Monday, September 27, 2010

Unconscious in the Emergency Department

As State Health Information Exchanges and Federal efforts (NHIN Connect/NHIN Direct) implement the data sharing technology that will enable all providers in the country to achieve Meaningful Use Stage 1, I'm often asked  "but when will this healthcare information exchange technology be able to retrieve all my records from everywhere when I'm lying unconscious in the Emergency Department and cannot give a history?"

Here are my thoughts about the trajectory we're on and how it will lead us to supporting the "Unconscious in the ED" use case.

Meaningful Use Stage 1 is about capturing data electronically in EHRs.  Getting healthcare data in electronic form is foundational to any data exchanges.   By 2011 we should have medication lists, problem lists, allergies, and summaries available from EHRs.

The data exchanges in Stage 1 are simple pushes of data from point A to point B - from provider to public health, from provider to provider, and from provider to pharmacy.   There is no master patient index, no record locator service, and no centralized database containing everyone's lifetime health record.

The Stage 1 data exchanges are the right first step.   Serving as chair for a health information exchange for 12 years, I can tell you that pushing data from point A to point B builds trust and breaks down political barriers to data sharing.   The policy needed to guide  push technology is straightforward, since the parties involved in the exchange are limited - the sender and the receiver.  Consent can be simple.  For a referral, this might be "do you permit me to send your data to a specialist so that your care can be coordinated?"

But how are we going to get to a model that supports the pulling of data by emergency rooms?

There are several engineering approaches.

One is to create a master patient index for a state or region so that the identity of patients seeking care is known.   Once we have a master patient index, we can build a record locator service to keep track of every site the patient visits.   Then, we can build a consent repository so that the patient can control what is shared.  During an emergency department encounter, summaries can be pulled from those sites listed in the record locator service which the patient has agreed to share.

I know that this approach works, because it is what I implemented at the hospitals of CareGroup in 1997 as the focus of my MIT thesis.

However, it does have scalability problems.   Given that there are 500,000 doctors and 5000 acute care hospitals, the engineering complexity to pull data from multiple, disparate sites is significant and it requires that all sites support real time queries 24x7x365, maintaining the necessary security and robustness to enable such interchange.   That's a tall order.

An alternative, that is simpler from an engineering perspective and achievable with the Stage 1 push technologies looks like this.

What if payers, providers, and private vendors offered something called "the electronic medical home"?  The patient needs to pick some hosting option they trust.  Just as Stage 1 of meaningful use supports push transactions from provider to provider, the same transactions could be pushed to the electronic medical home designated by the patient.    This has the added advantage of providing a means to  support the patient engagement provisions of Stage 1 (deliver summaries of ambulatory encounters within 3 business days, deliver inpatient summaries upon request, deliver lifetime health summaries upon request).   If every producer of data (lab, pharmacy, hospital, eligible professional etc) pushed a copy of the data they generated to the electronic medical home chosen by the patient, then the patient could become steward of their lifetime medical record hosted by the trusted agent of their choice.  Medical home data would be complete and constantly updated by data producers.

Patients could store their electronic medical home designation on a card in their wallet or on a medical alert bracelet.   If patients visited the emergency room of a hospital they had registered at previously, the hospital would have a record of the patient's electronic medical home selection since the hospital would be one of the medical home data sources.

Ok, but what if the patient arrives in the Emergency Department, naked and unconscious so there is no electronic medical home designation on their body?   I can tell you that in all my years of practicing emergency medicine, I have never treated a naked and unconscious patient, but let's think about the scenario.

There could be a national or a federated regional database which keeps a record of the designated electronic medical home - an "electronic medical home locator service" that is easy to implement because it only has to point to one place, not to every location which has records about the patient.

When the patient arrives naked and unconscious in the emergency department, assuming someone knows the name/date of birth of the patient, the electronic medical home locator service is queried and once the selection of medical home is known, that one location could be queried to retrieve all of the records.

Finally, just to play out the complete use case to its most absurd, if no one knows the patient identity, then we could suggest the electronic medical home locator service should contain a hash of a fingerprint, so that  a biometric - a scan of the patient's finger - could be used to identify the electronic medical home designation and then retrieve the  lifetime record summary.


1.  In Meaningful Use Phase 1, we implement push transactions by 2011.   Google Health, Microsoft Healthvault, Hospitals, and Payers should create electronic medical home repositories capable of receiving push transactions so that copies of inpatient, outpatient and lifetime summaries can be sent there.   The transport mechanism used to push data to the medical home could be REST, SOAP, or SMTP, just as the NHIN Direct project has suggested.   Ideally electronic medical home providers will issue each patient a Health URL, making it easy for EHRs and  health information exchanges to route data to the correct location.

2.  When an electronic medical home designation is elected, a copy of the designation and a hash of the patient's fingerprint is sent to a regional or national electronic medical home locator service.

3.  When the patient arrives unconscious, the name and date of birth  of the patient is used to query the electronic medical home locator service, retrieve the electronic medical home designation and then retrieve the lifetime record summary.

4.  If the patient is naked and unconscious, a fingerprint scan could be used to retrieve the electronic medical home designation and their records.

5.  All of this is empowered by Stage 1 of Meaningful Use as currently written.   All that needs to happen is that providers, payers and vendors need to offer a place to push transactions on behalf of the patient.  

Some have suggested that we should abandon the NHIN Connect, NHIN Direct, and Stage 1 push exchanges in favor of an engineering optimal solution of creating one large database of all electronic health records in the cloud.

Sometimes the engineering optimal solution is not the social/policy optimal solution.

The idea of creating a voluntary, opt-in electronic medical home supported by many hosts - payers, providers, vendors etc.  is achievable and appealing.

In the US, we are wary of any top down, government mandate.   We  are suspicious of centralizing anything.  Maintaining patient control of healthcare data and letting the market provide multiple hosting options seems like an achievable architecture that builds upon what we have already implemented and the meaningful use regulations already in place.


Richard Dale said...

As always, an interesting post. The notion of federating this amount of data (whether real-time or push-to-medical-home), something even rich unitary commercial enterprises (think Banks), are unable to do internally, strikes me as Herculean at best, and more likely, Sisyphean.
However, if we head down this path, then the last part of the puzzle should not depend on fingerprints, well known to be distorted by environmental problems, especially likely to occur with a naked unconscious patient. I recommend Iris scans (or Retina which is more robust but under-developed) to accomplish that magic identification stage. I understand that in some areas of the world where names are complicated identifiers (e.g., in some Arabic cultures), hospitals are moving towards Iris scans as the primary ID because confusion around name variants is so high.

Anonymous said...

Technology around MPI's is terrific, but you are still dependent upon humans to make it work. It's typically the lower level staff person (i.e. registrar) or a busy nurse in the ED registering a patient. How do you change that culture to embrace the technology so if you ever woke up unconscious in the ED, they were pulling right patient?

John@Chilmark said...

John, this is a nice follow-up to our discussions last week and concur that this approach is a viable option for consumers, payers and providers, both from a technology & policy standpoint, for a model that provides access to a record in an emergency situation.

Those situations, especially ones where the patient shows up naked and unconscious are very rare. A more likely scenario where this will be useful is for the snowbirds who spend half the year in the northern climes and the cold months in Florida, AZ etc. For them, I can envision that a patient (consumer) record medical home would be very valuable as they transition care from one provider to another every 6 months or so.

But one big challenge remains: Education for both providers and consumers. For providers, it is with regards to trusting the data. For consumers, it is seeing the value in establishing/creating such a record medical home and feeling comfortable with the privacy and security of their data.

There may also be one other wrinkle, state laws on privacy that are far from uniform in this country. Hopefully, this last point can be addressed through federal leadership such as widespread adoption by states of the NHIN DURSA.

Gerald Beuchelt said...

Great idea. We have been talking about something very much like the medical home in the context of hData: in our presentation on patient-centric health exchange, we have been proposing a systems based on Oauth and UMA that can discover and authorize access to distributed health stores.

It is my firm belief that this approach can be generalized into working with many different transport protocols.

IAMPLE said...

The "electronic medical home" concept makes great sense. While everybody likes to use the emergency department example, there is a real need to coordinate care among multiple physicians. The electronic medical home makes that possible, and an indexing capability might not be needed. The electronic medical fome for the record will probably be the patient's primary care provider.

Robert Reed said...

I think that some biometric ID might be good to have even if patient is fully clothed and conscious. Just in my day to day operation I have seen a dozen examples of patients with the same name and DOB (MAYBE a different middle initial). Sure it's rare but still the chance for identification error is there and will be greater as a MPI gets larger. Something unique and not consciousness-required would be useful I think.

A I said...

Wishful thinking: Imagine a perfect world where all of the patient's health information is gathered and made available electronically when the patient is unconscious in the emergency room.

In that perfect world, imagine a situation where the patient may be very close to death and the care provider needs to act quickly. Will having a tonne of the patient's health information help in anyway? Is there time to look at all the recent patient health information?

How about presenting the data visually to the care provider, as a series of color coded graphs or charts? Imagine a chart with surgical history, with red highlighted text markers pointing out complications, yellow denoting neutrals and green denoting no complications. I am a software engineer and this is as much as I can imagine in the ER world.

If I was going to give a title to such a technology, I would call it, "Visual ER Decision Support" or "Visual ER Adviser". Is this too much to ask?

Lygeia Ricciardi said...

Hi John –

Thanks for the thoughtful post.

Your description of a patient “electronic medical home” sounds a lot like what Google & Microsoft are already attempting to provide. They just haven’t been widely used—at least in part—because the surrounding health system isn’t yet up to speed. I hope implementation of Meaningful Use Stage 1 will tip the balance by A.) Getting the data digitized in the first place, and B.) Making it accessible to patients or the services they designate. As you say, it’s a foundational stage.

As John@Chilmark points out, though, much education is needed for both patients and providers if health IT is to succeed. A big part of the challenge is getting the public to grasp the need—and the ER scenario itself is one that particularly resonates for many. It’s my hope that Meaningful Use will start to shift the public mindset as patients get used to being included in the health information loop. Meaningful Use will succeed best if it changes expectations as well as processes.

Lastly, I like AI’s comments about the visual presentation of health data. When we move beyond the current (necessary) fixation on accessing data, we’ll need to shift our focus to how to make sense of it all--not just in an emergency scenario, but so providers and patients alike can make use of it on a daily basis.

Anonymous said...

I love the idea of having an explicit home designated for each patient's information- it is so in line with what we know is necessary for better care delivery in general, and where health reform is expected to take us.

To paraphrase the HIT Policy Committee Privacy and Security Tiger Team, providers are in a unique position as trusted guardians of patient information in a way that health plans, employers, government, and vendors aren't.

Especially at the system or referral network level, I think it will be very easy for most patients to know the delivery network that they want to register as the keeper of their care (while individual PCPs may change).

Two extensions may be worth considering,

1) Health Information Service providers (HISPs) for small unaffiliated practices (information-only IPAs?)

2) HISPs for patients who want to be their own keepers, eg who are anticipating frequent transitions (Personally Controlled Health Records).

Also, regulations/ gov't oversight may be needed to ensure that electronic medical information homes do share patient information when appropriate request is made, in the patient's best interest. (Of course best government action would be to align market forces so it is in their interest to do so)

Alan Viars said...

Hi John:

Nice post. I agree with John Chilmark and Lygeia Ricciardi's comments about the importance of patient education. I agree there is a need for a standardized, nationwide, master patient index (MPI). I advocate for the second simpler "push" approach rather than a record locater approach for the reason's you mentioned.

In any case, we are going to have to contend with many different file types including a mix of structured data (CCR/CCD/OMHE Microsyntax) and semi- structured data PDFs with tags such as ICD-9/CPT codes. So viewing things in "bank statement style" where each data element is just a unique transaction occurring over time makes the most sense. Some of you might know I've been spending a lot of time lately building open source solutions with this goal in mind. I'm fond of REST as an SOA approach.

-Alan Viars, CEO

Alan Viars said...

I'd like to address your statements about biometrics, an area where I have subject matter expertise.

1. Making a predictable hash or password from a fingerprint is next to impossible because the sample changes every time. Its fuzzy so the hash will change. You can, however, unlock a password when a close match is found. This is how the DoD's ID cards work. The match algorithm is on the card and if a match occurs, then card releases its "secret".

2. A single fingerprint is not enough data to look up a person in a population over ~5000 (relying on level-2 minutiae data as almost all fingerprint systems do to date). This is why the FBI/DoD take 10 rolled and flat prints to distinguish large populations. In short, if you want to use just a single fingerprint, you have to provide a token such as an MPI. This is a 1-1 match as opposed to a 1-N biometric search.

3. Not all portions of the population have easily usable fingerprints. Troublesome portions of the population include children, mason workers, Asian women, the elderly, and to some degree women in general.

4.Many people have reservations to biometric technology, like RFID, because it is seen as "big brother" or "the devil". Fingerprints have a criminal connotation because they are acquired when a person is booked.

I'm not saying biometrics isn't a good and workable solution, but it is important to plan for “exception handling” for those who abstain based on philosophical or religious beliefs or simply are not enroll-able because of physical characteristics.

If you ask me, it's not "meaningful" until the patient is the hub for exchange.

Alan Viars, CEO

Anonymous said...

I love the idea of a patient controlled electronic medical home. For the tech savvy patient, the one who uses electronic bill-pay through their bank, and probably comes in with a DDx in mind after reading about their SXs on the Web, this is a great solution. The thing that is still missing though is the support for the many patients who visit EDs who are not tech savvy.

It is probably more likely that an un-domiciled, psychotic or polysubstance abusing patient would present in the extreme scenario of naked and unconscious, and I would hazard to guess that most of these patients have not designated an electronic medical home. But just to step back from this extreme scenario a bit, the majority of less-well-educated and economically disenfranchised patients are also less likely to designate an electronic medical home. We take care of many of these patients in the ED, most of who are uninsured or underinsured. Trying to educate all of them and get them to take an active roll in maintaining an electronic medical home is a monumental if not impossible task.
I am not trying to be overly paternalistic or negative. I definitely think there is a place for this model. I’m just not convinced that it will cover this use case for enough of the > 100 million people who visit the ED every year.

It also may not lend itself as easily to other use cases that require data aggregation:

• De-identified and/or aggregate data for cohort identification and retrospective research

• Along the same lines, pharmacovigilance/stage 4 drug trials

• Ability to rapidly deploy new biosurveillance metrics not already being reported through a push

• Patient locator service in the event of a mass casualty event (i.e. did my loved one show up in any of the > 100 hospitals in the NYC metropolitan area after a terrorist attack)

I think there is still a place for the MPI/RLS model, and having RHIOs becoming the nodes for exchange of standardized data once a patient leaves a particular geographical area. This would still likely require some form of national MPI, or at least a standard by which to query any RHIO’s MPI, but it should require a nationwide RLS and all of the scalability issues mentioned.