Can software reach the system holding your record?

Where the chain breaks

Each step is a subset of the one above it. The candidate row sits outside the chain on purpose and is never added into coverage.

1. Active practitioners with an NPI

   230,837100%

   Everyone the NDH lists as practising in this state.

2. Has an active PractitionerRole

   100,91843.7%

   Without a role there is no organization, so no path to an endpoint at any confidence. This is the binding constraint. See the finding

   −129,919 lost at this step

3. Role names an organization that resolves

   94,84941.1%

   The reference points at an Organization that exists. See the finding

   −6,069 lost at this step

4. Organization carries an NPI

   94,84941.1%

   An organization NPI is what lets the org be joined to any other federal or vendor source.

5. Role names a location

   91,81139.8%

   The site the patient actually visited. No NDH Location carries an endpoint or a parent, so the place layer stops here.

   −3,038 lost at this step

6. Reaches a FHIR endpoint

   43,06018.7%

   Includes 5,720 reachable only after the vendor-published NPI fill, 31,636 only after name or brand resolution, and 3,243 only via the practitioner's CMS-enrolled group, of which 299 carry no PractitionerRole at all. The NDH alone reaches 2,461. See the finding

   −48,751 lost at this step

7. Name-matched endpoint candidate (not a linkage)outside the chain

   3,1831.4%

   An endpoint is published against an organization name that matches, but nothing deterministic connects them. This is where the large health systems land: the vendor names the brand, the NDH holds the practitioners under separate legal entities, and the NDH carries no hierarchy to bridge them. Triage material, not a directory linkage. See the finding

8. Endpoint attributed to an EHR vendor

   41,38117.9%

   The vendor is derived from the endpoint host and cross-checked against the file that published the URL. See the finding

Where the chain breaks, geographically

The same practitioners as the funnel above, placed by postal code. Colour is always a rate and never a count, so the map does not simply redraw where people live. Switch layers to change the encoding; the geometry stays put.

What established each link

Bands describe method, not quality. Green means the NDH itself carries the whole chain. Yellow means the last link came from a vendor-published NPI that matches an NDH organization: deterministic, but not the directory's own statement. Red means the practitioner reaches an organization and stops. None means no organization at all. Candidate is weaker than red on evidence and is never counted as connectivity: an endpoint exists under a matching name, with nothing deterministic tying it to this organization.

Deterministic

2,461

Vendor NPI

5,720

Name/brand resolved

0

CMS-enrolled group

0

Name candidate

3,183

Stops at organization

48,960

No organization

135,634

Which EHR the reachable practitioners sit behind

Vendor is derived from the endpoint host and cross-checked against the file that published the URL. It covers only the 41,381 practitioners who reach an endpoint at all, so it describes the connected minority, not Pennsylvania’s EHR market.

The web, and the holes in it

The 50 largest systems, the endpoints any of their members reach, and the EHR vendor behind each endpoint. Systems with no edge are the finding: they hold practitioners and nothing public connects them to a system that serves records.

Health systems, and how we know

Health systems, best evidence first. `cms-ownership` groups come from CMS enrollment ownership data, which states ownership and names holding companies explicitly, but covers hospitals only. `nppes-parent` groups come from NPPES is_organization_subpart plus parent_organization_lbn: real, but sparse, sometimes stale, and sometimes a program rather than an owner. `brand-name` groups are inferred from the organization name and cover the rest. The directory's own OrganizationAffiliation resource was tried and rejected: see affiliation_graph below.

A system reaching an endpoint is not a routing answer. Large systems publish several: UPMC has an Epic endpoint for its physicians and separate athenahealth endpoints for individual surgery centres. Every endpoint found in a system is listed with the member that carries it, and none of them is presented as the one serving a given practitioner.

The work queue: who to fix first

Organizations with 25 or more practitioners that no method reaches: not by the NDH's own attribution, not by a vendor-published NPI, not by an exact name match. This is the work queue. Every row carries an NPPES link so each one can be checked against the primary source.

These 100 organizations hold 110,128 practitioners between them. For several of them a FHIR endpoint is already published by their EHR vendor, under the brand name rather than the legal entity name the directory carries. Nothing public connects the two.

Largest organizations, and what reaches them

Ranked by how many Pennsylvania practitioners name them. A candidate endpoint is shown in its own column so it cannot be read as a resolved one.

Hospitals

Facility-level connectivity comes from the separate hospital audit, which starts from the CMS hospital list rather than from the directory. See the full hospital view.