Ben,
1. Getting patients with diabetes into care.
They are identified with an carefully developed algorithm run
automatically against the data warehouse. If they have a pre-existing
relationship with one of our care teams, they are invited by (or in the
name of) that care team to come in for a visit, where the diagnosis is
reviewed and a care plan negotiated. If they have a non-Geisinger PCP,
we try to support the patient's connection with that PCP, providing the
information we have to the patient and PCP.
2. Observational data is still observational data, no matter how much
of it there is.
Absolutely.
That is why I suggested the (rough) stratification earlier.
a. If observational data suggests that a me-too drug like Vioxx is
dangerous, the that data is fully strong enough to support removal from
a formulary.
b. Observational data is fundamentally inadequate to motivate
preventitive interventions in asymptomatic patients, as HRT should have
taught us.
c. If a patient is very sick and the best evidence regarding
appropriate care is based on observational data, that is the evidence to
act on--cautiously and with attentive monitoring.
3. Data mining is limited by more than the fundamental limits that
apply to observation data.
a. Most fundamentally, GIGO--garbage in is still garbage out. Data
mining will certainly reveal the inadequacies of our information
creation and recording practices; the rest is currently promissory
note.
b. There may also be limited cases where data mining suggests useful
hypotheses, but current limitations of precision and recall (except in
the narrowest of domains) mean that such a possibility continues to
recede into the future.
3. At Geisinger, the genomic data and phenotypic data in the EHR will
be used by a robust research team to conduct experimental studies, some
of which will yield evidence that will come to guide care. Observations
made in this data system will be regarded as hypothesis-generating.
Best regards.
Jim
James M. Walker, MD, FACP
Chief Health Information Officer
Geisinger Health System
>>> vijaya madhavan <[log in to unmask]> 11/7/2011 8:17 AM
>>>
Ben
Would appreciate a copy of the paper.
I share your reservations about the genomic analysis efforts on the
grounds
that I cannot concieve that any health system has the time resources
to
adequately interpret such data against the extreme variation that
healthcare interventions at the user interface generastes.
The grander the vision the higher the complexity generated and the
harder
it becomes to reach meaningful end points.
Vijaya Madhavan
MD- Personalised Intelligence Ltd
On 7 November 2011 12:44, Djulbegovic, Benjamin
<[log in to unmask]>wrote:
> I think we need to separate the role of the quality
improvement/monitoring
> (where quick access to medical records should prove invaluable) from
> discovery efforts (where the use of EMR will not revolutionize
clinical
> research as it is often portrayed). As we argued in our paper [
> Implications of the Principle of Question Propagation for
> Comparative-Effectiveness and *Data Mining* Research (JAMA Jan
19, 2011)],
> and which I sent to Jim in a separate e-mail, data-mining of EMRs to
> discover effect of new treatments etc will, due to fundamental
> epistemological reasons, always remain hypothesis-generating
exercise
> (invariably requiring further prospective data collection to refute
or
> corroborate the retrospective nature of EMR data-mining). I
personally
> believe that the current fad about collecting blood and tissue
samples for
> genomic analysis to be later correlated with EHR records is huge
waste of
> resources (and this is going on in every single institution across
the US
> and probably elsewhere). This, of course, is not criticism of Jim's
> phenomenal efforts and his reputable organization- I am only
reminding
> people of basic principles of philosophy of science that tend to be
> forgotten as new gadgets are being invented.
> Let me know if you need a copy of the paper
> Ben
> Sent from my iPad
>
> On Nov 6, 2011, at 11:17 PM, "Dr. Amy Price" <[log in to unmask]>
wrote:
>
> > Hi Jim,
> >
> > Thank you, This is how I was hoping the results could be used for
as
> > improved outcomes in conjunction with evidence based care
processes. I
> like
> > the concept of genomics inclusion a lot, in fact I wondered if this
was
> > possible. I am interested in the diabetes and how you brought them
into
> > effective care, this is really such an important contribution and
the
> > cardiovascular intervention improvements are huge.
> >
> > I understand people emphacizing the qualitative aspect and their
fears in
> > this regard but I see a lot of people where critical information is
just
> > missing and that makes any kind of intervention less effective and
more
> > costly in time and resources plus I think identifying patterns of
> successful
> > care and rejecting those that look good in the lab but don't work
in real
> > life would help everyone.
> >
> > I appreciate this information and your response,
> >
> > Amy
> >
> >
> > -----Original Message-----
> > From: Evidence based health (EBH)
> > [mailto:[log in to unmask]] On Behalf Of Jim
Walker
> > Sent: 06 November 2011 10:39 PM
> > To: [log in to unmask]
> > Subject: Re: Evidence-Based Medicine in the EMR Era
> >
> > Hi Amy.
> >
> >
> > First, to the very far from easy--granting the accuracy of what
> Marguerite
> > says:
> > For most useful analytics, we combine information from the EHR and
12 or
> so
> > other databases (including claims) into a data warehouse, where all
but
> the
> > simplest analytics are performed. The normalization of this data is
a
> > painstaking process. Recommendations are then fed back into the
EHR, the
> > only interface that most of our clinicians use. (Our Keystone
Beacon
> > Community has developed a community data warehouse to which
different
> > organizations contribute their data to support shared care
processes
> across
> > five rural Pennsylvania counties.)
> >
> >
> > Second, most of our analytics begin with evidence-based care
processes,
> not
> > trolling for correlations. (Kaiser has the numbers for doing this
more
> > effectively.) Beginning with process measures that have been
validated as
> > leading to improved outcomes (e.g., retinal exams for people with
> diabetes),
> > we create care processes to assure that the interventions are
offered to
> > patients 100% of the time. Then we measure process performance and
> patient
> > outcomes looking for correlations (or lack thereof) to guide
process
> > refinement.
> >
> >
> > For example, our cardiovascular surgeons identified 38
interventions they
> > believed to have been demonstrated to reduce complications in
patients
> > undergoing elective CABG. At the outset, our patients got them 69%
of the
> > time (better than national benchmark). Within 3 months of
re-designing
> the
> > process and fitting the EHR to support the team, we were at 95%. We
have
> > been above 98% for the last 2 or 3 years. Outcomes such as return
to
> work,
> > rate of wound infections, etc. have tended to improve, some of
them
> > statistically significantly. Correlations between individual
> interventions
> > and outcomes are performed regularly, with adjustments in
interventions
> > based on the results.
> >
> >
> > Using this approach in our patient-centered medical home, we have
> decreased
> > all-cause admissions in a population of 22,000 Medicare patients.
> >
> >
> > We currently are using this approach with 15-20 acute problems
(e.g.,
> > cataract surgery) and another 15-20 chronic problems.
> >
> >
> > We also scan the database (primarily EHR data) for, e.g., patients
with
> GFR
> > < 60 but without any record of assessment or management of CKD and
> patients
> > with HbA1c > 7 and no management of diabetes. We have identified
> thousands
> > of such patients and brought them into effective care processes.
> >
> >
> > We have collected 100,000+ blood samples for genomic analysis to
be
> > correlated with EHR records regarding, e.g., age, gender, problem
list,
> and
> > response to various drugs.
> >
> >
> > If I'm talking past your question, please let me know
> >
> >
> > Best regards.
> >
> >
> > Jim
> >
> > James M. Walker, MD, FACP
> > Chief Medical Information Officer
> > Geisinger Health System
> >
> > The best way to predict the future is to invent it.
> > - Alan Kay
> >>>> "[log in to unmask]" 11/05/11 10:13 AM >>>
> > Dear Jim and Marguerite,
> >
> > Could you elaborate on this? Easy would be awesome but effective
with
> > benefits is even better! My imagination has been sparked by all
this and
> > although it could be an IT nightmare at this point maybe it is an
> > information goldmine. Banks etc mine data relentlessly for their
own
> ends to
> > good advantage, it makes sense to Find ways this data could be
useful to
> > progress science and help people
> >
> > Amy
> >
> > Amy Price PhD
> > Empower 2 Go
> > Building Brain Potential
> > http://empower2go.com <http://empower2go.com/>
> > Sent from my iPad
> >
> > On 5 Nov 2011, at 09:57, Jim Walker wrote:
> >
> >> Thanks, Marguerite.
> >> It is very far from easy, but you and we do it all the time with
> >> considerable benefits.
> >> Jim
> >>
> >> James M. Walker, MD, FACP
> >> Chief Medical Information Officer
> >> Geisinger Health System
> >>
> >> The best way to predict the future is to invent it.
> >> - Alan Kay
> >>>>> Marguerite Koster 11/04/11 4:29 PM >>>
> >> Hi -
> >>
> >> Kaiser's decision to remove Vioxx from its regional formularies
occurred
> >>
> >> before the implementation of the organization's EMR. I should add
that
> >> EMR's are still in their infancy, and usually built as clinical
> >> management
> >> systems, rather than systems for robust data extraction and
analysis.
> >> People sometimes have the notion that extracting data from an EMR
is an
> >> easy process. Far from it, though, especially if you want a clean
and
> >> accurate dataset. There are also issues with data that is only
> >> available
> >> from EMRs in text format.
> >>
> >> Marguerite
> >>
> >> NOTICE TO RECIPIENT: If you are not the intended recipient of
this
> >> e-mail, you are prohibited from sharing, copying, or otherwise
using or
> >> disclosing its contents. If you have received this e-mail in
error,
> >> please notify the sender immediately by reply e-mail and
permanently
> >> delete this e-mail and any attachments without reading, forwarding
or
> >> saving them. Thank you.
> >>
> >>
> >>
> >>
> >> "Djulbegovic, Benjamin"
> >> Sent by: "Evidence based health (EBH)"
> >>
> >
> >
> >
> >
> > IMPORTANT WARNING: The information in this message (and the
documents
> > attached to it, if any) is confidential and may be legally
privileged.
> It is
> > intended solely for the addressee. Access to this message by anyone
else
> is
> > unauthorized. If you are not the intended recipient, any
disclosure,
> > copying, distribution or any action taken, or omitted to be taken,
in
> > reliance on it is prohibited and may be unlawful. If you have
received
> this
> > message in error, please delete all electronic copies of this
message
> (and
> > the documents attached to it, if any), destroy any hard copies you
may
> have
> > created and notify me immediately by replying to this email. Thank
you.
> >
> > Geisinger Health System utilizes an encryption process to
safeguard
> > Protected Health Information and other confidential data contained
in
> > external e-mail messages. If email is encrypted, the recipient
will
> receive
> > an e-mail instructing them to sign on to the Geisinger Health
System
> Secure
> > E-mail Message Center to retrieve the encrypted e-mail.
>
|