Topics discussed:

• The CMS National Provider Directory landed in April as the first CMS provider directory published in FHIR, covering six resource types — Practitioner, Organization, Location, Endpoint, PractitionerRole and OrganizationAffiliation. Endpoint is the one that didn't exist before: associating an endpoint with provider data used to be work you did yourself, and it's what turns a directory entry into an answer to "where do I go to get my data". Gene Vestel loaded the 40 GB bulk drop into BigQuery and built AINPI on top of it, open source and welcoming contributions — CMS has since pulled the public download after Social Security numbers were found in the data and reported, so Gene's copy, with the SSNs redacted, is currently a route to a dataset the source has taken offline while it fixes it.
• AINPI is really an audit framework — a data quality dashboard, state-level drill-downs, and findings pages that publish their methodology so the checks themselves can be argued with, aimed partly at states responding to a federal request to audit their Medicaid provider data. The cross-source search puts the master data management problem on one screen: a single provider turns up in the national directory, the legacy self-reported system and two payer directories, is absent from two others, and the locations and qualifications don't agree between them. Gene's framing is that this is a distributed data problem — everyone holds a piece, and nobody has a good answer for how an update to one propagates to the rest.
• John Grimes asked the question that punctures the optimism: fine, you found the endpoint — how does it know you are who you say you are? Mostly it doesn't, beyond a portal username and password you set up once and then reset ten times because you never use it; newer services lean on IAL2 identity verification providers, but most EHRs haven't implemented that, and Arjun Sanyal expects the standardised identity-provider question to be settled by regulation later in the year. Gene's wider map: TEFCA and the QHINs are the national infrastructure but not everything is on them yet, so which route gets you your data still depends on where you happen to see a doctor.
• Arjun played Josh Mandel's recorded demo of SMART Health Check-in, a protocol letting a clinic's website request structured data — demographics, insurance card, US Core clinical resources, and a visit-specific questionnaire — from a wallet app on the patient's phone in one flow, entered either from a pre-visit link or a QR code at a kiosk. The clinic asks for what it wants and the patient can fully fulfil, partially fulfil or decline each part independently. It works on iOS now by wrapping the request in ISO mDoc — machinery Josh described as either important enabling infrastructure or wire protocol baggage depending on your perspective — which simply wasn't possible when he first prototyped this a year ago; the remaining gap is that the W3C Digital Credentials API only lets natively installed apps respond, so web-based wallets need the experimental cross-tab path he sketched.
• Josh's reason for caring is that health information networks don't reach everywhere and may hit roadblocks nobody has found yet, so a protocol with the patient in the loop is the escape hatch when they don't work — and it's the only route for data that never lands in an EHR at all, like activity data from a phone. Matt Pfeffer picked that up as the patient becoming a source rather than a subject, patient-reported outcomes being a long-standing priority that has stayed stubbornly hard to untangle. Alex, building an open source personal health wallet, described an offline-first take on the same idea: data held locally on phone and desktop, shared through proximity- and time-limited sessions, any input transformed into FHIR, and structured by clinical specialty rather than by resource type so it's useful at the point of care.