A Model Mashup Environment for Healthcare Decision Support · A Model Mashup Environment for...
8
IBM RESEARCH © 2009 IBM Corporation INFORMS 2009 A Model Mashup Environment for Healthcare Decision Support Peter J. Haas Paul Maglio Pat Selinger
Transcript of A Model Mashup Environment for Healthcare Decision Support · A Model Mashup Environment for...
IBM RESEARCH
© 2009 IBM CorporationINFORMS 2009
A Model MashupEnvironment forHealthcare DecisionSupport
Peter J. HaasPaul MaglioPat Selinger
IBM RESEARCH © 2009 IBM Corporation 2
I’m back at IBMand seeing the world in a differentway!
We data folks thought we were really smart when we:Made relational systems do high performance
transactionsDid data mining and business intelligence on transactional
dataIncluded unstructured data, web, text, …Included streams of dataScaled to terabyte / petabyte scaleAdded semantics, annotations….Exploited metadata….
IBM RESEARCH © 2009 IBM Corporation 3
But….Understanding the world and deriving amodel of its aspects by just analyzing dataalone should be a last resort
We can understand SO much more if wemake our analysis and predictions bothmodel-driven as well as data driven
Especially when solving complex systemsinvolving systems of systems
IBM RESEARCH © 2009 IBM Corporation 4
Healthcare: A complex “system of systems”
Education Health careProviders
Health carePayers practices Health care
ClientsDemographics Disease modelsPharmacompanies
Governmentpolicy Treatment
standards Cultural andSocial influences
Recreational facilitiesTransportationAdvertising
Agriculture
…and more
Economics
SupplyChain
ImmigrationPatient perceptionand satisfaction Consumer/patient
behavior
Need a platform and processes for integrating models andsimulations for healthcare-related policy, investments, and planning
Climate
Hospitaland clinicoperations
Many systems have been studied individually by domain experts,using statistical and simulation models
CommunityHealthcarenetwork
Comparativeeffectiveness
IBM RESEARCH © 2009 IBM Corporation 5
Example:
Dunn and Bradstreetbusiness data
Insight: Nearby location of large chain grocery stores reduced obesityrates
Tax incentives for chain stores to move to obesity “hotspots”?10% incr. in fast food prices -> 3% incr. fruits and vegs -> BMI -10% in
obese youthChaloupka FJ, Powell LM. Price, availability, and youth obesity: evidence from Bridging the Gap.
Prev Chronic Dis 2009; 6(3). http://www.cdc.gov/pcd/issues/2009/jul/08_0261.htm
Obesity incidence andtreatment
Public Policy Investment Decision Support
US Census data
Model of accessto physical activity Model of transportation
Diet choices and costsmodel
But:Statistical models onlyResults integrated by hand
IBM RESEARCH © 2009 IBM Corporation 6
SPLASH! (Smarter Planet Platform for Analysis and Simulation of Healthcare)
A platform and service to “mash up” data, models, simulations Support complex decision-making for healthcare policy, planning,
and investment Share, integrate, correlate deep-domain models and data with those
of others Exploit the tools, models, simulations, data and analytics of others
Benefits
Significant new insights frommashing up proprietary andopen models from multipledisciplines
High quality of results fromcollaboration, interoperability,and integration
Rapid exploration of a largenumber of alternative inputsand outcomes
IBM RESEARCH © 2009 IBM Corporation 7
The Challenges of Integrating Models
Exploring the landscape of model combinationNot all models and simulations can be combined.Which ones can?
Need methodology for describing models so thatwe can determine which models are “compatible”
Building a shared framework with easyintegration tools and connectors
Building a community of platform users Must provide motivation to participate Must build trust in the models and data
IBM RESEARCH © 2009 IBM Corporation 8
We’re Looking for Input and Partners
Model-integration technologyScenarios and use casesSimulation models
• Discrete-event, agent-based, systemdynamics, differential-equation, …
Statistical models• Regression, time series,decision/classification
Datasets• Clinical, econometric, operational, …
Pat Selinger: [email protected] Maglio: [email protected]
Peter Haas: [email protected]
