UC-Rex: A model for cross-campus collaboration and data sharing UC-CSC Meeting San Francisco August 4, 2014 Doug Berman UCSF and Ayan Patel UCLA ucrex.org Today’s discussion: • • • • What is UC-ReX? Who is involved and what did we do? What technology was involved? How we worked together: • Project structure • Sponsorship, leadership, governance, workgroups and coordination • Results and outcomes • Results • What we learned about working together 2 ucrex.org UC- Research Exchange (UC-ReX) Background • During the past few years UC Medical Campuses have made significant investment in Electronic Health Records • Research is a key mission at each campus • We recognize the power in working together in research • Medical campuses may share data in order to achieve a large population for our work 3 ucrex.org UC-ReX Goals • Five-Year Goal: Enable researchers and quality improvement specialists to query and analyze clinical data collected at the point of care at all UC medical campuses for research or quality improvement purposes under a common crossinstitutional IRB approval process (Trust/Rely) and in a manner that preserves privacy. 4 ucrex.org UC-ReX Sponsors • UC Office of the President Funding for Cross-UC Data Sharing $5 million/5 years (July 2011- 2016) • UC - BRAID Biomedical Research Acceleration, Integration & Development • Local CTSA’s (Clinical Translational Science Awards) • Campus CIO’s 5 ucrex.org Demonstration : UCReX Data Explorer (SHRINE) https://ucrexi2b2.ucsf.edu/ ucrex.org Governance Structure Working Groups Technology Strategy (Lisa Dahm, UCI) UC BRAID PI from each UC CTSA Budgetary oversight Review quarterly status reports UC ReX Steering 1 voting + 1 nonvoting member from each UC Simple majority Rotating Chair http://www.ucbraid.org/informaticsmdashuc-rex.html Technical Implementation (Lisa Dahm, UCI) Data Harmonization (DaveraGabriel, UCD) User Support (Mini Kahlon, UCSF) Data Quality (Doug Bell, UCLA) ucrex.org Working Groups Technology Strategy Technical Implementation Data Harmonization User Support Data Quality • Specify use cases • Provide functional gap analysis • Propose technology roadmap • Ensure that infrastructure & critical software are deployed & maintained • Ensure semantic interoperability • Oversee data quality • Define processes, create SOPs • Coordinate user training & support • Design UCReX Website, roll-out pilot • Review completeness and consistency of data across campuses • Identify opportunities to improve data quality ucrex.org Technology Strategy • Identify use cases • Review and select technologies and partners • Sets technical direction 9 ucrex.org UC-ReX Use Cases Support Clinical Trials and Recruitment • Clinical studies are challenging • It is difficult to identify and recruit appropriate research subjects. • Clinical studies may take years to recruit sufficient populations to support conclusions; many fail for lack of patients Performing studies in larger populations may make many studies possible. Quality comparisons among sites to identify best practices Research Questions on Retrospective data 10 ucrex.org Technical Implementation • Open source software developed at Harvard • i2b2 – Informatics for Integrating Biology and the Bedside • • Scalable informatics framework that enables clinical researchers to use existing clinical data for discovery research https://www.i2b2.org/ • SHRINE - Shared Health Research Information Network • • System for enabling clinical researchers to query across distributed hospital electronic medical record systems https://open.med.harvard.edu/wiki/display/SHRINE/SHRINE++Basics • 1 Proxy Server, 2 Application Servers, 1 Database Server • Proxy Server – Apache • i2b2 Application Server – jBoss • SHRINE Application Server – Tomcat • Database Server – Oracle/SQL Server • UCSF, UCLA, UCD – Oracle • UCI, UCSD – SQL Server 11 ucrex.org What’s at each site • Unique ETL - Moves data to harmonized dataset • i2b2 database • Web-site for queries • Management agent • Local provisioning for access • Local support for users 12 ucrex.org UCReX Network Topology Central Node (located at UC Davis-MC) Data remain at each UC • The Central Management Node (CMN) is a webapp deployed to complement SHRINE – as agent or manager • Provides a central point for monitoring nodes and gathering information ucrex.org UC-Rex Architecture 14 ucrex.org Data Harmonization • Develop/determine ontology to reference source data from each site • In order to query all sites, we all must speak the same language • 5 different EMR implementations (4 sites with Epic EMR – Not much help) • • • Clinical workflows implemented differently Different modules at different stages of deployment What about legacy data? • Types of Medical Data • Demographics – Local source data mapped to various standards from CDC, WHO, ISO, HL7 • Diagnoses and Procedures – ICD9 Standard Terminology used consistently across all sites for billing/finance • Lab Results • • Different laboratories with different equipment Different reference ranges and units for the same lab • Medications • Maintaining Consistent Ontology • GitHub repository set up to ensure each site has the same ontology 15 ucrex.org User Support • • • • Communicate with local users Develop websites, documentation Provide training and support Develop support protocols and communication among site-based support groups • Implement processes for sharing identified data (IRB approvals, data sharing agreements, request process and secure delivery of results) • Receive feedback from use community 16 16 ucrex.org Data Quality • Discovery data anomalies by querying the i2b2 databases • Look at trends of counts of data types by year from each site • View distribution of demographics • One site had a unreasonably high percentage of a certain race, upon investigation it was discovered that that race was used as a default value • Lab Results • Look at medians and means for each lab type • Discovered some lab units were not converted to the appropriate unit agreed upon in Data Harmonization • Continue to slice and dice data and look at it from different perspectives • Investigate potential issues - feedback discoveries to Technical Implementation and Data Harmonization 17 ucrex.org Lessons learned • Balance Project Goals • Research project versus IT project to support research • Focus on delivery • Expect institutional differences • Infrastructure, organization, approval and change processes will differ at each institution • Time lines needed to adapt to diversity among institutions 18 ucrex.org Lessons learned (continued) • Project management was key • Program management - central coordination and decision making • Site level – planning resources and deliverables • Virtual work – Conference calls and screen shares are effective • Team members from each site work together directly • ‘Perfect is the enemy of good’ 19 ucrex.org UCReX –Team UCD Kent Anderson UCSD Elizabeth Bell Nicholas Anderson UCI Davera Gabriel Samuel Morley Hyeon-eui Kim Lucila Ohno-Machado Travis Nagler Paulina Paul Yi-Cheng (Andrea )Hwang UCSF Douglas Berman Lisa Dahm Robert Hink Ray Pablo Bhuwan Karki Dana Ludwig UCLA Douglas Bell Vijay Rayanaker Robert Follett Kimberly Romero Leslie Yuan Ayan Patel Marianne Zachariah Program Manager Lattice Armstead 20 ucrex.org Questions 21 ucrex.org