Allscripts Open App Challenge Cardinal Team Cardinal Optimization turning theory into reality Canary: An Analytical Tool for Reducing 30-day Readmission Rates Contact: Holly Jin, President [email protected]
Allscripts
Open App
Challenge
Cardinal Team
Cardinal Optimization turning theory into reality
Canary: An Analytical Tool for
Reducing 30-day
Readmission Rates
Contact: Holly Jin, President [email protected]
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 2/15
Healthcare Problem Addressed
• Healthcare/Business problem addressed: – 71 million+ individuals in the US admitted to hospital each year
– Over $30 billion spent on unnecessary hospital admissions (2006)
– Lack of effective tools for healthcare professionals to accurately
predict high risk patients for hospital visits in order for them to
take preventative measures to get the early treatments patients
need to avoid or shorten hospital stays
• Canary app addresses both the following categories: – Applications that improve management of high cost chronic
diseases, such as diabetes, stroke, renal, cancers, and others
– Innovative approaches to effectively reduce 30-day readmission
rates
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 3/15
High Level Solution Description
• Use of patients’ historical claims records in building
analytic models to effectively predict future hospital
visits of high risk patients in particular in the high
cost categories of chronic diseases
• Imbed the Canary inside an Allscripts app fully
taking advantage of Allscripts’ huge provider
/patients database
• High risks patients for next 30-day hospital
admission are identified with index 1 through 10 to
indicate risk factors
• Healthcare providers equipped with Canary apps
utilize the predictive results in order to take actions
to prevent unnecessary hospitalizations
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 4/15
Novel Approach
• Support Vector Machines (SVM)
– A large-scale, parallelizable framework for rapidly
classifying sets of data
• Bayesian analysis
– A set of statistical algorithms that allow a user to compute
probabilities of future events of interest
• Statistical regression
– Applications of high-dimensional calculus and linear
algebra for establishing relationships between variables and
predicting likely future events
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 5/15
Usability and Design
• PCPs equipped with the Canary apps receive
realtime (daily) warnings of relevant high risk
patients for hospital visits for the next 30 days
• PCPs take preventive actions on these high risk
patients leading reduced number of patients for
hospital visits or the length of stays
• In particular, the Canary app also provides high risks
factor index to the PCPs to assist them with
diagnosis and preventive actions
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 6/15
Development Stage
• A working prototype has been developed
– Prototype built utilizing 3 years of patients claims data from
the Heritage Provider Network
– Code development already in place for building effective
predictive algorithms used in Canary app
• Clustering analysis and prediction of hospital stay
risk index for Renal patients as pilot applications
developed
• Further analysis and expansions to other high cost
chronic diseases that resulting in frequent hospital
visits including heart disease, diabetes, cancer and
others
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 7/15
Integration Description
• Allscripts integration
– Canary app integrated with Allscripts EHR in order to obtain
relevant historical claims data
– Canary app push relevant data back to Allscripts EHR
• Additional data in refining Canary engine (Phase 2)
– Laboratory, clinical and drug data from Allscripts
– Integration with IMPACTMeds medication adherence and
community pharmacy data
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 8/15
Allscripts Integration: APIs/Web Services
Canary App AWS/Unity
WCF/Webservices
Canary App <-> Unity SDK data flow
1.Start session authenticating via Unity GetSecurityToken()
2.Data retrieval/update using Unity Magic commands
e.g. GetChangedPatients(), GetPatientActivity(), GetPatientIDs(),
GetProcedures(), GetPatientList() etc.
3.Close session using Unity RetireSecurityToken()
Canary DB Allscripts DB
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 9/15
Business Partners and Customers
• Support from IMPACTMeds Nephrology platform
• Dallas Nephrology Associates
– The second largest group of practicing nephrologist in the
country and will participate in multiple ACO’s.
– IMPACTMEDS received endorsement from Dallas
Nephrology Medical Director and Fresenius Medical Director
• Dialysis Clinics: interested as Beta customers
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 10/15
Market Strategy
• Pilot to Dallas Nephrology Associates
– An Allscripts customer
– 72 Nephrologists in 30+ locations
– 30,000 patients
– Offers pilot participations in using Canary app to predict
next 30-day hospital stays for patients under the care of a
nephrologist
• License Canary app to DaVita and Fresenius as beta
customers (Planed in Phase 2)
– Awareness of the doctors and staffs to high risk patients
benefit them in keeping patients out of hospital as business
goal
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 11/15
Market Strategy +
• Hospitals and specialties of internal med using
Allscripts will be targeted as customers and users
• Seeking Government healthcare agencies in
providing incentives to vendors for utilizing Canary
resulting bigger savings to the government
• Seeking healthcare insurance companies in
providing incentives to vendors for utilizing Canary
app resulting huge reductions to the healthcare
insurance cost
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 12/15
Business Model
License the application to organizations as a per
member per month fee.
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 13/15
Video Demonstration
• Link: http://www.youtube.com/watch?v=iafvbkDgAqc
• Description of what the demo shows: A preview that
shows how Canary can be used to forecast future
patient hospital stays, analyze historical patient
information, and identify the “risk factors” that affect
patients’ needs to use the hospital.
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 14/15
Why this solution be selected !
• Great market need and significant social impacts:
– Improving patients’ quality of life and saving lives
– Effective management tool to healthcare professionals
– Resulting in huge savings in the billions of dollars to
Medicare/Medicaid cost
• Ease of use:
– User friendly interface
– Intuitive statistically sound results
– Easy integration to Allscripts customers
• Committed pilot customers on Allscripts, Beta
customers lined up
• Simple and affordable pricing model
Cardinal Optimization Proprietary and Confidential (Not
permitted to distribute to third party without prior consent) 15/15
Why this solution be selected +!
• Novel solutions for a complex and difficult problem:
– Innovative technology: uses of cutting-edge research
developed from Cardinal team members who specialize in
computational mathematics, data analytics, and statistical
analysis
– Accuracy: expected improvements over existing methods
by up to 15x (see submitted video)
– Scalability: algorithms are parallelizable and can be applied
to huge datasets (thousands of dimensions/millions of
entries)
• Award-winning innovative team members:
– team members are highly innovative area experts
– PhDs credentialed from Stanford, Harvard, and Cambridge