Tip:
Highlight text to annotate it
X
Male Speaker: Okay. Well, welcome back, everybody, and we
have Marc Williams' slides up to introduce the eMERGE presentation on EMR integration
and Genomic Medicine Implementation. Marc, it's yours.
Marc Williams: Thank you, Dan. At this point, batting seventh
in the line-up, I think we're going to experience what Yogi Bear described as déjà vu all
over again. I won't read the title of our group because it would take my 10 minutes.
Needless to say, we were asked to cover a lot of different subjects so what you're going
to see is a very highly selected group of activities that have been ongoing.
So the EMR Integration Group, which I co-chair with Justin Starren, has been primarily focused,
as you've heard, on pharmacogenomic representation and clinical decision support, and this is
a list of the different variants that are being returned across the sites. We've seen
this before, so I'm not going to go through it. But we also, as you've heard, do have
some non-PGx initiatives that, again, I'll just project here briefly, and you can take
a look at the different things that are being done. And all of these are, to one degree
or another, using the electronic health record as a way to provide appropriate notification
or support to clinicians to utilize the information.
One of the things that we referenced earlier was the eMERGE Infobutton Project and there
are two objectives to this. And I thank Casey Overby who is leading this project, along
with Luke Rasmussen, for their slides. Objective 1 is to develop a new information resource
based on eMERGE II and PGx scenarios. So, infobuttons are a way to use the context of
where a provider is in the electronic health record to essentially pre-ask the question
that the provider is likely to have, and then, with one click, allow access to a resource
that specifically addresses that question. So -- and there's been formative research
on this that looks at the ontology of clinical questions, and you can map this pretty well.
So, for PGx, it's actually pretty easy, because if you're on a drug-order entry system and
you click on a drug to order, then we can pretty well decide the questions that you're
going to have about that particular medication, and, specifically, if we're interested pharmacogenomics,
we can present the relevant information about pharmacogenomics very quickly. So this talks
about the different parts of this objective that we're involved in. This is ongoing, so
we've completed a collection of scenarios and have designed a template. What we're trying
to do is to use a standardized format for this information sheet that can be used across
the sites. We're in the process of having our content developers complete the eMERGE
template, and then we're going to be using that to engage end users and evaluate the
resource.
The second objective is actually then to implement infobuttons within the electronic health records.
Again, you can see the status of this at the present time. One of the reasons that infobuttons
are a good target is that they are required by meaningful use to be included, and there's
an infobutton standard that exists. And so if we can actually develop infobuttons that
work, this is a way that we could implement this information within any certified electronic
health record. And we've been collaborating with University of Utah, particularly Ken
Kawamoto and his group, who leads the OpenInfobutton system. So, this is something that's ongoing,
and we hope that this will be the first example of something that comes out of eMERGE in the
electronic health record space that could be truly generalizable with little, if any,
local customization, but we remain to -- it remains to be seen. And Ken just messaged
me. It's actually Guilherme Del Fiol who's leading OpenInfobutton, so, sorry about that,
Guilherme, even though he's not on the phone.
The challenges, we've heard this one. Implementation of research informatics into the clinical
EHR system is really hard. I decided to illustrate this by a photograph. This was taken at the
end of a recent meeting here at Geisinger where I presented some of our implementation
ideas to the clinical informatics group. There were no serious injuries or fatalities but
it was a close thing. One successful implementation equals one successful implementation even
at the same site. And I think there has been a certain amount of tension and frustration
that's been experienced, given that a lot of other eMERGE activities are moving forward
much faster than the glacial pace that EHRI is moving forward. We are working now on looking
at some network outcomes related to EHR integration, but almost all of these are process outcomes,
even though we all know we would like to have clinical outcomes that we can look at, and
there are some sites that have developed some ability to look at clinical outcomes.
So we heard this from Justin earlier, the Clayton aphorism: Each system is built three
times -- the first time to see can it be built, the second time to determine how it should
be built, and the third to actually build it. So, our status, using the Clayton-Starren
algorithm here, is that we really -- most of what we're doing here in Phase 2 relates
to Clayton first stage, with some hope of learning enough to proceed with Stage 2, and
maybe at some individual sites to actually proceed into Stage 2. And I think the PREDICT
program at Vanderbilt is an example of an EHR implementation that is actually into Stage
3, but we don't have anything system-wide at eMERGE that we can point to at that level
of maturity.
As has been referenced previously, the EHRI group works closely with two other eMERGE
work groups: the Return of Results and CERC, which we've already heard from. There's also
an informatics group associated with the Clinical Sequencing Exploratory Research efforts which
are funded through the NHGRI as well. And we have had joint meetings between our two
informatics groups and have agreed to produce a white paper on where in the EHR this information
should appear.
So, future direction and opportunities. We think there's a research agenda around actionable
clinical decision support, so what are the optimum ways to centralize and distribute
standardized evidence-based clinical decision support. I referenced the Clinical Decision
Support Consortium earlier. This is work that they had done through their AHRQ-funded grant.
To put it in perspective, and I don't think Blackford would get horribly angry with me
if I said this, but they have six years of funding with literally a who's who at the
top clinical informatics groups in the country, and at the end of that, they had one clinical
decision support role that could have been said to meet that first bullet. So this is
really challenging, but, again, they have decided to sign up for a second go around,
and, as mentioned, they're very interested in engaging with us, and we've had some communication
with them at a previous work group session at an eMERGE meeting, and we will look to
do more of that going forward.
We also want to determine how accurate -- and there're different ways to define accuracy
which I'm not going to go into now -- but how accurate does clinical decision support
work across genomic medicine use cases? In other words, are there certain standardized
ways of doing clinical decision support that are more robust and are less likely to need
customization, because when we talk about implementation, there are really two issues
that need to be measured. One is fidelity: Does it work the same as Site B as it did
in Site A, and how much difference is there? And then customizability, which is how much
customization does it take at Site B to make it work like Site A? And those are critical
issues to study.
And then, ultimately, we would want to move beyond process outcomes to relevant clinical
outcomes for selected genomic medicine use cases. We want to be able to study the ability
to extract real-time patient-level data from transactional electronic health records to
fire clinical decision support for selected genomic medicine use cases. So, the common
example we've been using with PGx is you want to go to order a medication, you want to order
simvastatin, you click on the simvastatin, the system recognizes that a pharmacogenomic
test has been done, it looks for the result, and if the *5 allele of SLCO1BI is presented,
then the clinician would be notified in some way that perhaps a statin other than simvastatin
should be used because of the risk of statin-related muscle injury.
But what we'd really like to move to is much more complex clinical context issues. An example
that I use is it is now routine in pediatrics to screen for autism between 18 months and
2 years. If, for some reason or another, we had a whole genome or a whole exome on a patient,
and they did that screening test and it was screen positive for suspected autism, one
could shortcut the diagnostic odyssey by immediately generating a query of all autism-related genes
to say is there a known deleterious mutation in one of these genes that could potentially
confirm the diagnosis and the specific cause of autism in an individual. So, again, that's
just sort of one example of a much more complex clinical context where you'd want something
lurking in the electronic health record to be able to fire and answer questions that
clinicians may not know even need to be asked.
And then, ultimately, to study how electronic health records, personal health records that
are tethered to EHRs, and patient portals can be used to enhance education of patients
and providers and measure the comparative effectiveness of different approaches. And
I think as we talked about in the prior discussion, it maybe extends beyond just education to
who is best able to manage the results. There has been a sidebar conversation about the
dangers of, you know, the consequences, if you will, of giving information to patients.
But the patient is the only person that's the constant in the healthcare delivery system.
So if we do a genome here at Geisinger on a patient and they go to see somebody at CHOP,
they won't have that genome to utilize, whereas the patient, if they have the genome, could
potentially use it in some way, shape, or form. That remains to be studied.
And that's the end of my presentation, so I'll turn the control back over to Brandy.