Do Hospital Directory Updates Reach Payer Directories?

When a hospital updates its roster - adding a newly credentialed physician, or dropping one who has moved - how quickly and completely does that change reach the payer directories that members actually use to find care?

That's the question I asked myself after last week, when I looked at how well hospital systems keep their own directories fresh. Indeed, system directory freshness is only half the story: a perfectly maintained hospital roster isn’t very helpful if the update never reaches the health plan directory a member is actually searching. So this week I follow the hand-off, from the system's directory to the payer's.

J2 Health maintains roughly 300 payer directories and more than 500 system directories, which gave me a broad, matched view of both sides of the hand-off: from that footprint I assembled a set of payers alongside the in-network hospital systems that feed them, and traced how roster changes move from one to the other.

The setup

I studied the Aetna and Oscar directories in the New York metro, and the Aetna, Ambetter, and Oscar directories in Dallas–Fort Worth. For five in-network hospital systems, I tracked monthly directory changes - providers added or dropped - and measured what share appeared in the corresponding payer directories over the following three quarterly scrapes.

A note on the sample for data nerds: I chose large systems whose directories list many office-based physicians rather than hospital-based providers (HBPs) since anesthesiologists, pathologists, and ER physicians are routinely suppressed from member-facing payer directories and would otherwise shrink and skew the sample. I also excluded a handful of highly specialized, referral-only fields (neurosurgery, transplant, and the like). On the payer side, I chose directories with a strongly documented baseline turnover rate, so I could distinguish genuine system-driven updates from ordinary payer-side maintenance (member complaints, secret-shopper calls, and the like).

Across the study I looked at about 3,300 system-side updates, tracing each provider added or removed on the hospital side into the payer directory scrapes that followed. Because every directory churns a little on its own, I scored each payer against its own background turnover rate, so coincidental changes aren’t miscredited as a system update flowing through. One caveat: the analysis does not yet account for edits withheld through network steerage or availability-based suppression - a dynamic I'll take up in the next piece.

The results

The payer directories showed evidence of updates driven by system changes, above the directory's own background churn, but the pass through of system edits to payer edits was far from one-to-one. In aggregate, about 33% of the expected edits we were tracking flowed through to Aetna's directories, about 24% to Ambetter's, and about 18% to Oscar's. (Again, by "expected edits" I mean the office-based, non-suppressed providers that should appear in a member-facing directory in the first place.)

Flow-through also differed by system. For example, in Aetna’s directory, additions from Mount Sinai and NYU Langone propagated most strongly (+29 and +15 points above baseline, respectively), while Northwell's changes were the most muted (+5). I won't over-read the causes, but the divergence is interesting and maybe instructive. Does it reflect how cleanly and promptly a system shares data with its payer partners, or rather the priority a given payer places on a particular system's feed? There may be more complex, technical causes at play too, such as other data sources overwriting one system’s updates at a higher rate than another system’s, due to greater overlap (e.g. a claims-based source, or a public dataset).

Provider removals registered more reliably above baseline than provider additions across every payer and system directory. From a member's perspective that's a healthy skew: it suggests payers are most actively clearing out "ghost providers," the phantom listings that have long plagued this space. This result did surprise me a bit, though: anecdotally, staleness remains rampant, and payers have historically had more incentive to over-represent their networks than to under-represent them. I'd be especially curious for any reader's insight on this dynamic. My initial read is that, beyond a genuine and welcome focus on culling ghosts, the weaker propagation of adds may partly reflect system-requested suppressions. More on that next week.

Why it matters

Payer directory accuracy is increasingly under the regulatory microscope, with the REAL Health Providers Act the clearest recent example. Getting it right depends on several links in a chain: the upstream systems, whose own data quality varies widely, and the payers, who then have to actually ingest and publish what they receive. The evidence here is that the hand-off is happening - but with meaningful leakage between what systems change and what payers ultimately show. Certainly, a chunk of that leakage isn't the payer's "fault": providers may be intentionally suppressed, or a system may hand payers a different roster than the one it posts on its own website (a dynamic Ron Urwongse has discussed on Nick Helfrich and Mitchell Gorodokin 's podcast). What gives me hope is that the two forces bearing down on this problem are finally aligned - the ability to collect payer and system data at scale, driven by AI, is arriving alongside real regulatory scrutiny. Together these forces should let us start seeing more clearly who is updating what; transparency has a way of driving better practices behind the scenes.

‍