and updated its directory or, the payer is overwhelmed with conflicting updates and is trying its best to keep up. The patient used the directory to find in-network care. There are reasonable explanations for each process breakdown, but they culminate in the patient spending more time seeking timely, nearby, in-network care.
What is missing is a shared definition of what accurate actually means — one that is not negotiated ad hoc in interactions between payers and providers. These patient-centric definitions need to be clearly established by an outside authority that can motivate both providers and payers to produce and process the right data.
A patient’s definition of accuracy should be simple. Is this provider seeing patients by appointment? Can an appointment be booked? Will they be at that location (virtual or physical) when I show up? The presence of attestations or ‘contractually correct data’ matters less when compared to the patient experience. The question is whether a patient relying on directory information will get the care they are looking for.
When CMS asserts a measurement methodology, it has an opportunity to communicate this clearly to both payers and providers. Directory accuracy is not defined simply by an attestation or by a contract. It is defined by whether a patient can find, book, and see a provider at the place and time the directory states. That definition should be the anchor for everything else: what gets measured, what gets published, and what triggers a correction.
Good data should be obtained correctly from their respective source
Every directory data cleansing method ultimately relies on a good signal emerging from the provider organization. Payer agreement across multiple directories works because enough payers receive accurate data from the provider in the first place. Phone calls work because someone at the provider’s number confirms the information. Claims work because a submitted claim signals that a provider is still practicing within a group and seeing members in a plan. Web data works because when a provider publishes their own directory listing or syndicates it through other channels, they are signaling they want to see patients at that location.
The goal of a better measurement framework is not to build smarter algorithms that compensate for bad provider data. It is to make good provider data flow at scale — so that when a provider organization has accurate information about providers, that information reaches payers and CMS reliably, without requiring the provider to re-enter it in fifteen different places or answer phone calls from fifteen auditors asking the same questions.
Provider directory accuracy has two distinct source truths that must be correctly joined. Providers own the practice reality: where they see patients, when they are available, whether they are accepting new patients, what their telehealth capabilities are, and whether they are bookable. Payers own the participation reality: whether a provider has a current, active contract with a specific plan, what the effective dates are, whether the provider is credentialed and participating, and whether claims are observed. A provider can be bookable and practicing but not participating in a plan. A provider can be contracted with a plan but not actually see patients. The directory is only accurate when both the provider and payer reality are current, correctly joined, and presented to the patient.
Data mastering has to happen in sequence. First, within the provider organization: credentialing data, scheduling data, and directory data are often maintained by separate teams with no systematic reconciliation. A provider who attests to a location that their own scheduling system doesn’t use has an internal data problem. Second, between the provider organization and the payer. If the provider org has clean internal data and submits it consistently, that data should flow into the payer directory without friction. Third, internally within the payer organization and across all of its departments and systems. Fourth, between all parties and CMS: plans and providers are already making data available to CMS (via Health Service Delivery tables, Medicare Plan Finder, PECOS, and NPPES). If that data is clean, CMS becomes an aggregated system of record that all parties can reference.
CMS has authority over Medicare FFS enrollment and participation. The Part B provider guidance the Act requires (directing providers on when and how to update NPPES) is the beginning of using that authority. If CMS can establish clear expectations for provider data submission, and make accurate submission the path of least resistance rather than an additional burden, the downstream data quality problem becomes more tractable. CMS has a menu of monitoring and enforcement options with providers: asserting an accuracy definition, actively monitoring submitted data, publishing the accuracy of that data, and making some threshold of accurate data a condition for Medicare FFS enrollment. CMS has understandably leaned more on the payers historically than providers to clean up data, but a more permanent solution will require an increased expectation of accurate provider-submitted data. If CMS can thread this needle with the provider community, it can begin to aggregate accurate directory data from providers in a central location and make it more widely available to parties across the ecosystem.
On the payer side, internal data reconciliation is equally important. Health service delivery (HSD) tables submitted to CMS, FHIR-based provider directory APIs, member-submitted errors, provider-submitted corrections, and claims data are all signals payers have. If a plan is paying claims for a provider, that provider is still participating. If claims have gone dark but the provider remains listed, that record is suspect. If a particular provider record has received 10 member complaints over the past month, and inbound update requests via the provider relations team, that record should be investigated. The Act sets an outer bound on the lag: plans must verify directories every 90 days and remove departed providers within 5 business days of determining they are no longer participating. Determination depends on someone flagging the change, and claims going dark and member error reports can surface that signal faster. Modern data platforms can harmonize these signals and surface conflicts into a central workflow. That is not a substitute for good source data, but it is a powerful way to identify where remediation is needed.
All three sides of the industry — payers, providers, and regulators — have work to do to get their data internally consistent and consistent across organizational boundaries. This requires a consensus agreement on what constitutes accurate data, a scalable and repeatable way of measuring accuracy, and sufficient enforcement to ensure that the good data is flowing at scale. None of this works unless all parties work to mature their own internal data mastery, agree who is the right source for what data, and ensure that the accurate data survives.
Access outcomes need to be measured alongside directory accuracy
Provider directory accuracy is not the end objective. Accuracy is only useful if it supports a Medicare beneficiary in finding a provider, making an appointment, and getting care without a surprise bill. It needs to be accompanied by outcome measures that test whether accurate directories are translating into patient access.
The REAL Health Providers Act contains the most important accountability mechanism in the entire law: the cost-sharing protection provision. If an enrollee relies on a directory listing that turns out to be wrong and ends up with out-of-network care as a result, the plan must hold the enrollee harmless. The rate at which members invoke that protection is a lagging indicator of directory failure and that is useful to monitor.
Beginning in 2029, CMS will publish plan accuracy scores in machine-readable format. Once those scores are public, they become targets — which is why patient access outcomes need to be reported alongside them. A plan with a high directory accuracy score and poor CAHPS access scores should draw scrutiny. That divergence is itself a signal that processes are being optimized for the measure rather than the outcome. CMS collects data on timely access to care, appointment wait times, and member experience scores through CAHPS studies. Connecting those measures to new directory accuracy scores, and publishing cost-sharing protection invocation rates, creates accountability anchored to patient experience rather than data accuracy alone.
And in markets where workforce supply is genuinely insufficient, better data helps identify these shortages in an empirical and scalable way. Better accuracy surfaces those exceptions honestly and can support plans in their interactions with CMS. Plans already have the ability to file network adequacy exceptions with CMS when supply is demonstrably limited. Better data makes that process more streamlined: plans can document genuine shortage areas with evidence, CMS can evaluate requests against a shared data picture, and oversight bodies can distinguish between network gaps that reflect supply constraints and gaps that reflect network design choices.
How CMS can roll out directory accuracy measurements
CMS should consider rolling out directory accuracy measurements in a sequence such that payers and providers can adapt, streamlined approaches can be introduced, and progress can be assessed along the way.
Now through 2027: Continue phone audits of payer directories, but limit them to statistically significant samples targeted at high-error specialties. Simultaneously, CMS should begin developing a reference accuracy framework — using provider-sourced practice data and payer-sourced participation data as the two inputs into a National Directory — and testing the relative accuracy of both sources. The goal is empirical: what payer and provider sources are producing the best data, and where are there errors that need to be corrected at the source? Are providers able to produce correct data on practice and availability? Are payers able to ingest that data and join it with participation data to create accurate directories? If CMS uses these tests to determine what data is ingested into a centralized resource and judiciously applies enforcement on both payers and providers, that increases overall industry trust in the data. In parallel, CMS should issue the Part B provider guidance the Act requires: establishing clear, simple expectations for when and how providers update their information, and making NPPES or the National Provider Directory the system of record. CMS can then, for the first time, provide constructive feedback directly to providers about their data and give them a single place to make corrections for the entire industry. As CMS provides constructive feedback, many providers will decide to source their directory data from systems purpose-built for patient access (i.e., their own directory and appointment systems).
2028: Assuming that CMS is able to work with providers to submit more accurate data (from ideal internal sources) into the National Directory, provider-sourced accuracy becomes the reference point for practice data, and payer alignment with that reference becomes the primary measurement. Providers attest to practice reality. Payers attest to plan-participation and contractual reality. The payers’ directories are measured against the correct intersection of both, from the patient’s point of view: are providers bookable, participating, at this location, in this plan? Phone audits continue as targeted validation for high-divergence records and for regularly scheduled calibration. Accuracy scores are published alongside cost-sharing protection invocation rates, CAHPS access scores, and appointment wait time data. The public picture includes both leading and lagging indicators, so that industry and the public can trust that high accuracy scores are aligned with better patient experience.
Beyond 2028: The longer horizon is provider-driven interoperability: more detailed scheduling data, roster data, and practice information that providers maintain once and that flows to payers and CMS via standards-based interfaces without manual intervention or phone calls. Directory and availability data are not only accurate and timely, but help patients make more sophisticated decisions on care selection. The measurement framework being built now should be designed with this endpoint in mind, so that the infrastructure investments payers and providers make between now and 2028 are building towards that end goal.
The through-line across phases is the same. The administrative burden has to go down. The good data has to flow from the provider efficiently. CMS needs to ensure the data conforms with its accuracy definition. Payers need to use the data. And the accuracy measure has to stay connected to what patients actually experience.
A call for a meaningful methodology
Many of the policy, technology, payer, and provider professionals who are attending the June meeting have been working on this problem for a decade or longer. This is a persistent problem and its persistence is not for lack of effort. It is for lack of a clear, shared, patient-centered definition of what an accurate directory means, a repeatable way to measure that, and a constructive feedback loop to all responsible for where they need to improve.
If successful, CMS could establish a durable, consensus understanding of what accurate directory data is that leads to a better patient experience, more timely access to care, and fewer surprise bills. If CMS does not roll out the policies in a thoughtful manner, it could make the problem worse with checkbox compliance, attestation overload, or millions of phone calls quarterly to practices across the country.
We have been working on this too long to settle for another checkbox.
CMS’s public meeting on the REAL Health Providers Act is open to the public. We encourage providers, payers, patient advocates, and technologists to participate and submit input. The methodology conversation starts now.