A CME program director is building a two-hour session on a new drug class and needs a headline speaker.
She does what every program director does. She pulls names from the last two years of publications in the relevant journals, checks who spoke well at the last national meeting, and asks a couple of trusted colleagues who they would recommend. Within a day she has a shortlist of four names, all legitimately expert, all with strong publication records in exactly this drug class.
She does not, in that process, open the federal Open Payments database and check whether any of the four have received meaningful industry money tied to the manufacturer of the drug she is teaching about. Nobody told her to. There is no field in her speaker-selection workflow that asks for it, no step where the two searches, expertise and financial disclosure, are supposed to meet.
She is not being careless. She is using the only tools that exist, and those tools were built by two different institutions for two entirely different purposes, and nobody has ever connected them.
The system that finds you the expert and the system that discloses what the expert was paid have never been joined into a single query, so the default experience of "find an expert" surfaces prominence with no financial context at all.
The conflict rate is not a fringe finding
Start with how common this actually is, because the scale is larger than the phrase "conflict of interest" usually implies to someone who has not looked at the underlying data.
A systematic review published in Mayo Clinic Proceedings: Innovations, Quality and Outcomes in 2021 pooled 37 studies covering 14,764 clinical-practice-guideline authors in total. It found 45 percent had at least one financial conflict of interest, with individual study rates ranging from 6 percent to 100 percent depending on specialty and methodology.
The more striking number sits inside a subset of that review. Among the 10 studies that specifically cross-checked authors' self-disclosures against independent payment records, 32 percent had industry payments they had not disclosed in the guideline itself. Not payments that were disclosed and judged immaterial. Payments that were simply absent from the record.
Where per-author payment amounts were reported, across eight studies, the average ranged from $578 to $242,300. That is not a rounding error in a disclosure form. That is, at the top end, a meaningful financial relationship that a guideline reader, a journalist, or a fellow physician weighing a strongly stated opinion has no easy way to see.
And the underlying dataset these studies are checking against is enormous. CMS's Open Payments program, the public database created to track exactly this kind of financial relationship, published 17.07 million records totaling $14.67 billion in payments and transfers of value in program year 2025 alone. That is not a small, obscure dataset. It is a massive, technically public one that remains almost entirely disconnected from every tool clinicians, journalists, and committees actually use to find and evaluate expertise.
Two datasets, two purposes, no bridge
Here is the structural reason the gap exists, and it is worth being precise about it, because the fix is not obvious once you see it.
CMS Open Payments is a compliance and transparency database, built to satisfy the 2010 Physician Payments Sunshine Act's statutory requirement that industry payments to physicians be publicly reported. It is searchable by an individual physician's name. It has no concept of "who are the leading experts in aortic stenosis," because that was never its job.
Expertise-discovery tools, meanwhile, know nothing about payments. A PubMed author search, a specialty society leadership roster, a conference speaker slate, or a professional network like Doximity surfaces prominence: who publishes, who speaks, who is visible. None of these tools carry a financial-relationship field, because building one was never their job either.
ProPublica's Dollars for Docs tool sits closest to a bridge, and it is genuinely useful. It lets a member of the public search a single physician's payments by name. But that is the limit of what it does. It has no expertise layer. It cannot answer "who are the top experts in this drug class, ranked, with their payment history visible alongside them." You need a name first, and a reason to look, before Dollars for Docs tells you anything.
So the actual workflow looks like this: find the expert using one system, and separately, if you happen to think of it, check the payment using a second system, by typing the same name in twice. Almost nobody does the second step, because nothing in the first step prompts it.
This is a graph-join problem, not a disclosure problem
The instinctive response to a 32 percent undisclosure rate is to demand better self-disclosure forms. That instinct is aimed at the wrong layer of the problem.
Self-disclosure asks an expert to report their own conflicts, honestly and completely, at the moment they are selected. It has no independent verification step built into the workflow where selection actually happens. A guideline committee chair or CME program director who wants to check has to leave their selection tool entirely, go to a separate government database, and manually search a name they already trust enough to be considering.
What is actually missing is a graph join: matching a verified expertise graph, who genuinely knows this subject, based on publication record, case volume, and peer standing, against a verified financial-relationships graph, what payments this same person has received, by name or National Provider Identifier, and doing it automatically, at the moment of selection, rather than as an optional afterthought that requires already having a reason for suspicion.
No institution currently maintains both graphs in a form that can be joined at query time. That is the actual gap. It is not that the data does not exist. Both datasets are public. It is that nobody has built the query that makes them answer the same question together.
Why the obvious candidates have not built it
Run through who could plausibly build this join, and each has a specific, structural reason it has not.
CMS has no incentive to build discovery UX. Open Payments exists to satisfy a statutory transparency mandate. Its job is publication, not product design, and there is no mandate pushing it toward building an expertise-search layer on top of its own compliance data.
Guideline organizations still rely on self-disclosure at the point of panel selection. Even reporting frameworks that have matured recently, like the 2024 ACCORD consensus-reporting standard, formalize how conflicts should be reported after the fact rather than building independent verification into the selection process itself. The standard tells you what to disclose. It does not check whether you did.
ProPublica has no ambition to build panel-selection software. Dollars for Docs is a public-interest journalism tool, built and maintained for exactly that purpose. Turning it into a workflow product for guideline committees and CME accreditors is a different mission entirely, and not one a newsroom-adjacent nonprofit is positioned to take on.
Doximity has a direct structural disincentive. A meaningful share of Doximity's revenue comes from pharmaceutical-facing physician attention and engagement. Prominently surfacing which of its most-followed, most-visible physicians are also the most heavily industry-paid runs directly against the interests of the advertisers funding the platform. This is not a matter of Doximity failing to notice the opportunity. It is a business model that actively discourages building it.
The join is technically simple, both datasets are public and already exist. It has gone unbuilt because the party with the expertise data has no reason to surface the payment data, and the party with the payment data has no reason to build the expertise layer.
Why now, and why AI raises the stakes
Two things have changed that make this gap more costly than it was a decade ago.
Open Payments keeps growing every year, both in scale and in the intensity of public scrutiny following well-documented guideline controversies. At the same time, the technical cost of joining two public datasets has collapsed. What used to require a manual, name-by-name lookup is now computationally trivial. The barrier to building this was never really the data. It was the friction of doing the lookup by hand, and that friction is disappearing regardless of whether anyone builds a dedicated product for it.
There is also a newer wrinkle worth naming directly. AI clinical-answer tools synthesize literature and expert consensus at scale, and it is an open, testable question whether those tools account for or disclose the conflict-of-interest status of the studies and experts underlying their synthesized answers. If they do not, AI adoption risks laundering the same undisclosed-conflict problem into a new, faster, more authoritative-sounding channel.
What would actually work
Automatic surfacing, not manual lookup. An expertise-search result should display, alongside each verified expert, their most recent Open Payments summary automatically, matched by NPI, without requiring the searcher to think to check a second database.
Context, not a binary verdict. The tool must present payment type, amount, and recency rather than reducing a complex financial relationship to a misleading "conflicted or clean" label. Research funding is not equivalent to a marketing speaker fee, and treating them identically would make the tool less trustworthy, not more.
Filterable thresholds for panel and speaker selection. A guideline chair or CME director should be able to set a payment threshold as a selection filter, the same way they already filter for specialty or publication count, rather than checking conflicts one name at a time after a shortlist is built.
Refreshed on the data's actual publication cycle. Open Payments updates annually; a serious tool has to track that cycle rather than presenting a stale snapshot as current.
Built on top of a real expertise graph, not a name index. The join is only as good as the expertise side of it. A verified graph of who is actually expert in a subject, drawn from publication record, case experience, and peer trust rather than mere visibility, is the harder half of the two datasets to build, and the one that makes the payment data meaningful once attached.
No implication of guilt from presence alone. The guardrail has to be explicit and structural: appearing in Open Payments is not evidence of wrongdoing. The tool's job is to make the financial relationship visible at the point of selection, not to render a verdict on it.
A route back into formal reporting standards. A tool like this should complement frameworks like ACCORD, giving guideline organizations an independent verification layer that checks self-disclosure against public payment records rather than trusting the form alone.
What you can do now
If you select experts, panelists, or speakers
Add the second lookup to your existing workflow today, even manually. Before finalizing a guideline panel or CME speaker slate, search each name in CMS's Open Payments database directly. It takes minutes per name and it is not currently a standard step almost anywhere.
Ask your own committee what "relevant" conflict means before you need the answer under pressure. ACCME and ICMJE both draw a distinction between relevant and irrelevant financial relationships. Decide your committee's threshold in advance, not while defending a specific selection after the fact.
Do not treat self-disclosure as sufficient on its own. The 32 percent undisclosure rate among studies that actually checked is the single most important number in this article for anyone who selects experts for a living. Self-disclosure alone has already been shown, repeatedly, to miss a meaningful share of real financial relationships.
If you lead a guideline organization or CME accreditor
Build the cross-check into your process, not just your reporting form. Following a disclosure standard like ACCORD is necessary but not sufficient if nobody is verifying what panelists report against the public record.
Publish your own transparency rate. State, for each guideline cycle, what share of panelists were cross-checked against Open Payments and what share of self-disclosures matched. That single practice would put your organization ahead of most of the field immediately.
Ask your AI-tool vendors whether COI status is surfaced in synthesized answers. If a clinical-answer engine your members rely on does not flag the conflict status of the experts or studies behind its synthesis, that is a gap worth raising with the vendor directly.
If you build clinical or research infrastructure
Treat the expertise graph as the harder, more valuable half of the build. Open Payments already exists and is comprehensive; a genuine expertise graph, verified by publication, case volume, and peer trust rather than visibility alone, is the scarce asset that makes the join valuable.
Design the guardrail into the interface. Present amount, type, and recency by default, and make a reductive binary label something a user would have to work to construct rather than something the tool hands them automatically.
Frequently asked questions
How many clinical guideline authors have financial conflicts of interest? A systematic review published in Mayo Clinic Proceedings: Innovations, Quality and Outcomes in 2021, pooling 37 studies and 14,764 guideline authors, found 45 percent had at least one financial conflict of interest, with rates ranging from 6 to 100 percent across individual studies depending on specialty and methodology.
What percentage of guideline authors fail to disclose their industry payments? Among the 10 studies in that same 2021 review that specifically cross-checked self-reported disclosures against independent payment records, 32 percent of authors had industry payments that were not disclosed in the guideline itself.
How do I check if a doctor has received payments from a pharmaceutical company? CMS's Open Payments database, created under the 2010 Physician Payments Sunshine Act, is the official public source and can be searched by physician name; it published 17.07 million payment records totaling $14.67 billion for program year 2025. ProPublica's Dollars for Docs tool offers a public-facing name search of the same underlying data.
Is there a tool that combines expertise search with conflict of interest data? Not currently. Expertise-discovery tools such as PubMed author search, society directories, and professional networks carry no financial-disclosure information, while CMS Open Payments and ProPublica's Dollars for Docs require an existing name and offer no expertise-ranking or discovery layer. The two systems have never been joined into a single searchable tool.
Why doesn't an existing platform like Doximity add conflict-of-interest data to its expert search? A meaningful share of Doximity's revenue is built on pharmaceutical-facing physician engagement, creating a direct structural disincentive to prominently surface which of its most-visible, most-followed physicians also receive the largest industry payments; this is a business-model conflict rather than an oversight.
Does having received industry payments mean a physician's expertise is untrustworthy? No, and any serious tool addressing this gap has to say so explicitly. Research funding, consulting on trial design, and marketing speaker fees are financially different relationships that carry different implications; the goal of surfacing payment data is context at the point of selection, not an automatic verdict of wrongdoing based on payment presence alone.
The bottom line
Two datasets already exist, both public, both large, and both maintained for good reasons. One tells you who is expert. The other tells you who was paid. They have never been asked the same question at the same time, and the 32 percent undisclosure rate found among studies that actually checked shows exactly what that costs: a meaningful share of financial relationships that self-disclosure alone simply misses.
Every institution that could plausibly build the join has a specific reason it has not. CMS has no discovery mandate. Guideline organizations trust the form. ProPublica has a different mission. Doximity has a business model that runs the other way. The gap is not a technical mystery. It is structural whitespace that nobody currently profits from closing.
The CME program director building her speaker slate is not doing anything wrong. She is using the tools that exist. Those tools simply were never built to talk to each other, and until they are, "find an expert" and "check a conflict" will keep being two separate errands that almost nobody has the time, or the prompt, to run back to back.
Part of a series on the missing professional infrastructure of healthcare. Previously: Evidence-Free Question Polling: What Do Other Centers Actually Do?
Evidence note: the core conflict-of-interest statistics come from Tabatabavakili S, Khan R, Scaffidi MA, Gimpaya N, Lightfoot D, Grover SC, Mayo Clinic Proceedings: Innovations, Quality and Outcomes, 2021, a systematic review of 37 studies and 14,764 guideline authors. CMS Open Payments program-year 2025 figures ($14.67 billion, 17.07 million records) are as published by CMS.gov and reflect that agency's own reporting; they should be understood as a snapshot from the most recent available release rather than a fixed constant, since the figures update annually. The ACCORD reporting standard (Gattrell et al., PLoS Medicine, 2024) and the companion consensus-methodology review (van Zuuren et al., BMJ Open, 2022) describe reporting-standard gaps in the field generally rather than measuring undisclosed-payment rates directly. Claims about Doximity's and Sermo's business models reflect publicly known revenue structure and positioning rather than an independent financial audit. Nothing in this article implies wrongdoing by any individual based on payment presence alone; industry payments include research funding and legitimate consulting relationships alongside marketing arrangements, and context matters in every case.
