Million Veteran Program Industry Day Phenomics J Michael Gaziano MD MPH Background Traditional cohorts Nurses Health Study Physicians Health Study Framingham Heart Study Womens Genome Health Study etc ID: 762149
Download Presentation The PPT/PDF document "Million Veteran Program: Industry Day P..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Million Veteran Program: Industry Day Phenomics J. Michael Gaziano, MD, MPH
BackgroundTraditional cohorts: Nurses’ Health Study, Physicians’ Health Study, Framingham Heart Study, Women’s Genome Health Study, etc.Consortia: Cancer Cohort Consortium, eMERGE, CHARGE, dbGaP, etc.Other large-scale biobanks: BioVU, UK Biobank, Kaiser, Geisinger, etc.
Nesting Population Research in the VA Healthcare SystemVA ideal setting for nested large-scale population researchStable and willing veteran population of 8 million using the system each yearOutstanding electronic medical record; fully integrated; data reaching back as far as 20 years; access to CMS and NDI dataResearch infrastructure with diverse expertise Prototypes for health system based research:Pragmatic trial of HCTZ vs ChlorthalidoneMillion Veteran Program
Million Veteran Program (MVP) Enroll up to one million users of the VHA into an observational mega-cohort Collect health and lifestyle information Blood collection for storage in biorepository for future research Access to electronic medical record Ability to recontact participants Million Veteran Program Veterans Health Administration Million Veteran Program (MVP)
The Future: “Big Data”More and more data is becoming available for health care research: is it a blessing or a curse? Sometimes, data warehouses resemble landfills more than libraries.
Million Veteran Program (MVP) Data Universe 6 VA - Clinical VINCI, VIReC, Self-reported MVP surveys Molecular Data Non-VA NDI, CMS, etc. Biospecimens MVP Participant
MVP Data SourcesMVP Consented DataSelf-Reported Survey Data:Lifestyle Survey Data (Personal Information, Well-Being, Activity, Health, Military Experience, Dietary Intake, Medication, Habits) Baseline Survey (Health, Military Experiences, family medical history)Future data collection, follow-up surveys, Genetic DataGenomic data (Sequence, genotype, etc.)Other omics Passive Data Non-VA Data National Death Index (NDI) Centers for Medicare and Medicaid Services (CMS) DOD Other novel data sources: particulate matter, geospacial, etc. VA Healthcare System Data 7
VA Data SourcesCorporate Data Warehouse DatabasesNational Patient Care DatabasesVital StatusDecision Support SystemNational Data ExtractBeneficiary Identification Records Locator (BIRLS) death fileNew England VISN-1 Pharmacy filesOutpatient Clinic File (OPC) Patient Treatment File (PTF) Inpatient and Outpatient Hospitalizations Clinic Inpatient and Outpatient Visits Diagnosis (ICD-9) codes Procedure (CPT) codes Pharmacy data and laboratory data Pharmacy Benefit Management (PBM) system databaseOEF/OIF and OND RosterVA Clinical Assessment Reporting and Tracking (CART) Veterans Affairs Surgical Quality Improvement Program (VASQIP) Veterans Affairs Central Cancer Registry (VACCR) 8 Special Data Access w/ Data Steward National Data Systems (NDS)
VA VINCI Strategic Roadmap
MVP System Architecture 10 Access Authorization by Governance System Vendor Molecular Lab Query Mart Query Portal Analysis Environment Consent Manager Study Mart Study Mart Study Mart Data WarehouseHonest Broker VA Non VA Clinical Data NDI, CMS Survey Data Molecular data Researcher
MVP Complex Phenotypes Strategy Develop data processing pipelines to accommodate a high-throughput EMR databaseQuantify phenotypes using a probabilistic algorithm approachCreate phenotype data library resource that is reproducible, reusable, scalable and efficientRepurpose data elements as the library grows.
Step 1: Define initial working algorithm (T1A)Step 5: Derive T2A Step 2: Create study cohort and apply T1A Step 6: Evaluate T2A to formulate T3A Step 3: Create Annotation Data Set Step 7: Develop probabilistic model and assign caseness Step 4: Create Phenomic Database through Data Processing PipelinesDeposit resulting algorithms to a central Phenotype Library
MVP Phenotyping ActivitiesComplex Phenotypes DiseaseMyocardial infarction (MI)StrokeUnstable angina with revascularizationAcute congestive heart failureDeath from cardiovascular diseaseVascular procedurePosttraumatic stress disorder (PTSD)SchizophreniaBipolar disorderTraumatic brain injury Depression Vascular dementia Cognitive impairment Type 2 diabetes mellitus Other Creatinine trajectory Glucose tragectoryAlgorithm DevelopmentValidation MethodsCore Variables DemographicsAgeSexRaceLaboratory valuesTotal cholesterolHDL, LDLAlbuminSerum creatinineTriglyceridesMedicationsOther characteristicsBlood pressureHeight/weight/BMISmokingAlcohol consumptionCombat exposure 13
eMERGE PhenotypesACE Inhibitor (ACE-I) induced coughADHD phenotype algorithmAppendicitisAtrial FibrillationAutismCardiac Conduction (QRS)CataractsClopidogrel Poor MetabolizersCrohn's DiseaseDementiaDiabetic RetinopathyDrug Induced Liver InjuryHeart FailureHeightHerpes ZosterHDLHypothyroidismFibromyalgia in an RA cohortLipidsMidSouth CDRN CHD AlgorithmMultiple SclerosisPeripheral Arterial DiseaseRed Blood Cell IndicesRheumatoid ArthritisSevere Early Childhood ObesitySleep Apnea PhenotypeType 2 Diabetes – 2 phenotypesWarfarin dose/responseWhite Blood Cell Indices
CHARGE Consortium Working GroupsAdiposity Hemostasis Atrial Fibrillation/PR-Interval Inflammation Aging and Longevity Insulin Like Growth Factor Blood Pressure Lipids CHARGE-S Analysis & Bioinformatics Microbiome CHARGE mtDNA+ Musculoskeletal Depression Natriuretic Peptides Echocardiography: EchoGen Neurology (specific subgroups on Stroke, Dementia…)Educational Attainment Nutrition EKG Pharmacogenetics Endothelial Function Physical Activity Entrepreneurship Pulmonary Epigenetics Renal Exome Chip Repro-GEN Eye Retina Sex Hormone Family Studies Sleep Fatty Acid Subclinical / CHD Gene Expression Sudden Cardiac Arrest Gene-Lifestyle Interactions Telomeres Glycemia/Diabetes Thyroid Function Hearing loss Tonometry Heart Failure Venous Thromboembolism Hematology
Evolution of Phenotyping:Current ApproachesManual Algorithm DevelopmentFully Automated PhenotypingBenefitsInput from clinical expertsAccelerated phenotype development: saves time and increases scaleLimitations Creating list of informative variables is time-consuming Manual medical record review causes a bottleneck Reliance on the data to direct phenotyping instead of clinical experts
Our Vision for Phenotyping in MVP:A New ApproachSemi-automated phenotyping combines features of manual and automated phenotype development
Overview of semi-automated method for developing EMR phenotypesLiao KP, Cai T, Gainer V, et al. Electronic medical records for discovery research in rheumatoid arthritis. Arthritis care & research. Aug 2010;62(8):1120-1127.Step 1: Create “Phenotype X Mart”Step 2: Develop gold standard training setStep 3: Identify variables important for predicting Phenotype X Codified + narrative data (extracted using natural language processing)Step 4: Develop algorithm Logistic regression with LASSO Step 5: Validation
MVP System Architecture 19 Access Authorization by Governance System Vendor Molecular Lab Query Mart Query Portal Analysis Environment Consent Manager Study Mart Study Mart Study Mart Data WarehouseHonest Broker VA Non VA Clinical Data NDI, CMS Survey Data Molecular data Researcher
Summary of Unmet NeedsData mappingData cleaning, cataloguingAutomating the curation pipelineLibraryViewing / moving dataVersioning controlTracking uses of dataRetaining