Using Novel Data Sources DCRI CENTER FOR for Real-world PREDICTIVE MEDICINE Impact
WHO WE ARE Why DCRI? The Duke Clinical Research Institute (DCRI), the world’s largest academic clinical research • World’s largest academic organization and part of the Duke University School of Medicine, is known for conducting research organization (ARO) groundbreaking multinational clinical trials, managing major national patient registries, and • Nonprofit focused on outcomes, performing landmark outcomes research. DCRI’s Center for Predictive Medicine (CPM) aims not revenues to improve patient care through the development and application of novel approaches to quantify and communicate risk. • Partnership with the Duke University Health System
• Access to Duke’s clinical WHAT WE DO excellence and thought The availability of electronic health records (EHR), clinical trial data, claims data, and other leadership shared data has opened exciting new opportunities in predictive medicine. Use of predictive • History of successful analytics to assess the risk of future disease offers tremendous benefits to individual patients collaborative projects as well as to the clinical research and healthcare communities as a whole. • Experience with diverse HOW WE DO IT technologies and indications
FRAME THE RESEARCH QUESTION • Talented statistics and programming teams Define the research question. • Collaborate with clinical partners to frame the research question and identify the • Recognized innovators population of interest. • Pragmatic approach to projects
TURN COMPLEX DATA INTO SMART DATA Identify data sources. • Apply expertise with novel and complex data sources to match appropriate data source to the research question and population. Champions of • Create and validate analytic datasets. Collaboration
Analyze the data. HOUGHT L T LEA • Apply statistical methods and techniques appropriate to the research question. ICA D N ER LI S C • Implement methodological research techniques related to meaningful use of EHR data, high-dimensional data, and risk prediction.
CENTER FOR I
PREDICTIVE N
F
Y
O
T
I MEDICINE
IMPACT CLINICAL PRACTICE AND PATIENT HEALTH R
M N
U A
T
M
Disseminate results. I C
M
I A
O
N
C • Collaborate with clinical partners to accurately interpret and present results. S • Facilitate integration of risk-prediction models into EHR. DCRI CENTER FOR PREDICTIVE MEDICINE
Find out more about Examples of our experience in applying novel data sources to answer complex questions the DCRI Center for include the following: Predictive Medicine. DUKE DATA CLINICAL TRIALS
David Gee • Link data sources within Duke’s health • Support primary and secondary analysis Assistant Director, system, including Duke Databank for of clinical trial data for more than 530 Cardiovascular Disease, Duke Echo Lab publications in peer-reviewed journals Business Development, Statistics Database, MRI and Cardiac Computed since 2000. Phone: 919-357-2449 Tomography databases, and healthcare • Pool clinical trial datasets around specific Email: [email protected] encounters from Duke’s EHR to provide disease indications to leverage increased analysis for publications. dcri.org power to answer research questions not possible from a single data source.
EHR DATA OBSERVATIONAL DATA • Validate electronic phenotype algorithms • Pool datasets from five NHLBI-funded for outcomes and procedures including cohort studies to develop and validate hospitalization for heart failure, stroke, risk-prediction models for cardiovascular acute MI, diabetes, CKD, HIV diagnosis, outcomes. revascularization, and death. • Use the CDC’s National Health and • Use Duke EHR data to provide clinicians Nutrition Examination Survey data to with real-time risk assessment of sepsis describe the distribution of biomarkers and other adverse outcomes in the in patients with a history of MI and hospital setting. correlate with standard risk factors.
CLAIMS DATA COMBINED DATA • Process MarketScan claims data using • Create and validate risk-prediction advanced programming techniques to models across multiple data sources, handle large datasets. including MarketScan, pooled NHLBI cohort data, and registry data. • Link Centers for Medicare and Medicaid Services claims data with clinical trial • Apply predictive value of discharge codes data. in Duke EHR data to calibrate estimated encounter rates based on MarketScan data.
Produced using paper from sustainable forests | DCRI Communications February 2017