5. Agenda
1) Social Determinants of Health (SDH)
- What are they?
- Why are they important?
2) External SDH data sets available
- What are examples?
- What’s the definition and importance of Precision?
3) Making SDH information the 7th Vital Sign
- What are the challenges?
- Why is that challenging?
4) A Road Map to actionable SDH information
- What do you need?
- How do you get it?
7. Agenda
1) Social Determinants of Health (SDH)
- What are they?
- Why are they important?
2) External SDH data sets available
- What are examples?
- What’s the definition and importance of Precision?
3) Making SDH information the 7th Vital Sign
- What are the challenges?
- Why is that challenging?
4) A Road Map to actionable SDH information
- What do you need?
- How do you get it?
11. Precision
State / Region
County / MSA
ZIP Code
U.S. Census Tract
U.S. Census Block Group
Individual Household
Individual Person
LevelsofSDHPrecision
High
Low
1
2
3
4
5
6
7
17. Precision
State / Region
County / MSA
ZIP Code
U.S. Census Tract
U.S. Census Block Group
Individual Household
Individual Person
MultipleLevelsofSDHPrecision
High
Low
1
2
3
4
5
6
7
18. Precision
State / Region
County / MSA
ZIP Code
U.S. Census Tract
U.S. Census Block Group
Individual Household
Individual Person
LevelsofSDHPrecision
High
Low
1
2
3
4
5
6
7
19. Agenda
1) Social Determinants of Health (SDH)
- What are they?
- Why are they important?
2) External SDH data sets available
- What are examples?
- What’s the definition and importance of Precision?
3) Making SDH information the 7th Vital Sign
- What are the challenges?
- Why is that challenging?
4) A Road Map to actionable SDH information
- What do you need?
- How do you get it?
21. The costs of adding social and
behavioral domains to EHRs, such
as programming, modifying
workflows, and intervening on
positive screens often fall on the
individual health practice or
hospital. The movement toward
population health management
and ACOs may address this
malalignment over time. In the
meantime, costs remain a barrier.
23. Agenda
1) Social Determinants of Health (SDH)
- What are they?
- Why are they important?
2) External SDH data sets available
- What are examples?
- What’s the definition and importance of Precision?
3) Making SDH information the 7th Vital Sign
- What does this mean?
- Why is this important?
4) A Road Map to actionable SDH information
- What do you need?
- How do you get it?
29. Patient Specific Recommendations
28
Who
else?
The patient may need transportation
assistance to the hospital or pharmacy.
The patient is experiencing financial
difficulty and may not be able to
afford prescriptions or follow-up care.
The patient is seeing many providers
and may have issues with care
coordination.
The patient is at high risk for
medication non-adherence and may
need counseling or follow-up.
The patient is experiencing financial
difficulty and may not be able to
afford prescriptions or follow-up care.
30.
31. Dr. Brian P. Goldstein
EVP and COO of University of North Carolina (UNC) Hospitals
Member of Forecast Health’s Board of Directors
Forecast Health’s predictive models are highly accurate
and identifies patient-specific risk factors. This means
that if a patient’s risk arises from affordability issues,
we’d know to provide extra medications from the
pharmacy stockroom. With this, Forecast Health helps
to provide higher quality care.
34. NEXT STEPS
Contact info:
Sohayla Pruitt, M.A.
Sohayla.Pruitt@forecasthealth.com
(919) 699-7424
Michael Cousins, Ph.D.
mcousins@forecasthealth.com
(804) 380-8616
Thank you!
www.forecasthealth.com
Editor's Notes
Maps have been used for many decades to explore linkages between health and social phenomenon.
Here, dark blue indicates areas with poor health outcomes; and when compared to dropout rates in yellow, it’s easy to visualize a potential linkage.
<toggle twice>
While a map is worth a thousand words, I have seen that it can sometimes be worth a thousand questions.
If you show this to a clinician or care manager
Who is trying to intervene on a patient’s health,
they will often ask: “what do I do with this?”
The goal of today’s webinar is help you discover:
Why we need to better understand the social determinants of health , but also
Why we must push our thinking beyond the context of basic map visualizations toward capturing this information discretely within the EMR--similar to the way we capture a patient’s vital signs.
In this presentation, you will notice a strong focus towards clinicians and care teams being the end user of this information,
so that appropriate interventions can be delivered at the point of care.
This is because we believe it is at the point of care where
the greatest potential exists to improve outcomes and reduce costs.
we will begin by providing an overview of the social determinants of health, and then
we will provide examples of SDH data sets available from both public and commercial sources at varying degrees of precision.
We will also discuss the importance of integrating SDH information within the EMR along with the challenges, and finally,
we will explain the typical road map for SDH use in a health care setting, as well as what we recommend in order to decrease the time it takes to realize value.
IOM, WHO, and others estimate that clinical care may only contribute 10% to the health of a patient, and there are a number of other drivers that must be better understood if we are to truly have an impact on the health of our populations.
Health systems are well underway in capturing clinical information within the EMR, and billions of dollars are being invested in the understanding of genes and biology, but those together may only contribute 20% to the health of a population!
What about the 3 categories that may contribute 80% to the health of a patient?!?!
If a health system wishes to truly change patient behavior, engagement, utilization, outcomes, and cost, they will need to better understand the impact of each patient’s physical environment, social and economic factors, as well as their health behaviors (these are the 3 we will focus on today).
When a patient’s address is geocoded, their specific location can be related to socio-geographic information; such as:
their distance to other features in their physical environment,
or whether they live in a geographic area that might have high pollen counts or be considered a food desert.
With regard to social and economic factors, there is a wealth of information published by the U.S. Census at various levels of geography; such as education attainment and income.
Specific person and household level information can also be obtained from commercial sources, which can help understand things like: a patient’s ability to pay or car ownership.
Similarly, there are a number of person and neighborhood-level data sources to help us better understand health behaviors,
which can be derived from consumer spending and lifestyle segmentation data sources, but
can also include device and mobile health data.
For all of these sources of SDH information, it’s extremely important to understand the differences in precision.
SDH information is often available at all of these levels, but the lower levels of precision may be misleading and simply inactionable in a healthcare setting.
Let’s take a look at some examples.
As you recall from the SDH pie chart, social and economic factors account for 40%.
Some may examine median household income as a measure of affordability and may want to understand whether or not affordability has an effect on health outcomes or utilization.
While Median household income is available at various levels of precision, here it is displayed at the ZIP code level.
For the same ZIP code area, here is how median household gets broken up by block group
<toggle back and forth twice>
Now, if we examine the highlighted block group with a median household income of: 162,715, and compare against what the ZIP code said the median hh income was <toggle forward>
.
You can see just how imprecise zip level data can be.
Similarly, you can imagine that individual people living in high income neighborhoods may experience financial difficulty from time to time, which
may not be captured well enough without looking at person-specific information.
It is quite possible that individuals could have a number of person level factors that may affect their ability to afford visits or expensive medicine.
This isn’t surprising since a zip code can have tens of thousands residents, and a block group between 500 and 1000.
The differences in SDH precision levels can be sizeable.
For this, in order to have an impact on health utilization, outcomes, and cost, we recommend
only the highest level of precision available, with the lowest level being the block group neighborhood.
There is a tendency in the marketplace today to offer SDH at these lower levels of precision; such as ZIP Codes.
For a community planner, ZIP Codes may be precise enough, but as we have shown here, they may be inactionable within the context of the EMR or
could result in misinformation to the providers and caregivers who are trying to make an impact on an individual patient’s health.
again, our focus is on clinicians and care managers, along with their ability to take action in a care setting.
Can we really integrate SDH information within the EMR cost effectively?
Let’s take a look at the challenges.
In a recent IOM report, the importance of capturing SDH information within the EMR is emphasized, but they acknowledge the potential for significant cost and challenges to a health system.
As a result, they recommended a set of SDH variables to be captured via patient intake questioning.
Health behaviors related to smoking and alcohol were some of the recommendations.
While this is a good first step, I remember when I was working with the transitions in care team at Duke Medicine, and …. (tell story)
So why might it be considered difficult to enrich the clinical information with SDH information gleaned from external sources?
lets take a look at the methodology required if a health system were to try to make this information the 7th vital sign in their EMR.
Methods needed
Data External to EMR must be acquired (so many sources, yet little is known about which are needed, whether they are affordable, or precise enough to drive ROI).
Automated extract transform and load processes must be built for each one.
Integration within the EMR may take several automated processes, which rely on a number of FTEs, specialized tools, and advanced methods (which in the case of median household income or distance to nearest fast food, would mean….).
And once this is complete, what will a health system do with it to drive value? For a clinician or care manager to appropriately intervene on any given patient at the point of care, the information has to be actionable.
Let’s discuss what we think is needed in order to make this information actionable.
We’ve developed this SDH maturity quadrant to help health systems better understand where they are, where they wish to go, and how they get there.
For this quadrant, lower levels of SDH precision prevent the information from being actionable. Another driver of actionability lie within the sophistication of the methods used to glean insight from the data and translate that into clinical intervention.
We typically see that MOST health systems who attempt to understand SDH fall within this lower quadrant; they have acquired SDH data (often open-source at lower levels of precision), stood up some basic visualization tools in order to better understand WHERE a given phenomenon is occurring, but are having difficulty in taking action on this information at the point of care. We also find this is where most health systems are with regard to their clinical information (they have clinical data, they have BI tools, and they are beginning to understand patterns and trends, but haven’t translated the information into action).
However, if you are a clinician or care manager, the only way for ROI to be realized is for your solution to be positioned within the upper right quadrant.
Let’s take a look at SDH capabilities within each of these quadrants.
In this lower left quadrant, you might see a use case that help visualize WHERE a phenomenon is happening in relationship to other SDH-related factors.
In this case, the ZIP codes with the highest number of flu cases are shown in blue, overlaid by the location of flu vaccine clinics. It can be powerful to visualize this information outside the confines of rows and columns in order to discover patterns, but can be difficult for a clinician or care team to translate the information into prescriptive action or identify specific patients at risk for not getting their flu shot.
For a community planner, however, it might be useful to take the information gleaned from the prior map and identify NEW areas with which to target neighborhood or community level interventions.
For example, the previous map lead us to believe that ZIP codes with a high number of flu cases, which do not have a number of flu clinics in close proximity, may be in need of mobile flu trucks that can deliver immunizations to those ZIPcodes in need (such conditions could be identified in new geographies using spatial queries to identify WHERE ELSE such needs exist.).
For a clinician, they may still have trouble taking action on this information….
would a clinician feel confident that a patient coming from one of these ZIP codes in need
will be at risk for getting the flu because they live far away from a flu clinic?
This is why it is beneficial for solutions to be tailored at higher levels of precision; such as the person, household, or block group neighborhood.
With more precise SDH data, a health system might be able to better examine descriptive statistics about their patient population at a person-, household-, or neighborhood level. With this capability, they will be able to visualize where their patients live, as well as understand some of the patient specific information available on that patient population.
Using the above descriptive dashboard, a provider may be able to quickly see that their patients who inappropriately use the ED, primarily have lower household incomes. This may spur the development of programs to target lower income patients in order to reduce their ED utilization, but again….without methods to assist with actionability and real-time idenification, a clinician cannot use this information at the point of care. They still need a way of identifying patients likely to become a super-utlizer, as well as the factors that drive each patient’s individual risk.
When SDH information is properly integrated within the EMR with high precision,
it means that advanced algorithms can be used to model thousands of both clinical AND SDH data variables contained within the EMR, so that
providers are able to understand the WHO, WHERE, and WHY…as well as identify WHO ELSE is at risk and
what interventions are likely to be best for each at risk patient.
The results: targeted, person-specific interventions to decrease utilization, and improve outcomes and costs.
(discuss slide: patients stratified by risk for readmission, and patient specific responses given—both clinical and sdh-based recommendations (gray vs. green)
We have seen great results in capturing precise SDH information alongside the clinical information and using advanced modeling techniques to inform real time decision support.
For example, we compared the ability of two predictive algorithms to predict patients at highest risk of 30-day readmissions.
One of the models is called LACE – it is a simple algorithm based on four clinical data elements. No SDH. It’s also a simple rules-based algorithm that Epic has built into their platform.
We compared the predictive accuracy of LACE to our predictive model that used both clinical and SDH data.
We found that the predictive accuracy – what’s called the “true positive rate” or “sensitivity” -- of the LACE model was 44%. Ours was 84%.
This is a 91% difference -- almost a 2-fold increase.
Putting these stats in terms of patients -- this means that we’re finding 84 true positive high risk patients per 100 high risk patients, whereas w/o our SDH model we’d only find 44 out of 100.
We’ve also benchmarked our SDH predictive models against other benchmarks including BOOST and the PARR model from NYU and we had similar outperformance.
If you’re interested details on these comparisons are available on our Website in a White Paper under the Blog Section.
This quote from Dr. Brian Goldstein, a practicing physician and COO of UNC Hospitals, nicely sums up the past 2 slides.
In summary, it might be helpful to think of this maturity quadrant as a road map.
Be careful though, because health systems often think they must start at the lower left of this quadrant and build up as they go.
Unfortunately, when done this way, the data collected and methods used often do not scale to support end goal.
They end up spending more time and money in the long run for specialized FTEs, software, and hardware, with limited ability to better the health of their population.
We have found that when we build a road map for a health system with the end in mind,
we are capable of providing all other capabilities covered in each of the other quadrants, but
in a streamlined fashion where a health system can more quickly realize full benefits and ROI.
What this means is we have developed methodologies, which automate all of what is shown in the purple box; including the:
Identification of SDH information relevant to the health care phenomenon being examined,
The keeping of all external information up to date,
The enrichment of the clinical data with SDH information (stored discretely within the EMR as if it were a 7th vital sign),
The advanced analysis of all of the clinical and SDH information together, and
The serving of solutions which integrate back into the clinicians/care manager’s workflow along with the workflow of planners and marketers.
We have found that by eliminating barriers to entry for a health system, they can more quickly and cost effectively improve outcomes and reduce costs.
If you are interested in more details on SDH please visit our Website and look for the SDH White Paper under the Blog Section.