Data Sharing: Biomedical Research Data versus Personal Health Data Belinda Seto, Ph.D. Deputy Director National Institute of Biomedical Imaging and Bioengineering

A secure, nationwide, interoperable health information infrastructure that will connect providers, consumers, and others involved in supporting health and healthcare. President Obama’s Health IT Vision

Improve coordination of care across providers Ensure that consumers’ health information is secure and confidential Enable consumers control and decision making Reduce risks from medical errors Enable evidence-based decision support systems Lower health care costs Nationwide Health Information Network

Healthcare is the Largest Sector of the Economy BCBS Medical Cost Reference Guide 2008 Medicare alone is currently 3.2 % of GDP and increasing rapidly!

National Health Expenditure (NHE), BCBS Medical Cost Reference Guide 2008 The NHE is projected to increase by more than 70 percent between 2007 and 2015, with healthcare expected to account for almost 20 percent of GDP.

The use of diagnostic imaging is increasing rapidly. Projected Growth in Imaging Procedures, US Market BCBS Medical Cost Reference Guide 2006

Aggregate Imaging Growth Was the Fastest for Physician Services:

Contract awarded to Radiological Society of North America: Image Sharing Network, PI: Dr. David Mendelson Grant awarded to Wake Forest University Grant awarded to University of Alabama Image Data Sharing: NIBIB Activities

Develop a patient controlled platform for medical image sharing Test a model using electronic keys to access data Integrate image data with electronic health records Include imaging facilities in rural and urban southeast U.S. Wake Forest Project

Establish regional health image exchange system among hospitals in Alabama Design a web accessible point for physicians and patients to view images Adopt standards of National Health Information networks, 2007 Initial targeting trauma patients Scalable Alabama Project

“HIT is not the end itself but a means to improving quality of health care”, Dr. David Blumenthal Data is the fuel that drives HIT Health Information Technology (HIT): The Means toward Better Care

Data Sharing to Support Better Decisions

Patient data need to be integrated and assessed to provide real-time, point-of-care information regarding the right care Improves clinical decision support with enriched data Develop algorithms to use comparative effectiveness findings to optimize outcomes Decision Support in Health Information Systems

Purpose: to improve health outcomes by providing evidenced-based information to patients and providers. Mandate: to conduct study of outcomes and to derive conclusions to inform medical choices/decisions. Comparative Effectiveness Research

EHR Health Information Exchange Clinical Decision Support Health Care Savings Cost Savings from Clinical Decision Support System

Open access: no personal health data, no identifiers Tiered access: data use agreements Sharing Research Data

“Restricted availability of unique resources upon which further studies are dependent can impede the advancement of research and the delivery of medical care. Therefore, when these resources are developed with PHS funds and the associated research findings have been published or after they have been provided to the agencies under contract, it is important that they be made readily available for research purposes to qualified individuals within the scientific community. This policy applies to grants, cooperative agreements, and contracts.” PHS Grants Policy Statement April 1994

NIH expects timely release and sharing of final research data for use by other researchers. NIH expects grant applicants to include a plan for data sharing or to state why data sharing is not possible, especially if $500K or more of direct cost is requested in any single year NIH expects contract offerors to address data sharing regardless of cost NIH Data Sharing Policy Effective with October 1, 2003 receipt date for NIH applications

NIH serves as central data repository A federated model with grantee institutions provide data repositories Data Sharing Models

Genome-wide association study GenBank Protein Cluster PubChem Many others at: NIH Central Data Repositories

Alzheimer Diseases and Neuroimaging Initiative

Major goal is collection of data and samples to establish a brain imaging, biomarker, and clinical database in order to identify the best markers for following disease progression and monitoring treatment response Determine the optimum methods for acquiring, processing, and distributing images and biomarkers in conjunction with clinical and neuropsychological data in a multi-site context “Validate” imaging and biomarker data by correlating with neuropsychological and clinical data. Rapid public access of all data and access to samples Goals of the ADNI: Longitudinal Multi-Site Observational Study

MCI (n= 400): 0, 6, 12, 18, 24, 36 months AD (n= 200): 0, 6, 12, 24 months Controls (n= 200): 0, 6, 12, 24, 36 months Clinical/neuropsychological evaluations, MRI (1.5 T) at all time points FDG PET at all time points in 50% 3 T MRI at all time points in 25% PIB sub-study on 120 subjects Blood and urine at all time points from all subjects; CSF from 50% of subjects 0, 1 yr, 2 yr (subset); DNA and immortalized cell lines from all subjects GWAS study Study Design

Goal is rapid public access of all raw and processed data Central repository for all QA’d MRI and PET [Laboratory of Neuroimaging, UCLA (LONI)] Clinical data base at UCSD is linked to LONI Databases- in the public domain, available to all qualified investigators Sample sharing-Resource Allocation Review Committee No special access Data Sharing & Publication Committee (DPC) – -ADNI Data Use Agreement Data and Sample Sharing

To identify common genetic factors that influence health and disease To study genetic variations, across the entire human genome, that are associated with observable traits To combine genomic information with clinical and phenotypic data to understand disease mechanism and prediction of disease To develop the knowledge base for personalized medicine Genome-wide Association Studies (GWAS): Purpose, Goals

All GWAS-funded investigators are expected to submit to the NIH data repository descriptive information, curated and coded phenotype, exposure, genotype, and pedigree data as soon as quality control procedures are completed at the grantee institutions. GWAS Data Sharing Policy

Serves as a single point of access to GWAS data To archive and distribute results from studies of the interaction of genotype and phenotype Provides pre-competitive data, no IP protection Encourages use of primary data from dbGP to develop commercial products or tests Database of Genotype and Phenotype (dbGP)

NIH does not possess direct identifiers of research participants; does not have access to link between data keycode and identifiable information; such information resides with the grantee institutions Research institutions submitting dataset must certify that an IRB and/or Privacy Board has considered and approved the submission Investigators must stripped the data of all identifiers before data submission Optional: Certificate of Confidentiality Protection of Research Participants: De-Identification

NIH expects specific discussion and documentation that participants’ genotype and phenotype data will be shared for research purposes through dbGP If participants withdraw consent for sharing individual-level genotype and phenotype data, the submitting institution will be responsible for requesting the dbGP to remove the data involved from future data distributions. Protection of Research Participants: Informed Consent

Requesters are expected to meet data security measures: physical security, information technology security and user training Requires signed data use certification: – Proposed research use of data – Follows local laws – Not sell data elements – Not share with individuals not listed in proposal – Provide annual progress reports Data Access

Open-access data includes: – Summaries of studies – Study documents, reports – Measured variables, e.g., phenotypes – Genotype-phenotype analyses dbGP Access: Two Levels

Requires varying levels of authorization Provides data on a per-study basis Controlled-access data includes: – De-identified phenotypes and genotypes for individual study subjects – Pedigrees – Pre-computed univariate association between genotype and phenotype dbGP: Controlled-Access

Requester must submit a Data Use Certification Access is granted by an NIH Data Access Committee Approval of proposed research use will be consistent with patient consent and data provider’s institutional terms and conditions Controlled-Access Data Requests

Discourages premature claims on pre-competitive information that may impede research Encourages patenting of technology for downstream product development, e.g., – Markers for assays – Drug targets – Therapeutics – Diagnostics Up to one year of exclusivity is allowed for the primary investigators to submit GWAS data analyses for publication Clock begins when the GWAS datasets is first made available to the NIH data repository Intellectual Properties?

“Data should be made as widely and freely available as possible while safeguarding the privacy of participants, and protecting confidential and proprietary data.” -- NIH Statement on Sharing Research Data February 26, 2003 NIH Viewpoint