Transcript Slide 1

Finding High-Risk HF Citizens in
BC to support Primary Care A Work In Progress Update
February 28, 2013
Ella Young, Director of Care Continuum
and Actuarial Analytics, VCH
[email protected]
Objective
Event 1
e.g.
Diabetes
diagnosis
Event 2
e.g. Heart drug
T-(24+y)
mos.
Prediction
T-(24+x)
mos.
T-24
mos.
Event 3
e.g. Knee xray
HF
Prevented or
Mitigated Event
Heart Failure
•
The inability of the heart to pump blood to meet the oxygenation and nutritional needs of the
tissues, with multiple systems affected and participating in the dysfunction - more than just a
weak pump
•
Affects men and women equally. Women tend to be older with a history of hypertension
(HTN) when first diagnosed; men tend to be younger at onset and have a history of
coronary artery disease (CAD)
•
More common as a person ages, and as more of the population ages the incidence and the
prevelance of HF in the population is expected to increase
•
Has an annual mortality anywhere from 5% to 50%, depending on the severity of the
dysfunction and associated symptoms (ie. pulmonary edema) along with other factors
(ie.co-morbidities)
•
HF is associated with numerous symptoms and causes and subsequently there are many
ways a person may present. It continues to be a syndrome that is difficult to diagnosis and
to treat effectively.
3
Analysis / Modeling
Exploratory Analysis
 Consider relative risk factors to predict future
expected events
 Avoid “non-response bias”
 Identify people with lower, emerging, risk
 Step 1
◦ Who? What common characteristics?
◦ What are the implications of those characteristics?
 Step 2
◦ scarce resource allocation for maximum ROI
◦ Who is intervenable or impactable?
N = 20,148
Pop = 4.5 m.
New CHF/Pop = 0.5%
Pre-HF Chronic Conditions of the 09/10 HF Incident Cohort
% of HF patients with another select Chronic Disease
90%
HTN
80%
70%
60%
50%
IHD
OA, DEPR, DM
ANGIO
40%
COPD
30%
CKD
OSTEOPOROSIS
20%
DEMENTIA
PTCA, CABG
10%
STROKE
RA
0%
-8
-7
-6
-5
-4
-3
-2
Average incidence time versus HF (years)
-1
0
How good are the various models
at figuring out if your patient is at
high risk for heart failure?
“ Predictions are hard,
especially about the future.”
Niels Bohr
Nobel Laureate in Physics
Preliminary Findings – No Surprises
•
Membership in chronic disease registries and lab tests are important
•
Of approx. 800 initial variables, about 15 remain (depending on strata)
•
Decision trees and path analysis found 100 significant variables from the
800
•
Of the General Linear Models, discriminant analysis did the best - it
achieved a classification rate of about 70% for age groups of interest
•
Neural network modeling was next at about 65-71% classification rate
•
Logistic regression resulted in a 60-65% classification rate
Preliminary Results - Survival Analysis
• Examines the time it takes for events to occur,
allows for time-varying predictors, and late entry
into the risk set
• Among the best results at 85%+ classification rate
on sample data sets and stratum of interest
• Indicates that MSP and drug amounts are important
• Drug costs related to lipids showed significant
protective effects
Conclusion
• To date, survival analysis achieves the
best result overall
– Over 9 years, it finds about 85% of 45+ yearold who became HF incident
– Over a shorter timeframe, predictive power
increases, so within two years it yields over
90% classification rate
Summary of
“Old Way” vs. “New Way”




Not focusing your interventions
on this year’s future high-cost
group
Relying on claims & pharmacybased approaches
Expecting that people will
participate because “it’s the right
thing to do”; weak, nonexistent, or
inconsistent incentives
Poor timing, data, and PM
logistics
 Predictive modeling: focus your
most powerful interventions on
the right people!
 Most companies realize the
advantage of linking data and
using advanced analytics
 Powerful incentives: Enough of
the right people have to engage
in the process to produce
enough benefit to SEE a positive
ROI
 The entire front-to-back PM
process has to incorporate the
best modeling, and be
implemented properly
14
Next Steps
• Identify VCH citizens at high risk of HF
incidence and communicate them with
providers
• Work with Primary Care/GPSC/BC HF SC
to provide evidence-based care for those
identified
• More value to model ??
Possible Clinical Next Steps
• Physician Profile Report
• Actionable conversations with impactable patients
• Risk-reduction and education e.g. if AMI, then
cardiac rehab, condition management, etc.
• Ask about swollen ankle, take BP each visit, etc.
• Care plan
• iCDM
• Blood test -> biomarkers ?
• HF clinic consult
• Develop validation study with GPs ??
• Other ??
Physician Profile Report
Q&A
Thanks for your feedback!
PROOF Centre
Blood Tests for Care of
Heart Failure and COPD Patients
■ Bruce McManus
■ Practice Support Program
■ Thursday, February 28, 2013
“The only way to keep your
health is to eat what you don’t
want, drink what you don’t like,
and do what you’d rather not.”
~ Mark Twain
PROOF Centre’s Solution
Blood Molecules
Tissue mRNA
or miRNA
White blood cell
mRNA or miRNA
Plasma proteins and
metabolites
Signatures from ~38,000 blood and/or tissue
components computationally analyzed
21
Test Development Process
Clinical
Need
Biomarker (Blood
Test Content)
Identification /
Replication
Develop Clinical
Laboratory
Blood Test
Clinical
Use
Clinician Driven
Health Economics Evaluation, Commercialization and
Implementation Strategies
22
50%
of heart failure patients die within
up to
5 years of diagnosis
more than
$4B
is spent on heart failure care
each year in Canada
23
Improving care of
heart failure patients
Develop blood tests that will
allow physicians anywhere to
diagnose chronic heart failure
and to monitor therapeutic
responses and outcomes
24
How will our Blood Tests be used
to manage Heart Failure?
Family
practice
Person
with
physician
heart
failure
symptoms
Diagnostic
blood test
Person with risk
for heart failure
DHF
SHF
Blood tests to
monitor
response to
therapy and
hospitalization
/ deterioration
outcomes
Admin Data Predictor (Ella Young)
DHF = diastolic heart failure
SHF = systolic heart failure
25
COPD
Exacerbations [lung attacks] are the
leading cause
of hospitalization in Canada
$1.3B/year is spent on hospitalization
Affects >200M people globally
26
Improving care of
patients with COPD
Develop blood tests that will
allow doctors to identify who
•will have frequent COPD
exacerbations [lung attacks]
•is having a COPD exacerbation
27
How Will Our Prognostic Test
Change COPD Patient Care?




High AECOPD Risk
Prognostic
Blood Test


Low AECOPD Risk




Evidence-Based
Treatment


28
Benefits for patients
with COPD
•Improved evidence-based
management of COPD
•More family medicine-based care,
fewer COPD hospitalizations
•By 2020, help >450,000
patients/year, save >50,000
QALY’s/year, and >$600M/year
29
How do we translate our results to
physicians and care teams?
Cell phone, iPad, computer, etc
Score: 1… 10
Short
Explanation
Mouse-over: Full
Recommendation
Our Passion
• Focused on each patient
• Grounded in clinical reality
• State-of-the-art analytic algorithms
and access to data
• Quality-assured technologies
• Health economics modeling
• Ability to engage key partners across
all sectors
31
THANK YOU
www.proofcentre.ca
[email protected]