Good morning my name is Dr. RanellMyles and I want to welcome you to the NIH Office of Disease Prevention
Medicine: Mind the Gap Webinar Series this webinar series explores research design
measurement intervention data analysis and other methods of interest to
prevention science our goal is to engage the prevention research community and
thought-provoking discussions to promote the use of the best available methods
intervention research and to support the development of better methods before we
begin I have some housekeeping items to submit questions during the webinar
there are two options first you may submit questions via WebEx by clicking
on the question mark in the WebEx toolbar and direct questions to all
panelists second you may participate by Twitter and submit questions using the
hashtag #NIHMTG at the conclusion of today's talk we will open the floor to
questions that have been submitted via WebEx and Twitter lastly we would
appreciate your feedback about today's webinar upon closing the WebEx meeting
you'll be directed to a website to complete a seminar evaluation and also
sent a link an email with a link to the evaluation your insights will help us to
continue to improve this webinar series at this time I'd like to turn things
over to Dr. David M. Murray Associate Director for Prevention and Director of
the Office of Disease Prevention thank you Ranell.
Dr. William Vollmer is a
biostatistician and health services researcher who's been conducting
collaborative multidisciplinary research at Kaiser Permanente Center for Health
Research for over 30 years his early work focused on the epidemiology and
management of asthma and chronic obstructive pulmonary disease within
that health plan as well as on a series of collaborative clinical trials and
nonpharmacologic approaches the control of blood pressure and weight which he
directed or helped to direct the data coordination more recently his research
has increasingly focused on the use of large pragmatic trials the test
strategies to improve population-based management of chronic diseases he served
as principal investigator for a two region study focused on improving
adherence to statins angiotensin converting enzyme
inhibitors and angiotensin perception blockers in adults with diabetes or existence in DD
the three regions study involved over 26,000 members Dr. Vollmer currently
serves as the statistician with two projects funded as part of the NIH
collaboratory study and pain management among members of three Kaiser Permanente regions and a study to improve colorectal screening among
members of 26 federally qualified health clinics it's my pleasure to welcome Dr.
William Vollmer and turn the session over to you thanks David for the
introduction and I appreciate the opportunity to present that in it so let
me just begin with some disclosures I. Work for Kaiser Permanente for 31 years
now and I've been a member of Kaiser for 35 years it handedly biases my view that
this is an ideal setting for doing prevention based research and leveraging
the EHR to do that and since it's a system that I that I know well and I'm
essentially working almost my entire career it's where I'll focus most of my
comments in terms of examples however I'll try to illustrate where you can
translate similar concepts elsewhere basically I think this model shows
you what's what's feasible sort of the universe the possibilities
are and in other settings which may have less rich elements you may not be able
to do the full scope of things that sometimes able to do so just as a
background when I join Kaiser 31 years ago the electronic health record was a
rarity certainly Kaiser didn't have one and there weren't many people around the
country that did that landscape is hugely changed and it was helped along
in 2009 when as part of the Affordable Care Act the federal government set the
adoption and meaningful use of electronic health records is a national
goal and they incentivized it now as of 2015 as estimated eighty seven percent
of U.S.
Physicians use some form of electronic health record and that I have
as a 2014 is that more than 80 percent of hospitals
have adopted some form of electronic health record so it's now out there
even if the ACA goes away I think this is a movement that's going to continue
to see more and more of it so it's going to be with us for the foreseeable future
so the objectives of my talk I want to highlight the opportunities and
challenges for using the electronic health record for prevention research
and I'll provide some illustrious of examples from my work and others of my
colleagues we're going to focus primarily as I said in the world of nonprofit managed care which is what I know best and I'm going to largely ignore a
much broader class of epidemiologic and health services research that's enabled
or facilitated by the EHR so we consuming is a long time a lot of it
wouldn't fall per se under the rubric of prevention research but I think it's
still quite obviously relevant research so what do I mean by the EHR so in my mind its two dimensions one is content what information is available to you to work
with but equally important is the context so what's the nature of the
population and/or care delivery setting about which we have information and it's
these two dimensions content and context that collectively define the types of
research that we can carry out and I'll try to illustrate that a little bit as
we go on so content what information is available obviously you're going to have
some form of healthcare utilization and that could include both usage and
outcomes and so by usage and I might get a lab test I might get a fasting lipid panel the outcomes would be but there one of the results of that test so
and sometimes you're only going to know that the test was done and other
situations you may have the detailed results of those tests and obviously you
can do more if you have the latter we have contextually inpatient outpatient
pharmacy labs and genetics this was wanted on everything
that you could imagine in the EMR could be available to you but not everybody
has access to the full panoply of information ideally you'd like to have
that if it's linkable over time that's certainly going to extend the ability of
what you can do with the data but that's going to imply the existence of a
consistent unique patient identifier so when I've worked in the past at the
National Hospital Information System a national database that's a hospital event-based record one record for hospitalization
and those records can't be linked over time so that's a situation where you
have a lot of that is a lot of same individuals with multiple records in
there but you can't tell who those are but when you have a situation like I
have a Kaiser with VA or many many other settings where you have patients that
you're following and they have a unique ID then you can can track them over time
you can also have patient reported outcomes so Kaiser collects data on
paying health status and they continue to add other health patient reported
measures over time I think the recent introduction of PCORI the
Patient-Centered Outcomes Research Institute has to do to certainly help to push this
movement and we're seeing it is increasingly more important to gather
your EMR that you're dealing with maybe a registry but also within a broader
electronic health record there may be pieces of it that are registry so Kaiser
keeps registries of patients with diabetes patients with heart disease
they have various high risk populations people who are high utilizers of care
regardless of the disease so there's a lot different registries that are being
kept and they help Kaiser do their job so you may have access to that to a
registry within a larger EHR or the registry maybe the EHR that you have to
work with you're going to get patient demographic and risk factor information
age sex race smoking BMI physical activity some subsets
at a minimum you can probably expect to find eight insects but hopefully some
race information and smoking is increasingly begun in common you may
also find data on insurance status whether your Medicaid or Medicare or how
your being covered in terms of population delivery setting there's a
managed care environment which gives you the sort of the richest population to
work with and causes an example but there's there are many many other
examples of that basically it's a situation we have a well-defined
population I think that's really key it enables an awful lot of additional work
that you can do there are comprehensive services we typically have minimal use
of outside care so you pay Kaiser a fee they manage your health and and they're not
going to pay for outside care unless it's urgent care so if you do have acute care
and you're not near a Kaiser facility you can file a claim you can reimbursed
for that and then we can pick up that that utilization from your claims
database so I here at Kaiser we do a good job of capturing outside care and
we believe from work they've done that there's really not a whole lot of
outside services that we're missing increasingly these days pharmacy may be
an issue where there's becoming more and more options for discount pharmacies
that may be more competitive with Kaiser or whatever health aid you happen to be
in so so that becomes perhaps issue and and you have pretty exhaustive EHR based
information including a lot of fields that are captured in text or remembered
form they're easy to abstract and but you also have a lot of text information
which can be searched now with natural language processing software so that's
an extra dimension moving away from to the managed care you might look at for
instance federally qualified health centers so I'm doing a study as part of
the HMO collaboratory that involves 26 federally qualified health
these individual health planners don't have well-defined populations they may
or may not offer comprehensive services they probably don't capture outside care
as well as Kaiser does and what's in the EHR may be limited certainly relative to
what I'm used to it at Kaiser and then it even going further away you might
look at a fee-for-service hospital setting again the population is not well
defined and now you may have no or very limited allocation so and there's other
examples along the way but this this gives you a wide range of spectrums and
you get well imagine that depending upon which setting you're in you're going to
be able to ask different sorts of questions so as an example how content
and context constrained the use I was involved with a network concert that was
set up to carry out comparative effectiveness research related to COPD
chronic obstructive pulmonary disease and we had a wide mix I think of a six
or eight different delivery systems around the country and the the diversity
of the settings added generalizability to our findings but they didn't have
well-defined populations and the same breadth of EHR data what we did have in
common that everybody I was inpatient data and so the limited the types of
questions that we could ask we certainly look at at studies of treatment for
acute exacerbations for presenting the hospital they can look at post discharge
follow-up studies put them in population-based disease management for
the whole collective of concert what's much more problematic we can do that
within a couple of sub entities that had similar systems we had a VA Center and
Seattle for instance that looks very similar in terms of what they had to to
Kaiser but but not all of them did and so it's just something to be aware of as
you're as you're going forward and thinking about using the EHR we're doing
research so I'm going to take a minute here and talk about different ways that
I might use the electronic health record to facilitate research one obvious told
it to the around forever is using it as a
recruitment tool so it becomes in a very efficient way to target specific
populations of interest and you can easily be populations that you might
identify and recruit independently so I. I would simply use the EHR to find
people and then I look phone them up or email them or some whatever contact them
to recruit them into a study but you could also use the EHR to recruit a
point of service so you can have pop-up set in sort of say here's a patient of
care and try to get them into the study right away or we used to send people
into the clinic so you get research fact that they'd be embedded in the clinic so
many people were coming through and you meet them in the clinics in the waiting
room and try to recruit them at that point so there are different ways to to
use it that way also once you've had been recruited you can use the EHR to
flag participation in a specific study might might facilitate the ongoing
provision of services in that study people are aware that you're in this
service and they're going to try to measure them they give you care this
compliant with whatever the protocol is from that study the other way we might
think of using the EHR and this is recruitment it's outcomes assessment and
a big strength here this has become a big factor with the movement or
pragmatic clinical trials it is to passively capture outcome information so
obviously we can passively capture that on healthcare utilization and cost of
care we can capture clinical outcomes these are one function test labs and
again we might capture the fact that they were done or ideally even the
results of those tests physical measurements patient reported outcomes
all the things that I was mentioning earlier we should be able to passively
capture from the medical record health behaviors demographics I would take a
minute to since this is a prevention oriented audience a colleague of mine Tom Vote
several years ago came up with something he called the prevention index and this
is a novel metric for quantifying in essence the proportion of time in a
given window given target interval that a person is compliant with the United
States Preventive Services Task Force guidelines for select preventive
services so let's take colorectal cancer screening I have someone and beginning
of year let's put the 2016 I say well do you meet guidelines where you in the
right age range and you go to already have colorectal cancers are like the
screen for if you like and know that notice that you have it or not but if
you in fact do me guidelines for screening and the next question is are
you up to date with your screening requirements and I can determine for any
given day in that year whether you are currently up to date and so over the
course of the year I might say for what proportion of the year was I fully
compliant with the guideline now this is in contrast to what the HEDIS
is might do HEDIS would simply say I. Look over the year and I say these are
people that should have been screened and were they screened during the year
and it doesn't really care whether that person was screened at the end of the
year or the beginning of the year and just says who has a service during the
year so the prevention index gives you a finer tune tools for looking at
prevention related questions and it has the advantage you can aggregate the
scores at the provider level or the clinic level or the system level to have
very obvious questions about the differences in services or provision of
a preventive care by these various dimensions as the HEDIS using a tool
like this requires the sufficient window of observation to define who even is in
need of a given service so again in the example of colorectal cancer screening
you don't like to have a ten-year history because if you had a colonoscopy
up to ten years ago you're currently covered and there's flexible
sigmoidoscopy five years ago so you'd like to be able to go back now we can't
necessarily go back that far for everybody but we will least try to look
back as far as we can but almost any measure going to use is going to have to
have some window to see well who meets criteria
for screening they begin and look for them during the screening window and see
whether they actually got that service so that is the requirement in all these
is you can sort of track individuals back in time and in terms of using the
EHR outcomes except just a few comments on challenges and doing this one is the
validity and outcome so long as you have a diagnosis of COPD or a diagnosis of
asthma or diagnosis of myocardial infarction is that a valid outcome
it's a good question it may or may not be relevant for the work that we were
doing so a lot of classic health services research just said whether it's
accurate or not it is what it is it is with the EHR shows and one of the
outcomes that people go with carry a diagnosis of X Y or Z in the EHR how are
they treated what are their outcomes you may get poor outcomes because the
diagnosis is bad and unless that is what's in there and how they're being
treated so a certain class of questions doesn't really care about that if you're
doing a clinical trial and your primary outcome is myocardial infarction or
heart failure or COPD wearing that exacerbation of COPD then they might be
very important to actually say let's validate that out so let's make sure it
really means rigorous criteria and a traditional tundra trial you're going to
have a a adjudication committee that's going to review the records and here
what we do is we can pull the outcomes from the EHR and pass those on to some
committee that can verify that you have to be careful now not just pulling
records for people who have the purported condition but pulling records
for other conditions that might have been where they've been misclassified so
if you're looking for COPD and maybe pull out asthma hospitalizations or
hospitals it is for other conditions that might be a mask as COPD so would
look care there analytically you only get challenges we
don't have sort of what we call rectangular data sets you could have a
varying number of observations and spacing of observations for individuals
and that's been have implications for how we analyze
the data details I won't get into here but just something to be aware of a
really critical issue becomes data harmonization over time and across sites
and by data harmonization I mean does the data mean the same thing over time
or does with that I mean the same thing from site A to site B to site C a very
good example of harmonization over time relates to simply how the code things
we've just gone from ICD-9 to ICD-10 and the would be coded thing that ICD-9 is no
longer the same in some cases we have we typically gotten more finer grain so we
can clump things at 10 and go back to 9 typically the case but sometimes you
group going forward and we no longer can separate a diagnosis and 10 it really is
a combination of two separate diagnosis in ICD-9 so that can pose some
challenges for us there's also a situation I dealt with years ago this
Kaiser changes the way when organization changes where they provide services and
I was planning our early studies we're looking at long term trends and
utilization for asthma and COPD and we found a drop-off and at hospitalizations
at some one point when you look at a broader definition of ED based care
episodes of ED based care which would include hospitalizations we saw that that blip
that drop-off didn't occur anymore and we did some further exploration and
discovered that at about that time Kaiser had introduced what they call
short stay units if you were hospital a few words put into the ER per se asthma
exacerbation we can keep you under observation in the ED for up to 24
hours without formally admitting you and so it no longer counted as a
hospitalization whereas the year before what it counted as a hospitalization and
so we think that that drop-off there was a total artifact related to this
movement towards these short stay unit over
I think that the time frame in the short stage is actually increasing that you
can be instructive 48 hours now I'm not positive on that so that's an example of
how care changes another thing is we didn't use to have
urban care services we had an emergency department and people came in to the ED
and they were triaged and literally one of our hospitals is sort of would
put in one of two sides of the central corridor as you came in and the more
urgent cases went to the right and the less urgent cases event to the left of this
corridor and and they realized there was a pretty predictable group of people
that were in that less urgent bucket they needed to be seen but not as
critically as the classic emergency department patients and we eventually
opened a an urgent care clinic right next to our emergency department and and
and that was sort of the evolution of that and you've grown with Kaiser and
elsewhere for any myriad so there's shades of urgency care services we have
after hours clinics now and a whole variety of things the people that we've
been seeing traditionally in one setting may be seen in different settings and
and looking again over time and across centers becomes challenging it to
reconcile well what are you calling this and what am i calling us what is a
hospitalization really mean what is emergent care and really mean what is
urgent here really means how do you define it
so all that harmonization has to happen and it's a big big deal missing data so
you may feel it if you have a traditional clinical trial you're going
to bring people in take the measurements and you have some control over that when
you're looking at the EHR you have what you have another kind of missing data in
some senses is that the data and they have it outside of Kaiser so somebody
really got a flu shot but they didn't get it in Kaiser and so I don't know
what happens I can't tell my analysis whether it didn't happen or it happened
elsewhere but if you're looking at predictors of who gets a flu shot you need
to be aware that people that don't have a flu shot in your system may not all
not really have a flu shot this may have gotten outside so factors
to be thinking about in the final piece here in terms of how we might use the
EHR as an intervention delivery tool and I think the simple way to think about
this it's not going to use the EHR to give clinicians what they need when they
need it you're false for evidence-based care I
think that's really the crux and it's certainly a Kaiser they've worked very
hard to the issue of only what you need when you need it because there's a huge
tendency and problem with overwhelming providers quickly primary care with too
much information information overload and the EHR really gives you the
opportunity to show you just what you need when you need it and that's that I
think really really important so types of ways that you might do this prompts
for guideline based services and get flu shots or screening for whatever needed
blood work point of service alerts or a variety of things including things like
the flu shot specifically for other other issues you can get system
generated reminders phone calls snail mail email and I'll top up that way
medication veto reminders those can either happen as an alert to the doc to
let you know or they can be sent to you directly pop up flags for medical
contraindications so you go in to order a given medication instead oh by the way
this patients taking X Y and Z as well and this may be contraindicated are you
sure you want to do this medication use profiles that have been a lot of work
with asthma over the years and one of the things that was very fashionable
for a minute was to look at the ratio of controller used to reliever
medication use and if you were a patient that was using a lot more relievers beta
agonist were all due to a controllers and a inhaled corticosteroid then we would
have suggested that perhaps you needed to you were a more severe patient and you
needed to be put on controllers or paid more attention to your regular
controller use because you were being overwhelmed and
having too much reliever use again you get flagged populations for a high the
high-risk populations for care and case management so these are all the examples
of the kinds of things that you might do using the EHR as an intervention
delivery tool I'm going to just a note here on sort of given the realm of
universe of what you can do how do you decide what to focus on and here I'll
give you some lessons some hard earned lessons from my 31 years working in your
organization when I first started out I. Had a position partner from the medical
school an academic researcher and we would cook up ideas and try as we would
sort of play in their sandbox they let it do their studies but begin to realize
they weren't really invested in our studies we're never happy to have us
look at it and if they were successful they may or may not pick it up but but
I've come to realize that where that where the bang is or where the bang for the buck is
looking at the organizational priorities I can do the organization tell me what
they're passionate about and see if I.
Can bring tools of research to help them
best that's where the fun is it's really a hard work but that's also where the
biggest payoff is and I think managed care organizations like Kaiser are
heavily focused on on primary and secondary prevention and and so we can
ask how they can help that as an example I did a sabbatical in our Hawaii Region
several years back I was talking to someone and they said yeah we've got
this new initiative we've got we've identified our top 1% of care utilizers
in the last year and it's about 10,000 of them we have money to go after all
10,000 but we're going to assign a care manager there's a top thousand in that
group and and we want to see you know what the benefit is what the savings are
for for doing a service plan I said well you know there's a position on his
regression to the mean to do any population based on being at the
extremes of utilization either high or low they're going to tend to come back
for the mean and subsequent years and so you
don't know how much of that is just was going to happen anyway and how much
Institute you what you what your intervention was and so I said since you
got 10,000 people you can't deal with them all anyway go and get with a
thousand item to randomly assign the top two thousand and the two groups and half
of them get the intervention then you have a true comparison group a very
comparable group and you can measure the passive sort of regression to the mean
and differ the incremental effect from your intervention that's a way to bring
some tools of research to what they're doing and they're going to do anyway
cannot slow them down it's always a big issue within it so we don't want it we
need to get this done we can't wait research is a congress activity that
takes lots and lots of time but you can just give them something to do some
tools and design thoughts to help them to being right away diabetes prevention
is something I think I'm very interested in in recent years and a met with the
head of our diabetes prevention diabetes management population management group
at Kaiser and say well what do you guys have doing and we're looking
to see whether we're ways that I can sort of help you with what you're doing
he said well you know a huge issue right now pre-diabetes we have a lot of
diabetics we have in five six times as many pre-diabetic since this is this
avalanche is coming towards us and we really want to get on top of this and
work to keep them from delving diabetes so classic profession and I said great
so what do you plan on doing and he said we have these ideas and I said so would you
be interested in looking at some alternative strategies for managing this
population segment a little more high intensity or lower intensity because it
to the cost-effectiveness of different strategies and they were very receptive
to that idea so again taking something that the organization was really cared
about was passionate about that how can I. Bring the tools in the search to help
you answer the questions that are passionate and important for you I think
following this strategy is really the best way to ensure that you have at the
end of the day a high relevant topic and are likely to get
good organizational buy-in and that latter really is important in doing some of these large trials so now I'm going to move into the variety case study but I'm
not going to talk about the outcomes of these studies and I'm going to more talk
about the design and some of the EHR. Based issues that we've faced and doing
and just to give you a flavor of the kinds of work that you might be able to
do so this is a study patient looking at booking and medication adherence is a
one-year parallel arm pragmatic clinical trial it set the two health information
technology based strategies versus usual care to improve adherence to
cardiovascular disease medications we randomized almost 22,000 patients with
diabetes or a thorough sporadic CVD from three different Kaiser regions into one
of three study arms in the usual care will be called interactive voice
recognition you get an automated reminder or enhanced IVR you got the
reminder plus you got various educational material including a
personalized health report periodically to see how those compare our primary
outcome was that here is measured through the medical records and
secondary outcomes again July no medical records or BP and lipid levels so some
key key EHR elements we wanted to set up a study to look the labels look in real
life but also so that if it worked Kaiser could just keep it rolling so we
didn't do a one-time enrollment we set it up to do ongoing enrollment so every
day we were refreshing that the databases who was newly eligible and
getting them randomized into the studies so we've set it up that way all done
electronically we created for this one our personalized health reports that
have information from the EHR not only your medication use but also
information on blood pressure and lipids hemoglobin A1c if that was
relevant for you and we might be able to save rent since we see that your blood
pressure is running in the high range and by the way your use of your blood
pressure medication is not as good as it could be if you were to increase a shoe
might be able to bring your blood pressure down so the hope was that the
help the personalized help that it could be a slope to improve motivation for
adherence we allowed some site flexibility and how the interventions
were delivered in order to fit stakeholder priorities in need so we had
an external service out of Boston that was doing the calling but our Georgia
regions said you know we have our own internal service to be used for this and
we'd really prefer to do that and so we were able to work with them if they will
let's add an extra arm in your site and we will actually have one service one
arm and uses our external calling service and one that uses yours and we
can compare how those results work out then they said that's great that meets
our needs and so again we allow some of that flexibility also that different
regions had different ways in which they would do programs reminder programs some
of them routinely called have people hold the medical records to make sure
people were really appropriate for this and that's how they did it so we said do
it the way you would do it normally that's what were' going to be looking at it gives
you a little more variability which is gets away from the classic language file
but it's very much in the nature and spirit of pragmatic kind of the files
it's looking at how people do things in the real world and we leverage the
existing virtual data warehouse to define population and outcomes and
consistent manner so Kaiser is actually is not one big homogeneous entity but a collection of different entities we have I don't know 12 or 13
regions and each region has a medical record based on epic sub they're all
what slightly different from one another and they don't inter communicate
seamlessly so it's part of a larger research network group the Healthcare
Systems Research Network we've developed a virtual data warehouse to define a lot
of elements in a common way and we're able to leverage that that gets into
this data harmonization issue that I. Mentioned earlier and we also work with
health plan to create custom fields in a EHR to capture
certain process data so when somebody called you add live callbacks at the
automatic callbacks didn't work and we wanted to know what was the nature of
this calls did they change your medications and the increase your
medications what actually happened there and so we got work with the pharmacy
department to say when you do these calls we'd like you to flag what you did
and put this in the system in a way with labels the weekend that's fine they
captured later on so they were willing to work with us to do that some
challenges is it any multiple site multi-site study managing is like the
site interaction fidelity is always a challenge but there is also the issue of
integrating information from multiple complex data sources as part of the
ongoing intervention delivery and outcome assessment student and so there
we had we were losing bbw data the virtual data warehouse we also had to be
shipping data to this calling center outside of Boston managed in that case
we were getting some fields from the medical record that were part of the VDW
and how we could pull that and integrate all these pieces and shift things around
from site to site so that was a policy challenge there colorectal cancer
screening Gloria Coronado is the PI.
In this study it's a cluster randomized
trial to improve screening for colorectal cancer it involves over
41,000 patients from 26 federally qualified health clinics though this is not Kaiser two study arms you get a usual care arm and an active intervention they
were mailed fit pits basically fecal occult blood tests and asked to complete those
and send them back in the notion is if I. Proactively send you these fit kits why
increase colorectal cancer screening as opposed
to just letting it happen passively in the clinics the outcomes are primarily
whether returns fit dip but we also looked at the re-aim as what Glasgow
concept for looking at a broader study implementation reach adoption
implementation of the intervention etc some of the TCH are elements we created
tools that were specific to old occupational screening within it that
can be these included a real-time updated registry of patients who are in
need of screening and a process for bulk ordering fit kits and for batch
communications the patients these are technologies that could be done through
epic but weren't being done at the time so we use the the features of every to
build these tools we created unique reports to assist the clinic's with
scrubbing their EHR data and identifying care gaps so we knew that there was some
information on external users creating services that wasn't necessarily being
fully captured and we work with them to try to improve that and catch up on that
as the intervention was totally embedded within the EHR so the challenge is we
have a lack of a defined population I. Mentioned this earlier as being really
important things we had to rely on a somewhat arbitrary set of rules to
define what be what is an active patient we said if you were seeing any campus in
the last 12 months we're going to treat you as an active patient we don't know
that that's true necessarily but it's the operation of old we had to go with
also because the intervention was totally embedded in the EHR it made some
traditional tracking that you would do in a research study a bit more
challenging we couldn't mint monitor at a day-to-day basis what was happening
and we had to rely on monthly snapshots pulled from the EHR to generate our
monitoring of reports pain management Linda Barr study
feedback another cluster randomized trial this time to improve pain
management it involves 851 patients with chronic pain were receiving long-term
opioids this is across three kaiser regions to
study arms a usual care arm and a multidisciplinary integrated pain
management on is embedded in the primary care setting and our primary outcome was
suffer courted pain severity and secondary limbs include opioid use and
false stimulated cost in the wall post in terms of TC
or elements we had a tear cautious for getting a Pierrot data patient reported
outcome pain data all the regions were trying to get pain data but they didn't
all use the same measure and they weren't all consistent about getting it
on a regular basis so we wanted to get that as much as possible so we built
systems into the medical record to try to do this so we had we have an online
patient portal if I can use it they have internet access so our first step was
every quarter you get an email reminder saying please go online use this length
to go complete your pain survey if they didn't do that then they got an
automated call your Kaiser full service to ask these questions and that was
captured in the EMR harness that didn't work then we had a live person call them
up and enter the information into the EMR so most of it was automated but we
did have a live backup at the end we also had national build so Tyler again
is mostly in different regions we were looking at three of them but they
decided that we're going to build something for three amazing builder for
everybody so they build national builds to add the spore item subsets of the
grief pain inventory our main outcome to the EHR they also had a national bill to
addicts and emulator disability questionnaire to the EHR and we used
some features of epic that allow for the attachment of images to the turn notes
so that providers could have access to detailed information summary information
for the intervention patient so the integration began by bringing we had
group visits a bunch of patients and this multidisciplinary provider team
that did an initial intake look at how you were doing and talking about pain
management strategies and the notes from this make sure the Doc's had be where we
able to scan them into the church and so they can see that and we finally added
tact itself as a flag that you were in this study and in the ER - without
greater visibility attacked visits in the truck so some challenges many of the
needed fields and variables that we wanted to use for the study work part of
the existing virtual data warehouse that I'd
mentioned before and they weren't always coated in a standard manner across the
region so we can't have to do this data harmonization for those one region
underline a substantial change in your epic implementation in the middle of the
study and we had to accommodate this and through our data systems and also I've
add there's many good tools that are available in epic and they wouldn't
facilitated the intervention delivery or outcome assessment but it weren't built
into the versions of epic that are we're using in some sense we have a penalty
for being trillion or puzzle but very early adopters for Kaiser of the epic
system and and there's a certain inertia so when a new version that comes out
there's a certain amount of work and bother costing Baga was updating to that
versus you had to stay with your older version and so we tended to have the
older versions and people who came on later to epic tended to get the current
versions with all the bells and whistles in it so there were things that we could
have used have been nice to have that we actually didn't necessarily have access
to medication safety at David Smith's a colleague of mine they didn't interrupt
the time series to evaluate the impact of computerized provider order entry
with clinic decision support in reducing the use of potentially contraindicated
agents medications and elderly individuals so this was Kaiser members
in Northwest 65 and over the intervention was decision support that
alerted collisions to preferred alternative medications when what they
were ordering were non divert ages that carried a potential contraindication for
the elderly now we didn't say you have to use this other medication but we left
it up to the provider but we gave them the alert that there may be a
contraindication to this for your patient and please consider it the
outcome is the rate of the used for both preferred and non-preferred drugs so the
alerts were fully integrated into the EHR they were presented whenever any non preferred drugs was perscribed just regardless of the age of the patient so
we looked at both older and younger patients to see the impact
and these levers existing computerized provider order entry functionality so
what that means is right now you no longer write a paper script you give it
to the patient you you type in the prescription the order into the EMR and
so that fact enabled us to top up the alert at that moment so we've gotten
away from paper prescriptions and they're now all electronically ordered
and that's what we leveraged to get the other some challenges there was a bit of
a lack of functionality in the EHR to do truly real-time in the moment alerts but
we what we did is every night or what they did is every night they refresh the
system so that it was at least current as in the last 24 hours
there was an unwillingness on the part of the clinic to randomize patients
develop this is a good thing to do and everybody should get it visually which
is what led to the interrupted time series analytic design and this is not
an uncommon situation where somebody says we want to do this is a good thing
to do but we want to evaluate its impact well this is an analytic tool strategy
that gives you an opportunity for looking at that the stepped wedge design
which I gather is coming up is the next topic on this seminar series is it
they further wrinkle on that at ninety-degree my beam but it's a real
issue was finding qualified epic programmer so health plan has plenty of
people who work on their epic programming but they're typically kept
busy full-time and in any given moment they've got a list of about 20
priorities longer to-do lists that they have to get to so if you come along with
something else you want to do while how's that going to fit in they've got
everything else because you gotta find somebody who's qualified your work on
that said who the system is willing to let come in and tinker with this just a
minute what does this let anybody come in and start messing around with their
EMR so that it seems mundane and routine but it's in fact a real challenge to be
thinking about in doing these kinds of studies and there was a concern about
alert fatigue I mentioned earlier title is really trying to be very aggressive
and using the EHR to promote prevention and and so you had to find an
apartment in this case the pharmacy department is a champion in the
intervention otherwise doesn't let me go anywhere because there's just there's
just too much going on and so that's that's an issue this is I think the
final study I'm talking about here vets and service use among patients with and
without serious mental illness this is a currently recently submitted paper again
by another college at the Center for retrospective cohort study involving
over eight hundred thousand adults served by tribes of Northwest or number
of community health clinics and one Dane it is dated put people into one of six
category Tuesday one of five adult categories disorders or you were a
reference group which had none of those that was the exposure and they looked at
used to provider services over time and the extent to which the ability varied
according to one's mental health categorization and the analysis you can
use you that as an existing in the EHR. Adjusted for age and gender
race/ethnicity medicaid medicare status and the
comorbidity index so these are all again flavors of things that you can do I
mentioned previously I just have two more slides the HMO research
collaboratory this is a big initiative at NIH to strengthen the national
capacity to implement cost-effective large-scale research studies that engage
the health care delivery organizations as research partners provide the
framework of implementation methods and best practices to the studies that I
mentioned were examples of collaborative studies and and they really develop a
wonderful set of resources for you there's a weekly Grand Rounds and those
are cataloged and you can look at them online they happen every Friday if at
ten o'clock nighttime I think it is and if I specifically one o'clock Eastern
there's a living text book and an ultra pilot or there's a link here and it's
also in other supplemental materials and recording that is the National patient
centered clinical research network this is part of the patient-centered outcomes
Research Institute or the quarry that is funded as part of the Affordable Care
Act it's designed to make it faster and easier and less
to conduct clinical shirts by couraging the power large amounts of health
plantation partnerships now all the work done by the court ad and all the work
done by the HMO research collaboratory won't all be prevention based
necessarily but it nonetheless these are really big players and movers and
shakers now and and using the EHR to facilitate research and much of which
will be prevention related and so I've encouraged you to to look into these
that's my comments the Damned again the ladies I've had a chance to present to
you and I think we are open for questions at this point I thank you dr.
Koh we're very much I enjoyed your presentation and I know our audience has
and we've been monitoring questions as they've been coming in a number of
people are interested in the issue of how to access how to get access to
electronic health record data and do research using those kinds of data you
happen to be in an excellent position of working for Kaiser Permanente one of the
large healthcare systems so you have automatic entree to the people the data
that the whole system what about somebody who's at a State University is
not part of Kaiser or guys injure or one of the other major health plans but is
interested in doing the study in that context what advice would you give them
for how to get started so it's a great question what you want to do is okay I
forget requests all the time people wanting to use our data and we're very
open to that but we don't want to just give it away so the idea is to form
partnerships to find somebody to partner with and you don't just have Kaiser that
kind of part of a very broad network called the healthcare systems research
Network formerly the HMO RN that all have summer these are large managed care
a network integrated care networks that have electronic databases and are
doing research and go to their website check that out look for somebody who's
perhaps got an interest area that dovetails with you go to the website of
an individual organization and find somebody somebody I get somebody with
email me a question and I'll say well this is not something that I do but I
will shop us around to mic to my colleagues so the key here though is to
it to try to say with what's not going to work to say give me some that so I
can go do whatever I'm going to do but rather say can i partner with you
they're very open to the notion of partnership they just don't want to give
the data away wholesale and that the cornet isn't in it this is meant to be a
large national resource so that I think anybody can be applying to work with
them one that I'm not positive on the details of that because I'm not closely
involved with it but again I think that's part of the purpose there is to
create these national networks that are going to facilitate research but the key
to it generally is partnering don't say I want it for myself I'm going to say
can i partner with you and that's what's going to open the door thank you
another important question for our viewers is how such studies get funded
that's a common question is that we have in our webinars so if someone has
they're at the State University they've reached out and formed a partnership
with someone at Kaiser or at one of the other organizations they've got a
research idea how do they get financial support to do the work do these health
care systems provide resources for that or how do these folks proceeding it's a
mix so I would say that where I am the Center for Health Research 80 plus
percent of our funding is federally funded either NIH or the agency for
Healthcare Research and quality we do some mccrory work
does fund the have some mechanisms for funding work internally it's pretty
competitive and it's not there's not buckets loads of money but there is some
increasingly my sense is that NIH is looking for health systems to to partner
certainly the quarry is very explicit about this that they're expecting health
systems to share part of the cost of doing this research so there's no sort
of single place that's the the ideal venue for this again most of why then
myself over the years is that NIH fund that I'd be interested if anybody else
out there in the audience or view or others have have thoughts on this as
well certainly people from some from the federal perspective feel free to chime
in on all on your own thoughts on this issue and we have another question and I
just want to point out that we have a lot of questions so if we don't get to
your question and we'll definitely reach out to dr.
Vollmer and see if you can
answer some of them and we'll post them on our website and so we've done so this
person has we've done some work with adding flags and content into EHR and
have got a good amount of resistance from practices particularly with adding
more clicks to epic did you have similar issues and how were they addressed I'm
not sure I understand the question can you read it one more time sorry I think
the question is asking how hard is it to program the prompts and the intervention
methods and the extra measures and other kinds of things that you want to do as
part of your research into epic or whatever EHR system you're using because
that's right yeah that's that's that's a great question I don't personally I can
answer that one do the program much myself one thing I have been impressed
with over the years is that think expressing it being easy aren't always
easy and something that I would have thought were hard to do wind up being
easy I do know that that is that some things you try to do a lot harder than
others the the one big message and it's not
directly a programming message but but what I hear from the docs will be the
pushback we get all the time is whatever you do needs to be to act as little the
doc as possible we're talking about doc sue and Kaiser we have a 15-minute slot
for 20 minutes lock or primary care appointment and and that's like 15
minutes and direct contact and five minutes write up your notes and and
they're looking to take five seconds off of their plates during that visit rather
than adding five seconds to it so the more clicks and the more harder you make
into fine things the more challenging it is the interventions that work best
would be things I think that are like pop ups I don't have to go click on
something else to get it but when I do something I'm going to do anyway I got a
top like this papa alert says by the way this medication may be contraindicated
for these reasons so the extent to which you can provide the providers with
information in the moment and that happens spontaneously a response to
something they were going to do anyway it's going to be most effective I know
that's possible to do but I don't have a feel for how difficult it is
computationally to do that looks long person answer that awesome well but you
did an excellent job and the person that ents the asset wrote back to us and said
you're answering the question exactly so okay again finger cancer I I'm afraid
we're running out of time unfortunately we're approaching the top of the hour
this has been a terrific presentation we have had lots of questions dr. Bober we
will be reaching out to you to send you the questions that we have so that you
can send us answers we will post those on our website as soon as we're able to
and I encourage people that have asked questions and and we weren't able to
raise them look at the website and find answers later the National Institutes of
Health are making a substantial investment in EHR based research dr.
Boomers described two of the studies that are part of the healthcare systems
laboratory the CRC stop he packed or both supported by the
collaboratory the expectation that NIH. Has is that EHR systems will make it
easier and far less expensive to do large clinical trials because much of
the data is being collected already you've got a system built for delivering
intervention many of the features that you described are things that NIH thinks
we ought to take advantage of so we are putting more and more resources into it
including a major new effort that literally started recruiting patients
last week the all of us research program which is the covert component of the
precision medicine initiative that will launch in a very large way over the
course of the summer and early fall will go national and that's that whole
program is reliant on the electronic health record shirts made major data
collection activity yeah so and I want to say thank you again for you know
coming out of retirement to do this for us really appreciate it
and good I like I saw level so the comment day today III think that I think
that it's inevitably through that the each are good facilitate cheaper per
patient studies regard final trial the critical perforation cost becomes
expensive but they tend to involve thousands of patients and they can be
collectively very expensive to do Asil and holding is all think one of the
learnings coming out with cloud Lori that everybody was finding is that this
is hard work I mean working with health systems to change how they do things is
is a really important word so there's not going to be a silver bullet just
because we have that we can now do these big great trials on the Chiefs what okay
but it puts it on one challenge yes yeah point well-taken okay so thanks again
and we're running out of time so we have to close and thank you to everyone who
participated in today's webinar on the Medicine: Mind the Gap website at prevention.Nih.Gov/MindtheGap you will find
several resources for this talk including the slides references and a
link to complete and evaluation your feedback is very important to us as we
plan future sessions thank you again for your time.
No comments:
Post a Comment