THE QUANTUM PATIENT: ADVANCED AND PREDICTIVE ANALYTICS IN HEALTH CARE



Similar documents
Creating a SimNICU: Using SAS Simulation Studio to Model Staffing Needs in Clinical Environments

How To Analyze Health Data

WHITE PAPER. QualityAnalytics. Bridging Clinical Documentation and Quality of Care

Clintegrity 360 QualityAnalytics

ADVANCING MEASUREMENT OF PATIENT- CENTERED OUTCOMES AND QUALITY METRICS WITH ELECTRONIC HEALTH RECORDS

Why is SAS/OR important? For whom is SAS/OR designed?

MSD Information Technology Global Innovation Center. Digitization and Health Information Transparency

PREDICTIVE ANALYTICS FOR THE HEALTHCARE INDUSTRY

Find the signal in the noise

Secondary Uses of Data for Comparative Effectiveness Research

Transformational Data-Driven Solutions for Healthcare

A leader in the development and application of information technology to prevent and treat disease.

A Proactive Approach to Capacity Management

Accelerating Clinical Trials Through Shared Access to Patient Records

Mastering the Data Game: Accelerating

The Six A s. for Population Health Management. Suzanne Cogan, VP North American Sales, Orion Health

SOLUTION BRIEF. SAP/PHEMI Big Data Warehouse and the Transformation to Value-Based Health Care

Saint Luke s Improves Patient Flow with Help from Apogee Informatics Corporation and ithink

ADVANCED DATA VISUALIZATION

An Introduction to Health Informatics for a Global Information Based Society

Disclosures. Real World Data Sources

How To Change Medicine

User Needs and Requirements Analysis for Big Data Healthcare Applications

A Glimpse at the Future of Predictive Analytics in Healthcare

The registry of the future: Leveraging EHR and patient data to drive better outcomes

MEDICAL DATA MINING. Timothy Hays, PhD. Health IT Strategy Executive Dynamics Research Corporation (DRC) December 13, 2012

SAP/PHEMI Big Data Warehouse and the Transformation to Value-Based Health Care

Demystifying Big Data Government Agencies & The Big Data Phenomenon

What Lies Ahead? Trends to Watch: Health Care Product Development in North America

I n t e r S y S t e m S W h I t e P a P e r F O R H E A L T H C A R E IT E X E C U T I V E S. In accountable care

Measuring quality along care pathways

Duke University Health System Electronic Health Record Update. Art Glasgow Chief Information Officer, Duke Medicine

Carolina s Journey: Turning Big Data Into Better Care. Michael Dulin, MD, PhD

Smarter Research. Joseph M. Jasinski, Ph.D. Distinguished Engineer IBM Research

Electronic Health Record (EHR) Data Analysis Capabilities

The Business Case for Using Big Data in Healthcare

Big Data Analytics- Innovations at the Edge

THE ROLE OF CLINICAL DECISION SUPPORT AND ANALYTICS IN IMPROVING LONG-TERM CARE OUTCOMES

LESSONS LEARNED FROM MOVING TO WEB BASED SURGICAL REQUESTS

KPIs for Effective, Real-Time Dashboards in Hospitals. Abstract

Big Data Trends A Basis for Personalized Medicine

Big Data An Opportunity or a Distraction? Signal or Noise?

Oracle Buys Phase Forward Expands Oracle s solutions for the life sciences and healthcare industries

KNOWLEDGENT WHITE PAPER. Big Data Enabling Better Pharmacovigilance

All Forecasts Are. Not Created Equal

January EHR Implementation Planning and the Need to Focus on Data Reusability. An Encore Point of View. Encore Health Resources

Putting IBM Watson to Work In Healthcare

Redefining Role of Business Analyst in the paradigm of Big Data in Healthcare

Patient Management Systems. Terrence Adam, BS Pharm,, MD, PhD Assistant Professor, PCHS University of Minnesota College of Pharmacy

Big Data and Real World Evidence

Big Data in Healthcare: Current Possibilities and Emerging Opportunities

MID STAFFORDSHIRE NHS FOUNDATION TRUST

HL7 and Meaningful Use

Big Data and Healthcare Information. Ed Reiner Quintiles Transnational

Bench to Bedside Clinical Decision Support:

UAB HEALTH SYSTEM AMBULATORY EHR IMPLEMENTATION

HITAM. Business Intelligence for Hospitals: Empowering Healthcare Providers to Make Informed Decisions through Centralized Data Analysis

The Blue Matrix: How Big Data provides insight into the health of the population and their use of health care in British Columbia

Use and Value of Data Analytics. Comparative Effectiveness Study Inpatient Rehab Hospital (IRH) vs. Skilled Nursing Facility (SNF)

ANALYTICS PREDICTIVE. Tool of Providence or the End of Coincidence? He who does not expect the unexpected will not find it out.

Process Intelligence: An Exciting New Frontier for Business Intelligence

Big Data Analytics in Health Care

!!!!!!!!!!!! Liaison Psychiatry Services - Guidance

DRIVING VALUE IN HEALTHCARE: PERSPECTIVES FROM TWO ACO EXECUTIVES, PART I

Planning and Scheduling in Manufacturing and Services

Analytics and Business Intelligence

96 PD Predictive Modeling: Now What? Moderator: Kara L. Clark, FSA, MAAA

Rehabilitation Network Strategy Final Version 30 th June 2014

Healthcare Informatics and Clinical Decision Support. Deborah DiSanzo CEO Healthcare Informatics, Philips Healthcare

How To Improve Health Care At Stevens.Org

Lessons from the Trenches: Transforming the Experience of People with Advanced Illness

Martin Bowles: using health information to improve care

Agreed Job Description and Person Specification

Master of Bioethics Degree Program

Watson to Gain Ability to See with Planned $1B Acquisition of Merge Healthcare Deal Brings Watson Technology Together with Leader in Medical Images

Oncology Nurse Care Coordinators as. Navigators. Improving cancer disease management and the patient experience

How can you unlock the value in real-world data? A novel approach to predictive analytics could make the difference.

Tapping the benefits of business analytics and optimization

Transcription:

THE QUANTUM PATIENT: ADVANCED AND PREDICTIVE ANALYTICS IN HEALTH CARE Mark Wolff, Ph.D. Health and Life Sciences SAS Institute Cary, North Carolina, USA

HEALTHCARE ANALYTICS THE QUANTUM PATIENT While the individual man is an insoluble puzzle, in the aggregate he becomes a mathematical certainty Sherlock Holmes

CONVERGENCE A central tenet of SAS industry strategy is the concept of convergence SAS understands that transformations needed in the life science and healthcare industries require shared data and insights across previously siloed markets (payers, providers, and pharma) an era of collaboration around analytics Outcomes Quality Value

CHALLENGES ADOPTING ADVANCED ANALYTICS IN HEALTHCARE - Data Access - Data Quality/Integrity - Unstructured/Text - Universal Healthcare Exchange Language - Interoperability - Privacy and Anonymization - Analytic Maturity - Easy of Use

DATA ONE MORE THING ABOUT COMPLEXITY Medicine is a probabilistic science Data quality, completeness, reliability Capacity to describe unexpected events Multi dimensional data Dealing with complexity instead of fighting it! CHRISTIAN LOVIS, MD MPH Professor of Clinical Informatics, University of Geneva Head of Medical Information Sciences, Geneva University Hospitals Geneva, SWITZERLAND

HEALTHCARE ANALYTICS WHY NOW? Maturity of advanced and predictive analytics to address many complex issues in health care Growth in the digitization and availability of health care data Changes in social, economic and legislative considerations related to availability, access and quality of care

ANALYTICS WHAT EXACTLY IS IT? Philip R. Bevington McGraw Hill, 1969

VISUALIZATION JOHN W. TUKEY Automatic analysis techniques such as statistics and data mining developed independently from visualization and interaction techniques The greatest value of a picture is when it forces us to notice what we never expected to see. Important shift from confirmatory data analysis (using exploratory data analysis charts and other visual representations to just present results) to exploratory data analysis (interacting with the data/results) John Tukey J.W. Tukey. Exploratory Data Analysis. Addison- Wesley, Reading MA, 1977.

ADVANCED ANALYTICS HYBRID ANALYTIC APPROACH FOR COMPLEX PROBLEMS Enterprise Data Known Patterns Unknown Patterns Complex Patterns Unstructured Data Associative Linking PATIENT Rules Anomaly Detection Predictive Models Text Mining Network Analysis PROVIDER REGULATOR EFFICACY SAFETY OUTCOMES EHR LABS PHARMA DISCOVERY EPI CLINICAL TRIALS RESEARCH Rules to surface known issues Relevant diagnosis code recorded Prescription not filled Algorithms to surface unusual behaviors New symptom presents # patients with event exceeds expected Identify patterns and relationships to anticipate future events Patterns of patient behavior for known issues Drivers for high cost Enhance analytic methods with unstructured data Extract knowledge from clinical narrative Integration of rich case file information Associative discovery thru automated link analysis across heterogeneous data Linked highly diverse elements Connected network of adverse events with contributing factors Rx PAYER QUALITY TEXT NARATIVE NOTES VALUE HYBRID APPROACH Proactively applies combinations of techniques at entity and network levels

DATA SHARING CHALLENGES IN ADOPTING ADVANCED ANALYTICS Making our medical records open for sharing will save 100,000 lives a year, Google CEO Larry Page told the TED conference in Vancouver today. "Wouldn't it be amazing if everyone's medical records were available anonymously to research doctors?" Page said. "We'd save 100,000 lives this year. Project Data Sphere Data Sharing in Cancer Research We're not really thinking about the tremendous good which can come from people sharing information with the right people in the right ways. Larry Page 19 March 2014 Publication and Access to Clinical-Trial Data The Patient Preference Initiative: Incorporating Patient Preference Information into the Regulatory Processes

GOOGLE BASELINE

CHALLENGES FROM BENCH TO BEDSIDE WHAT HAS CHANGED? Translating the progress in molecular medicine into new therapies has met with limited success; the route from idea to drug has many hurdles and is a very fragmented process. This fragmentation is evident at all levels academia, industry and governments and across geographic boundaries What is driving the change toward a new Collaboration Model? Availability of Massive Amounts of Data Commoditization of Computational Power Availability of High Performance Analytics Newfound Focus on Outcomes Nature Medicine 15, 1006-1009 (2009) Qian et al. Clinical and Translational Medicine 2012 1:25

CHALLENGES UNDERSTANDING, PREDICTING AND INFLUENCING BEHAVIOR The role of patients has changed from being passive receivers to becoming actively involved in their own health care. Patient engagement can improve patient satisfaction and increase adherence to treatment. Facilitates to better clinical outcomes and increased efficiency in health care.

CASE STUDIES ADVANCED AND PREDICTIVE ANALYTICS Patient Reported Outcome Measures (PROMs) Unstructured data and text analytics Clinical Decision Support Clinical knowledge management technologies. Optimization The traveling salesman problem. Discrete Event Simulation Model the operation of a system as a sequence of events in time.

FDA SOLICITATION The objective of this requirement is to provide FDA with the resources needed to use social media to inform and evaluate FDA risk communications. Specifically, the objective is to provide FDA with: Analyses of social media that provide baselines on consumer sentiment prior to FDA communication and that depict changes in social media buzz following FDA communications In-house capability for social media monitoring; and Surveillance through social media listening for early detection of adverse events and food-borne illness. The scope of work includes social media buzz reports, a social media dashboard, and quarterly surveillance reports related to specific product classes.

GOOGLE TREND THIS IS NOT THE WAY

TWITTER RECENTLY PUBLISHED RESEARCH Drug Safety Digital Drug Safety Surveillance: Monitoring Pharmaceutical Products in Twitter Published online: 29 April 2014 "The resulting dataset contained a high volume of irrelevant information, but provided a useful starting point." "We did not seek to verify each individual report as truthful, but rather to identify overall associations between Twitter and official spontaneous report data as a preliminary proof of concept."

TEXT ANALYTICS EMBRACING COMPLEXITY IN UNSTRUCTURED DATA How to build a Lie Detector for the internet; Semantic Field Normalization/Contextualization for Self-Reported Symptom- Treatment-Outcome Measurement in Web-based Media Sources Adaptation of Semantic Nets to Establish Veracity of Symptom-Treatment Outcome Reports in Health Related Web Interactions Behavioral context as a pathway to crafting semantic field normalization mappings in Clinician/Patient Reported Outcomes Data (C/PROM)

DATA ON LINE DATA SOURCES

VISUALIZATION FROM NARRATIVE TO COHORT TO INSIGHT MULTIDIMENSIONAL TEMPORAL ANALYSIS

Patient preferences considered for the first time in FDA decision to approve first of kind obesity device RTI Health Solutions partnered with the FDA to conduct a study on patients preferences which contributed to the Agency s regulatory decision to approve a firstof-kind device to treat obesity This was the first time a patient preference study impacted a new device approval Incorporating patient-preference evidence into regulatory decision making Surgical Endoscopy January 2015 Martin P. Ho, Juan Marcos Gonzalez, Herbert P. Lerner, Carolyn Y. Neuland, Joyce M. Whang, Michelle McMurry-Heath, A. Brett Hauber, Telba Irony

MENTAL HEALTH CARE THE CHALLENGE OF IMPROVING SUICIDE RISK ASSESSMENT Suicide risk assessment in mental health usually relies on data from the patient to identify those at low, moderate or high risk. Most people who die by suicide are low risk and most high risk people do not kill themselves. It would be valuable to examine the notes made at the time of assessment and subsequent care, yet this is time consuming and is often done only during formal inquiries when a suicide has already occurred.

SUICIDE RISK PROJECT UNIVERSITY OF OTTAWA & BRAIN AND MIND RESEARCH INSTITUTE Can we identify and change high risk systems of mental health care? OBJECTIVE: Phase I Phase II Phase III Improving suicide risk assessment through automated and semi automated knowledge extraction from clinical notes and narrative associated with patient assessment of suicide risk in acute and care management scenarios. Text Analytics and Machine Learning applied to clinical notes/narrative Develop Linguistic Rules for risk assessment based on mathematical approaches and clinical best practices Validate, test and compare approaches against historical clinical data

TEXT ANALYTICS IMPROVING OUTCOMES A study of over 300,000 free-text machine-readable documents in the Stanford Health Care EHR finds that text-mining tools can be used to detect unplanned care episodes documented in clinician notes or in coded encounter data. Researchers believe their methods could be used for quality improvement efforts in which events of interest occur outside of a network that allows for patient data sharing.

OPTIMIZATION THE HOSPITAL GAME Optimizing Surgery Schedules to Save Resources, and to Save Lives Three key steps to a successful Operations Research Optimization Problem Identify real-world challenge as relevant to mathematical optimization formulation Does one understand the real-life situation to allow for a close-enough mathematical formulation, such that the solution will be relevant to real world decision making. Translate a real-life problem formulation into mathematical language Develop preliminary mathematical statement, including control variables, objective and the constraints. Formulate complete, in-depth understanding of all known influencing factors and outcome expectations. Develop and refine mathematics and algorithms: Solve problem by pushing it toward a desired optimum. https://support.sas.com/resources/papers/proceedings13/154-2013.pdf SAS Global Forum 2013 Paper 154-2013 Andrew Pease and Aysegul Peker, SAS Institute, Inc.

SOLVER MIXED INTEGER PROGRAMMING (MILP) PROC OPTMODEL offers the following types of solvers for different types of mathematical programming problems: Linear programming (LP) solver Mixed integer linear programming (MILP) solver Nonlinear programming solver Quadratic solver Integer programming solver The Master Surgical Schedule problem falls into the mixed integer programming problem category. The solver implements a LP-based branch-and-bound algorithm. Branch-and-bound strategy divides the feasible region in to disjoint sub-problems. The solver then incorporates pre-solving, cutting planes, and additional heuristics to increase the efficiency in finding an optimum.

RESULTS LENGTH OF STAY Figures represent length of stay in the nursing wards for a schedule that was generated for a single operating room in a two-week period This simulation assumes that the nursing wards were empty on the first day and gradually fill up. Optimized schedule creates a much more steady buildup in occupancy, demonstrated on day 2. At end of end of the second week, there are numerous discharges in both plans. Buildup is more stable as related to manual scheduling.

RESULTS OCCUPANCY Reduction in the number of beds required to meet care needs Reduction in the variability of occupancy rates Indicates a slightly higher average occupancy

SUMMARY THE HOSPITAL GAME Surgical procedures and scheduling are highly variable between hospitals Successful implementation of an optimization is related to how well business requirements are formulated. Implementation must strike a balance between approximate and exact implementations. Too many business restrictions can result in an unstable solution, whereas Too few can result in an irrelevant solution that does not work. Any analyst who wants to develop a schedule optimizer for a master surgical schedule can follow the steps in this paper. Important benefits of an optimized master surgical schedule include; Data driven understanding of the maximum number of resources that are required downstream for post-operative nursing care. Achieving a more steady utilization of available resources.

SIMULATION A SMALL BIG ANALYTICS PROJECT A Small Application of Big Data Analytics in Healthcare Create a discrete-event simulation model for a NICU's patient mix (SAS Simulation Studio) More accurate nurse scheduling Better match between nursing ratios and acuity Better preparation for changes in status Optimize balance between cost and quality Chris DeRienzo, MD, Chief Patient Safety Officer, Mission Health System David Tanaka, MD, Professor of Pediatrics, Duke University Medical Center Emily Lada, PhD, Senior Operations Research Specialist, Advanced Analytics, SAS Phil Meanor, Senior Manager, Advanced Analytics Division, SAS SAS Global Forum 2014 - Paper 1361-2014 http://support.sas.com/resources/papers/proceedings14/1361-2014.pdf

DUKE NICU OPTIMIZING VALUE Duke Hospital NICY q24 Staffing needs: Three Neonatal Fellows Four Attending Neonatologists Five Pediatric Residents Five Respiratory Therapists Nine Neonatal Nurse Practitioners OVER SIXTY NURSES

DATA SIMULATION INPUTS Number and Type of Admissions per Day Inborn / Transfer and Gestational Age (GA) Predicted Length of Stay (LOS) / Model Exit Probability envelopes for major morbidities Timing of events Affect on LOS Temporal affect on acuity Model Exit Death Discharge Transfer Constraints Number/type of beds, number/availability of transfer beds, number of nurses

PROBABILITY ENVELOPES PREDICTED LENGTH OF STAY (LOS) / MODEL Exit Probability envelopes for major morbidities Timing of events Affect on LOS Temporal affect on acuity Model Exit Death Discharge Transfer Constraints Number/type of beds, number/availability of transfer beds, number of nurses

SIMULATION MODEL LOGIC

SIMULATION VALIDATION

THE ANSWER TO THE BIG QUESTION Contrary to the current belief that reduction of ALOS implies a reduction in hospital resource utilization due to improved care, the exact opposite appears to be true. Hospital administrators should seriously consider high fidelity modeling before initiating a one size fits all approach to cost containment strategies.

ANALYTICS HAS THE TIME FINALLY ARRIVED It is predicted that digital electronic computers will assume an increasingly important role in medicine William R. Best, M.D. (1962) Since the earliest days of computers, health professionals have anticipated the day when machines would assist in the diagnostic process Edward H. Shortliffe, M.D., Ph.D. (1987) Computer Programs to Support Clinical Decision Making Reply In Brazil and India, machines are already starting to do primary care, because there s no labor to do it. They may be better than doctors. Mathematically, they will follow evidence and they re much more likely to be right. Robert Kocher, M.D. (2013) 1959

THE FUTURE HAS ARRIVED Mathematical models out-perform doctors in predicting cancer patients' responses to treatment "If models based on patient, tumour and treatment characteristics already out-perform the doctors, then it is unethical to make treatment decisions based solely on the doctors' opinions. We believe models should be implemented in clinical practice to guide decisions. Cary Oberije, Ph.D. MAASTRO Clinic Maastricht University Medical Center, Maastricht, The Netherlands. 2nd Forum of the European Society for Radiotherapy and Oncology (ESTRO). Abstract no: OC-0140, "Clinical 2 Lung and Head & Neck cancer" Geneva, Switzerland 19-23 April 2013

HEALTHCARE ANALYTICS THE QUANTUM PATIENT When schemes are laid in advance, it is surprising how often the circumstances fit in with them. Sir William Osler, New York Academy of Medicine in 1897