Med Biquitous Annual Conference 2013 IBM Watson in

Watson Workload Optimized System (Power 750) – Original 90 x IBM Power 7501 servers 2880 POWER 7 cores POWER 7 3. 55 GHz chip 500 GB per sec on-chip bandwidth 10 Gb Ethernet network 15 Terabytes of memory 20 Terabytes of disk, clustered Can operate at 80 Teraflops Runs IBM Deep. QA software Scales out with and searches vast amounts of unstructured information with UIMA & Hadoop open source components SUSE Linux provides a cost-effective open platform which is performance-optimized to exploit POWER 7 systems 10 racks include servers, networking, shared disk system, cluster controllers Note that the Power 750 featuring POWER 7 is a commercially available server that runs AIX, IBM i and Linux and has been in market since Feb 2010 1 2 © 2012 International Business Machines Corporation

Watson Today 3 1 to 4 p 750's 16 cores 256 GB RAM Enhanced core algorithms Large case question Health care specific reasoning Health language normalization New health care specific scoring New NLP processors New Data Engine New Data Model © 2012 International Business Machines Corporation

Key Differences Between Jeopardy and Healthcare 4 Jeopardy Healthcare Single, known correct answer Multiple possible answers, “best” may be debatable No substantiating evidence allowed Substantiating evidence essential Single dimensional input from one source Multi-dimensional input from multiple sources No dialogue/interaction by/with system Interaction by/with system – “prompting” for information Fixed repository Dynamic repository Evidence profile and underlying algorithms based on trivia questions Evidence profile and underlying algorithms based on medical questions 08/10/11 Consequence: Significant development required © 2012 International Business Machines Corporation

Administrative Simplification Process with Watson Physician records symptoms, medical history, and vitals Patient 5 Physician determines patient is at risk for breast malignancy and wants a breast MRI. Staff requests MRI pre-authorization from Well. Point Utilization Management (UM) Clinical staff Utilization Management Registered Nurses (RN) UM RN reads the pre-auth request (fax/email/transcribed vm) and creates a summary. RN queries Watson and reviews Watson’s recommendation Utilization Management Registered Nurses (RN) Medical Policies / Care Guidelines RN makes a decision: authorize or forward request to WLP MD Utilization Management Registered Nurses (RN) Clinical staff © 2012 International Business Machines Corporation

Oncology Advisor Use Case Workflow Patient Discuss Treatment options Patient Data Nurse Clinical Coordinated collaborative care P C P Oncologist Presumptiv e Diagnosis EMR Create tailored treatment program and adjust as needed Claims Query for patient specific treatment options (e. g. , mortality, morbidity, likelihoods) Office Staff Treatment options S U R G E R Y T H E R A P Y Comprehensive case management Notify additional care givers of treatment authorization Care Management Authorize full treatment program instead of each episode of care Longitudinal Patient Health Information Medical Literature Care Guidelines Morbidity and Mortality Info Genetic Testing Medical Policy Cost Information Treatment Analyzer Summary Benefits Increase evidence based treatment Streamline pre-auth / claims administration Improve coordination of collaborative care 6 Pre-Authorization Automated claims authorization for providers in patient’s care delivery team Claims Management © 2012 International Business Machines Corporation