(bright upbeat music)
Welcome, my name is Lou Brooks.
I'm the Vice President
of Commercial Analytics here at Optum
and today I'm going to talk to you
about creating a clearer picture of the patient experience
through the integration of claims
and electronic health record data.
We're all well aware that the health ecosystem continues
to evolve as we collectively try to manage
the cost of healthcare,
develop new compounds to treat diseases
and get consumers to take a promo more proactive role
in managing their health.
We continue to drive innovation in the areas
of drug development, health technology, surgery,
care models, reimbursement and analytics,
but much of this innovation is based upon
existing data strata utilizing either claims
or electronic health record data in isolation.
We have the opportunity today
to further accelerate our understanding of healthcare
and drive noble changes
to many of our fundamental constructs
by altering the underlying data foundation
to many of our processes.
It starts quite simply by integrating claims
and electronic health record data at scale
to provide a unique data foundation
that enables us to see a more comprehensive view
of patient care, outcomes and costs,
become more tailored in addressing healthcare questions
in a more cost effective manner
with a single source of data,
and improve on our existing models and methods
to shift our focus to joint clinical and cost outcomes,
to identify those treatments and care pathways
that lead to the best clinical outcomes
for our patients at the lowest possible cost.
How does it all begin?
When we think about
the data itself,
from that standpoint,
on the next slide,
you'll see that we have a wide range
of data potentially available to us.
As we continue to work to transform healthcare,
a perfect example of that transformation
is machine learning models
and identifying patients that may be at risk
for a certain condition
and proactively intervening with those patients
to positively alter their health trajectory.
However, even the best of
those models miss a fairly large portion
of the appropriate patients and falsely identify others,
leading to some level of healthcare waste
and missed opportunities
to positively engage with patients on their health.
Why is that?
Many of the models today are based solely on claims
or solely on clinical data.
It's like using unleaded gasoline
when premium is needed in a sports car.
It runs, gets the job done,
but it isn't firing on all cylinders optimally,
so you aren't quite getting peak performance.
Integrating gait as the premium gasoline
for that sports car,
and it doesn't just stop at integrating claims
and electronic health record data.
As this graphic shows,
there's a wide range of additional data available
to bring into healthcare research.
The ultimate goal for healthcare researchers
is to obtain that comprehensive view of the patient,
including healthcare data, attitudinal information,
consumer profiling and purchase data
and leveraging all of those elements at once
to truly understand health outcomes and costs.
We aren't there yet today,
but the integration of claims
and EHR data is a big step in that direction.
Imagine for a moment how
integrated claims electronic health record data
could change the way we engage
in existing analysis across
the entire spectrum of healthcare.
Let's start by imagining being able
to fully assess a clinical trial protocol
for inclusion and exclusion criteria,
and then be able to drive
that directly back to site selection.
How many times have you used claims data
to engage in that process
only to be unable to identify the very specific criteria
that's essential to your protocol,
or be really targeted
at your protocol assessment using clinical data,
but be unable to identify the best sites
for potential trials.
The integration of data, claims
and electronic healthcare record data,
has the potential to support the entire work stream
with one set of data.
Let's move forward a bit and talk
about comparative-effectiveness studies.
Most of that work today is done with claims,
data and various methods analyzing either total cost of care
or the cost of the disease of interest.
How much better could those models be
if we were able to integrate clinical information
in case matching methods
and stratify the analysis by disease severity
or other clinical metrics of interest
that could provide additional illumination
and insights on product performance.
Integrated data also has the potential
to change the way we evaluate product performance over time.
Imagine for a moment,
being able to stratify patients being put on your drug
by clinical severity, measure with lab results,
and then tracking the clinical performance
and cost of those patients over time.
Imagine that information being available at your fingertips
when you sit down with a payer
to discuss the total cost of care
of the patients on your drug compared to competitors.
How could that change that discussion?
Both clinical and claims data
have their own individual
strengths and weaknesses.
Claims data is missing critical clinical data elements.
Things like lab results, observations,
notes information where a doctor
has recorded specific pieces of information on care
that have occurred during that particular instance.
Those clinical insights could provide a better understanding
of what's going on from a care management standpoint.
On the electronic health record front,
we're missing claims,
data information such as costs, eligibility,
and filled prescriptions.
As this graphic illustrates in that gray box in the middle,
we've got a great deal of data in common to both,
but the integration of the data allows us
to gain access to all the information.
As a result of that,
we fill in missing pieces of information
from either of the individual unique sources,
resulting in a new data foundation for healthcare analytics.
Let's look at an example.
Very simple, very straightforward,
but it gives you some insight as to what happens
when you can integrate claims
and electronic health record data together
from an analytical perspective.
Let's imagine that we have a patient
who has visited the emergency room.
With claims data,
we know that they had a claim for an outpatient visit,
they went to the emergency room,
we saw that it was billed related to diabetes
and the cost was about $7,000.
With EHR data,
we can easily confirm the visit and the diagnosis
because those data elements are in common
to both sources of data.
In EHR data however,
provides a wider range of additional data on the patient
that would be unseen
in just a straight-claims-based-analysis.
Including basic observational data such as height,
weight, blood pressure,
as well as extremely high A1C level,
and what medications were prescribed to the patient.
This additional information only scratches the surface
of what's available
from an electronic health record perspective,
including things like symptomology,
patient reported medications
and other health data
that will provide context
around the specific interaction
and the specific treatment decisions.
Let's go back to claims for a moment.
With that integration of data,
we're now able to connect that fill information
from the pharmacy,
with the electronic health record information
and see that the patient actually followed through
with the three written prescriptions
from that emergency room visit
and filled those three prescriptions.
And this journey can continue from there,
and you can demonstrate a wide range of impacts
from a treatment decision standpoint with providers.
We'll talk a little bit more about that in a moment.
The integration of data does come with its challenges
and its limits.
Perhaps the biggest three challenges are first, sample size.
The more disparate data sources we integrate together,
the smaller the sample of data that you have to work with.
So it's essential to work with the largest sets
of individual claims and electronic health record data,
to give you that largest intersection you can
from an analysis standpoint.
Completeness can also be difficult.
On the claim side,
many claims sources are eligibility controlled,
so you know what you're missing.
But as you start to integrate various sources
of electronic health record data,
you may or may not be missing particular pieces of data.
The same can hold true with claims sources,
depending on whether or not they're open or closed sources.
The final challenge is really around resource competency.
You need to rethink the way that you model
and you analyze data once you start integrating it.
The old claims based algorithms don't hold true perfectly
and that research that you've been doing
from a clinical perspective, isn't perfect either.
You have to rethink the way
that you are working with the data
and your researchers need to be retrained.
We must also remember
that privacy is a fundamental component
and a responsibility of all of us
as we're working with these integrated data source.
While it doesn't stop the integration of data,
it does complicate it
and it complicates what you can do with it.
You need to make sure
that we're working to look at that data
and make sure that we are hyper compliant as we work
through all of this integration standpoint.
Because, while it limits what we can do,
it doesn't prevent outright analytics
with that integrated data source.
So let's change gears now.
We've talking theoretical for the last 10 or 15 minutes,
let's move into some actual use cases
to show you what we can do with integrated data.
I've got a few examples of how
that integrated data can be utilized,
translated into analytic value
and change how we generate insights.
The first example is perhaps the easiest.
It's the "low- hanging fruit".
And we've touched briefly upon it
with the emergency room example previously.
The integrated data provides a one stop shop
for truly understanding the journey
of a patient to get treatment from that interaction
and written prescription in the office,
through the subsequent fill and refills
of those prescriptions over time.
In this example,
we get to see a piece of information
that isn't normally available in a claims analysis.
That information is
the actual written prescription or order.
The physician's intended treatment for that patient.
Once we have that information,
we can move to claims.
Claims now allows us to cycle
through the administrative process
of filling the prescription
from the point of presentation to the pharmacy,
you through the utilization management programs
that might exist
and the fill and subsequent pickup by the patient.
The claims data also allows us
to track subsequent prescriptions and adherence,
and if treatment changes do occur,
the electronic health record data gives us
those reasons for change
such as the example in the lower right hand of your screen.
So as we can demonstrate with the three product examples,
we can see
not only how many prescriptions were originally written,
how many of those were presented to the pharmacy
and ultimately how many of those got into
the hands of the individual patient.
Having insight along this pathway offers many opportunities
for all of us providers, payers,
and life savings companies alike,
to develop and target programs to maximize
the number of patients
that they're get their prescriptions
and stay on their medications.
Imagine for a moment,
just one simple example.
A provider has access to the integrated data.
They can follow a patient who doesn't present
that written prescription to the pharmacy
and contact them to discuss treatment again,
and depending on the reason why
that patient decided not to present the prescription,
work to overcome that barrier to filling
and getting that patient treated.
Now let's look at a integration of the data,
utilizing both clinical metrics and cost.
We're going to look at the correlation between cost
and clinical outcomes
for type 2 diabetes patients.
We took a very simple approach to this,
identifying type 2 diabetics in 2016
and calculating their baseline,
A1c and BMI levels at the end of 2016.
We then had a year's worth of data for them throughout 2017,
and we tracked all of their healthcare interactions
and expenditures over the course of 2017,
and we deciled them.
Decile one being the most expensive or highest cost segment,
and decile 10 being the lowest.
The 80-20 rule does hold true,
so you'll know that the smaller deciles
are in the higher costs of one, two, three, four, and five,
and the larger segment memberships are
in deciles eight, nine and 10.
We then looked at their clinical metrics at the end of 2017,
to evaluate what the change was in those metrics
and how it related to cost.
We show that there's a positive correlation
between higher spend
and better clinical outcomes moving down
and to the left on the graphic.
But it is far from perfect correlation.
Just opens up a wide range of questions
and potential future analytic opportunities.
Imagine for a moment doing a comparative study
and being able to control for a wider range of confounders
and develop metrics on the most cost
and health outcome-effective therapies.
Imagine for a moment being able
to show a payer
your more cost-effective
and have a greater impact on clinical outcomes.
Imagine then being able to take that information
and build better pathways
to get a more impactful management of population health,
and a lowering of costs simultaneously.
We're going to get a little heavier
into the clinical metrics associated
with integrated data in this next example.
And our goal is to really evaluate
how integrated data can gain better insight
into the impacts of interventions.
Repeated health measures are a integral part
of the integrated data.
And they very succinctly allow us
to evaluate interventions based on
the actual elements of interest.
Let's take the case of bariatric surgery.
In the example in the lower left hand graphic
on your screen,
we have segmented bariatric surgery patients
over a one year period after surgery,
based on their BMI pre
and through out the entire year.
We know that the average cost
of a bariatric surgery runs about $27,000,but in this analysis we found that 8% of those surgeries
and 8% of the patients that received them,
saw no appreciable weight loss in the course of a year.
Imagine for a moment that we could build that
into some type of monitoring
to help alter that performance trajectory
and perhaps even change the performance metrics
to only reimburse physicians when we saw a positive gains
in healthcare outcomes related to the surgical intervention.
Let's take it a step further.
We subset these patients to just type 2 diabetics.
And we examined how they were able to manage
their type 2 diabetes during
that one year period post-bariatric surgery.
In the example on the right hand part of our screen,
we found that 64% of the patients
that were type 2 diabetic
and had bariatric surgery,
also had uncontrolled diabetes
during that one year period.
They also cost on average three times more than
the patients in the lower right hand part of that graphic,
that had the best performance in both weight loss,
and were better able to control their diabetes.
We've found that there were many differences using
the EHR data in healthcare engagement
between those groups as well.
Differences in engagement with providers,
diet, and exercise,
and the integration of the data allowed us
to see all of that.
And allows us now, to set to identify those issues,
evaluate the performance of these interventions,
set up new interventions, policies and protocols
to improve patient care overall.
A fourth used case is more targeted
towards life sciences companies,
and one of the greatest challenges that they face
in evaluating and understanding the investments
and the return that they they've gotten
from those investments in terms of product performance.
These companies spend millions,
hundreds of millions of dollars developing drugs,
and ultimately more marketing them
and bringing them into a space where we can utilize them
to improve the quality of life for our patients.
The problem is that monitoring many
of these markets is difficult because
the data itself resides in different silos.
Let's take the case of "The Immunology Space".
Truly understanding market share on a weekly
or even a monthly basis is difficult because
while many of the products go through the pharmacy channel,
still many others are administered in a physician's office
and paid through through the medical benefit.
Those two claims streams are independent
and they have different time lags associated with them.
Integrating claims and electronic health record data,
allows organizations to gain clarity in their markets
by combining all of those products together
in a single view in near real-time.
As I mentioned,
those medical claims tend to lag in some cases,
days, weeks, and even months,
which means that you can't evaluate properly market share
or product performance until all of
that information comes into play.
EMRs, and in this case EHRs,
are updated on a more routine basis and as a result,
we can integrate all of the data together
to give us all of the medically administered products
on the same case as the pharmacy products
and demonstrate market share in real-time.
Is to also opens up on the next slide,
the ability to track other things with the data as well,
looking at indications,
and why products are being prescribed from that perspective.
This type of market clarity on a weekly basis
can enable better reaction to competitive threats
and maximizing new opportunities for Life Science Companies.
Our last example
is all about tactics.
I spoke earlier about the resources that you need.
You need to be able to change them
as you work with integrated data.
And it's likely the case today
that your organization may not yet be at a point where
you can tactically implement programs, pathways,
or other strategies using integrated data.
It may be that you only have claims data
to leverage for your operations,
or perhaps only clinical data.
If you recall in my earlier in my presentation,
I showed you a graphic of the different types
of data found in EHR and claims,
and I noted a group of variables that are in common to both.
One of the advantages of integrated data is
you can build models using that common subset
to move from the strategic to the tactical,
even if you don't yet have the infrastructure in place
to leverage integrated data at scale for operations.
As we round this out,
I searched for a comment on healthcare data and change,
but was unsuccessful.
So I defaulted to one of my favorite quotes
about Charles Darwin.
And it really is,
is ultimately focused on change
and our ability to survive by being able to adapt to change.
As we are well aware,
as we are developing new models
and new methods to solve problems in healthcare,
integrated data will help fuel that change
and enable us to realize greater efficiency
and insight generation.
Imagine the ability to address value based
on clinical values and costs simultaneously.
Think of how we could develop new value based contracts,
reimbursement models and care pathways
all using one integrated source of data.
It is the future of healthcare analytics.
I hope you've found today's session informative,
and I thank you for your time.
(bright upbeat music)
Integrated data is the driver of change
Integrated data can help us realize greater efficiency, new reimbursement models and more proactive care pathways for better outcomes. Join Lou Brooks, vice president of Commercial Analytics at Optum, as he discusses how integrated data is ushering in the future of health care analytics.