Population Health Management

Population Health Management:
Real Time State of Health
Analysis
YESTERDAY: CLAIMS-BASED PREDICTIVE MODELS
For years, healthcare insurance companies (payers) have mined
claims data for chronic patients and have built predictive models
to identify high-risk patients.

While this approach has seen some success, limitations far
outweigh merits.

Data used by payers to flag high risk patients is historical claims
data — primarily costs, admissions, and diagnoses. Furthermore,
regression and time series risk models are typically updated only
annually.


Most physicians are highly skeptical of claims based predictive
models because they have no clinical basis, and give no
consideration to an individual's current state of health.

Moreover, there is a complete lack of causation, "Why is a
patient considered high-risk? What are the clinical reasons for
the score? How do we lower the patient's risk score? How does
the score measure the effectiveness of my care management
program?“
http://healthcarecostmonitor.thehastingscenter.org/kimberlyswartz/projected-costs-ofchronic- diseases/
http://www.ahrq.gov/research/ria19/expendria.htm
These models lack a correlation to clinical information.

Claims-based risk scores are created with regression analysis at a
population level to predict scores at the patient level.

Not only are today’s calculations unsuitable for determining a
patient’s true risk, they provide no insight on how an individual’s
score improves or deteriorates after each clinical visit.



FURTHER CONSIDERATIONS
Current thinking and efforts create a disproportionate focus on
existing chronic patients.

A better approach is to monitor all patients, healthy and chronic,
for risk of hospitalizations.

Unfortunately, current claims-based predictive risk models allow
no room for this approach.


VITAL PROGRESS
Today, most large physician groups and medical homes already
use at least a basic EHR system.

CMS predicts that by 2014, more than fifty percent of all eligible
medical professionals in the U.S. will use EHR.

This is a transformational shift, because for the first time in
history, clinical information is digitally available in real time, with
reasonable availability of laboratory results and patient vital
data.
CLOSED-LOOP CMP
Using real-time clinical data
from EHR records, health care
providers now have the
capacity to design a closed-
loop population care
management program (Figure
1). A well-designed program
delivers primary care to drive
higher quality, reduce costs,
and deliver greater
value in health care.
Population SOH Stratification
State of health stratification provides actionable and measurable
information about actual health status at the population and
patient levels, with visibility of controllable and non-controllable
factors.

SOH is a “risk predictor”. However, it is also an indicator of the
quality of care delivered.

If the score trends down, the quality of care is good,
because health is improving.
Origins of SOH Models
Nationally accepted clinical models are the basis for state of
health models.

SOH scores are calculated at the patient level and rolled up to a
population level (Figure 2).

In this example, each row corresponds to a physician's patient
population. It shows the patient count, the number of office
visits (encounter) and the average population SOH score for each
chronic disease.

Figure 2 Population SOH (Risk) Stratification by Physician
Chronic Disease Management
Patients who comply with prescribed care programs are typically
more successful in managing chronic conditions.

This is where care coordinators play an important role.

Monitoring gaps in care established by evidence-based care,
patients’ SOH trends, and underlying clinical drivers over time,
care coordinators can identify patients that need their attention.

Care Coordination
Physicians who improved the state of health for their population
(i.e. lower the score) over a one to three year period established
and used better clinical protocols (i.e. best practice care
management programs).

In one instance, one physician’s CHF population risk increased to
55%, while another’s dropped to 5%.
Figure 3 - Effectiveness of two physician CHF populations.
Use best practices within the risk group for evidence-based care
coordination: medicines, treatment levels, frequency of visits; by risk
group.
Population performance: Map patients on quality and total cost across
the continuum-of care (ambulatory and acute). Identify optimal
preventive care levels to minimize lifecycle cost over a time period by
chronic condition.
Incentive management
If financial incentives for health care professionals are not
aligned with performance, success may be temporary and hard
to sustain.

Effective incentive programs distinctly drive higher quality and
reduce costs for greater value in health care.

Incentive programs reward care teams for reducing population
risk scores, improving patient satisfaction scores, and reducing
overall patient costs.

Continuum of care dashboards (ambulatory and acute) are
useful in designing incentive programs and illustrate risk-cost-
quality details for each patient (Figure 5).







Figure 5 - Continuum of Care Analysis by Patient, Preventive Care Impact on Acute Care Costs
Monitor how much total inpatient and outpatient care (cost and quality)
is being provided to the risk panel; identify patient outliers.
Patient SOH scores can be rolled up to population averages.

For example, one incentive program dashboard maps
physician/care coordinator teams on a cost-quality grid.

Each bubble corresponds to a specific physician- care
coordinator team, and the size of the bubble illustrates the size
of the population they manage. The distance of each bubble
from the crosshair indicates the positive or negative variance
from the target and is proportional to each team’s bonus or
penalty.( Refer Fig.6)














Figure 6 – Physician value index used for incentive management for care teams.
Report shared savings by plan by physician on a periodic basis and show
the impact of actions on their “pocketbook”.
Validating the SOH Model APPROACH
To validate the models, researchers compared the new SOH
model against that of a leading claims-based risk model (the
payer model).

For the SOH model, researchers used real-time clinical data. The
SOH model did not include past ER or IP admissions data.

Next, researchers calculated a SOH score for each patient using
historical data over two years
Inpatient Admissions
Figure 7 shows total
hospitalized patients as a ratio
of the total diabetic patients
for that SOH band.

At very high scores, all patients
were hospitalized. Thus, Figure
7 validates the accuracy and
predictive power of the SOH
score.









Figure 7- Ratio of Hospitalized Patients to Total
Diabetic Patients
Creating a SOH Composite
Figure 8 shows the
relationship between
the payer risk scores and IP
admissions.

Similarly, at higher risk scores,
the predictive power of the
payer’s model
is only 50% whereas the
researchers’ SOH model is
closer to 100% accurate

Figure 8 - Relationship between the payer risk
scores and IP admissions.
WORK SMARTER USING SOH MODELS
State of health models are highly accurate and predictive, and
ideally suited for chronic care population management by
chronic condition.
Using SOH scores, care coordinators can correctly identify and
focus on high risk patients with a great risk of hospitalization in
the short term.

Given the rapid adoption of EHRs among primary care physicians
and groups, the data required to build SOH models is readily
available now, and will continue to expand over the next two
years.
Healthcare providers can enable continuous improvement using
SOH models together with care management programs. This
approach has already been institutionalized in a number of
leading medical homes like Medical Clinic of North Texas
(MCNT).
MCNT has pioneered the SOH-based population management
approach.

MCNT experienced a stellar FY 2010 performance with Total
Medical Cost trend.

Overall performance index improved in Facility Outpatient (-5%),
Other Medical Services (-6%), and Professional (-1%) categories,
relative to the market. An enviable performance considering the
challenges healthcare provider markets are facing with the influx
of market changes.

SUMMARY
To lower health costs, physician networks and medical homes
must employ a closed loop population management program
that focus on patient SOH stratification, chronic disease
management, care coordination and incentive management.

To become masters in their population management programs,
they need decision support systems such as population SOH
(risk) stratification and predictive models.