Wednesday, September 2, 2015

Solving the Provider Directory Problem for the Country

In many previous posts, I’ve written about the importance of enabling infrastructure to accelerate interoperability.   The standards are not the rate limiting step, but the lack of a provider directory, patient identifier, and consent registry are.

David McCallie of Cerner has solved the provider directory problem of the country.

He downloaded the NPPES national provider database.

He created a FHIR-based Application Program Interface to the database by writing 300 lines of Python code and put it live on Amazon Web Services (for $15/month)

You can try it yourself here:

Just look up last name Halamka (or any other physician know)

Some important caveats:

It’s based on loading the imperfect and often out-of-date CMS national provider database. database.

The national provider database does not contain Direct addresses, so to run it as a real national service, Health Information Services Providers (HISPs) would need to submit a comma separated value (CSV) file of Direct Addresses and National Provider Identifiers at reasonable intervals.

Each night, the database could be re-loaded  using the then-current CSV files.  That would easily allow a HISP to remove or correct names, or even drop out (submit an empty CSV.)

As I wrote about in my recent post, Trajectory not Position, we all need to be doers.

The Provider Directory for the country issue has been solved, we just need to get HISPs involved in updating it.

I look forward to including the David McCallie Provider Directory FHIR implementation in upcoming national standards recommendations.


Matthew Tobias said...

This is a promising first step! One issue it would not solve though is finding Direct Addresses associated with a location only. One of the first things we discovered when rolling out TOC was that many people refer to a place, not knowing or perhaps not caring what provider the patient will ultimately see. We're getting more requests to publish place DA's than provider DA's now.

I don't want to downplay the excitement here though. A national directory is exactly what we need. Having every organization maintain data from tens of thousands of other organizations is unsustainable.

Adrian Gropper said...

Here's a post about unique patient identifiers:

How much privacy do people who are also licensed professionals deserve? Should we be encouraging Google to collect as much information about each licensed professional, including the NPPES and public legal outcomes and put together a bigger directory than David's?

Bristow said...

Some comments from the perspective of a specialty clinic that has over two thirds of its patients from Medicaid and Medicare with institutions like the VA, military, state mental hospital etc. dominating the rest.

1) You definitely need to be able to search by clinic name or facility as some institutions (military bases for example) experience enormous provider churn. Being able to see which providers are associated with a facility important. You also need the directory to support showing the address for organizational npi numbers so you can map facility-wide addresses as well.

2) The ONC needs to force EHR vendors to participate and contribute their directories on a nightly basis if they want to be certified. I just had the lovely experience today of asking a vendor of several providers we consult for if (a) we could get a copy of their provider directory as an excel doc to submit our vendor for inclusion in our directory (they said no), (b) our providers could be listed in their directory for the convenience of the multiple customers we work with (they said no again, their directory is for EMA providers only - and there's no address book feature or way to save and retrieve direct addresses for providers who use another EHR), (c) we could get a copy of the direct addresses and npi numbers for the specific clinic on the call (no, we don't have a way to export that). The practice administrator was ready to throw their vendor into traffic.

3) The ONC needs to add to the certification test scripts verification they can transmit direct messages to other certified EHRs (randomly-assigned, no advance notice) to avoid these islands where EHR vendors either charge or make it close to impossible to send a direct message to someone outside their vendor ecosystem.

4) The ONC needs to add to the certification test scripts a step to verify that users can search the above directory.

And one bug report for the linked directory:

It appears that the search is assuming there is only one direct address per NPI number, however I can tell you that it is pretty routine for providers to have at least two (or more). Providers who have remote EHR access are being auto-enrolled in Direct by the large health centers, but they usually have their own practice as well or work for several community-based clinics. Each of those clinics may also have direct addresses assigned if they are doing MU. So that NPI number could have two, three, four, etc direct addresses associated with it that are all equally valid simultaneously. However, the search does need to accept from the EHR vendor directory submission a facility name and address(es) so the sender can verify that they have Dr. John Smith at the right clinic.

Angelique Mattin Russell said...

Fantastic resource, thanks for sharing.

In addition to the aforementioned concern about multiple Direct addresses it is also problematic when organizations have set up "ghost" direct addresses that don't truly go anywhere. A read receipt was not a requirement for Meaningful Use, just a confirmation that it arrived at the destination. A few organizations in my local market used dbMotion to set up inboxes for several local independent physicians. These inboxes were to enable exchange between the hospital and their providers, but if the provider never bothers to log in or integrate with their own practice EMR you are pretty much sending to a black hole. It would be a shame if these black hole addresses make a directory unreliable.

I would also like to offer a counter-perspective: I do not agree that nationwide directories or national patient identifiers are essential for interoperability. I would argue instead that patient-centric information systems hold the key to the future of healthcare data management. If a platform had Facebook-like dominance and ease-of-use it could be a place where connections between providers are managed. I also think hosted solutions need to die a quick death. If the patient-centered platform contained not just provider Direct addresses but facility EHR details including the API information it could broker encrypted access into a single medical record. So when I'm in an ER 3,000 miles away, I use my smart phone to identify myself as a patient at Random Distant Hospital. Random Distant Hospital receives a ping with my patient identifier. Random Distant Hospital has now inherited rights to open distant web-based EHR's that show my medical records in detail, as well as TOC summaries collected from CCDA. Safe, secure, patient-controlled.

We talk about EHR's being designed for hospitals and then poorly adapted to outpatient and vice-versa but the real problem is all EHR's are provider-centric, not patient-centric. The fluffy "patient portals" we have today are not going to cut it--we need to rethink healthcare information systems from the ground up. Many of the gains we fail to realize from technology are due to inflexibility of EHR's trying to be everything to everyone--highly specialized EHR's actually make a lot of practical sense, but are prohibitive because of the interoperability problem. A dominant and universal patient-centered system could solve the interoperability problem and allow greater innovation in HIT.

Brandt Welker said...

Great idea, we do this locally, also with a FHIR RESTful API and we’d love to participate in a national resource; does anyone think clinicians are going to have a problem with this? Do clinicians want to be contacted via Direct?

See excerpt from Applicability Statement for Secure Health Transport "A receiving STA MUST NOT send an MDN unless it is prepared to take on legal responsibility for receipt and delivery of the message." This will hold true for the receiving party as well.

David said...

Very cool! Thank you John and David. I tried it on several of my own providers, and it found them (though in a couple of cases it showed a different address than I know, but it's possible the doctor has multiple offices). I hope that momentum will build around operationalizing a national PD and (especially important) keeping the date up-to-date.