Transcript Introduction

Shared Health Research Information Network
Three axis for rapid production grade deployment:
1. POLICY
2. TECHNOLOGY
3. RESEARCH SCENARIOS
Andrew McMurry
Sr. Research Software Developer
Harvard Medical School Center for BioMedical Informatics
Children's Hospital Informatics Program at Harvard-MIT HST
Andrew_McMurry(@) hms.harvard.edu
https://catalyst.harvard.edu/shrine
Outline of topics covered
Policy

History of success cross-institutional IRB agreements
 Integrated health care entities
 Across independent HIPAA covered entities
Technology



SHRINE Architecture
Current status and roadmap
Development Challenges and Opportunities
Intended future translational research scenarios



for Translational Research Requiring Human Specimens
for Population Health Surveillance
for Observational Studies of Genetic Variants
History of cross-institutional IRB agreements
Integrated health care entities

Partners RPDR  i2b2 Clinical Research Chart
Everyday patient encounters  huge research cohorts
Shawn Murphy et all (wont steal their thunder here)

Centralized Research Patient Data Repository shared among


Massachusetts General Hospital (MGH),
Brigham and Women's Hospital (BWH),
Faulkner Hospital (FH),
Spaulding Rehabilitation Hospital (SRH), and
Newton Wellesley Hospital (NWH)
History of cross-institutional IRB agreements
http://spin.chip.org/irb.html
Across independent HIPAA covered entities

SPIN: Federated query over locally controlled de-identified databases

Distributed pathology database shared by
Brigham & Women's Hospital*
Beth Israel Deaconess Medical Center*
Cedars-Sinai Medical Center
Dana-Farber Cancer Institute*
Children's Hospital Boston*
Harvard Medical School*
Massachusetts General Hospital*
National Institutes of Health
National Cancer Institute
Olive View Medical Center
Regenstrief Institute
University of California at Los Angeles Medical Center
University of Pittsburgh Medical Center
VA Greater LA Healthcare System
* Participate in live “Pathology Specimen Locator” collaboration
History of cross-institutional IRB agreements
SHRINE approach : leverage has worked in the past


Secure IRB approvals for I2b2 local database at each site
Separate set of approvals for federated queries across sites
SHRINE governance principles



Hospital Autonomy: each site remains in control over all disclosures
Patient privacy: no attempts to re-identify patients
Non compete: no attempts to compare quality of care across sites
SHRINE Technical Architecture
Bird’s Eye View

Leverage local i2b2 deployments

Broadcast queries and aggregate responses across autonomous sites as
if they were “one clinical data warehouse”

There is no central database

Connect sites in a peer-to-peer or hub-spoke fashion
SHRINE Technical Architecture
Technical Architecture
Architecture, “cell” view
2009 deliverable
Architecture, sequence diagram view
SHRINE Technical Architecture
Current

Status
Harvard Effort
Prototype system running live at Harvard across BIDMC, Children’s,
and Partners representing both BWH and MGH.
Uses 1 year of real patient data
Demographics and diagnosis
Under tight IRB control
SHRINE Technical Architecture
Current Status

National Effort: west coast partners
University of Washington
UCSF
UC Davis
Recombinant
End-to-End Demo March 18th (3 week turn around time)
SHRINE Technical Architecture
Current Status

National Effort: sleep study partners
Case Western Reserve Institute
University of Washington-Madison
Marshfield Clinic
(potentially others as well)
I2B2 users interested in using SHRINE for sleep studies
SHRINE Technical Architecture
I2b2 single site query demo

http://I2b2.org/software
SHRINE multi-site demo

http://cbmi-lab.med.harvard.edu:8443/i2b2
SHRINE Technical Architecture
Timeline and Roadmap

By end of 2009, Harvard SHRINE queries for aggregate counts
Demographics + ICD9 Diagnosis

Current work
Polishing demostration software for relase
Medications and Labs

Next Steps
Browseable random LDS datasets
Downloadable LDS
No plans for PHI
Development Challenges and Opportunities
1. Grid computing makes multi-threading look
simple by comparison



Politically impossible to send patient data to each ‘grid’ node
Grid computing and federated queries are VERY different
Pre-processing can be used effectively as shown in our use cases
2. Open Source strategy
1. Writing plug-ins for the SHRINE network
Development Challenges and Opportunities
1. Grid computing makes multi-threading look simple by
comparison
2. Hosted retreat to address Open Source strategy






Harvard CTSA, CHIP, I2B2, Partners, DFCI, private companies
Science Commons, jQuery
Actively launching an open source portal
Test driven development with continuous integration
Release early release often
All milestones measured by what we can get IRB approved and
deployed with real clinical data
3. Writing analysis plug-ins for the SHRINE network
Development Challenges and Opportunities
1. Grid
computing makes multi-threading look simple by
comparison
2. Open Source strategy
1. Writing analysis plug-ins for the SHRINE network
•
Using I2b2 Java Workbench
Using I2b2 Web Querytool
•
By pre-processing results when required for patient privacy *
•
(Shawn Murphy et all)
(Griffin Weber et all)
* http://www.jamia.org/cgi/content/abstract/14/4/527
SHRINE: Intended Investigation Use Cases
For translational studies requiring human specimens
For Population Health Surveillance
For Observational Studies of Genetic Variants*
Examples shown here reflect current projects which will use the
SHRINE infrastructure
for Translational Research Requiring Human Specimens
NCI
vision
2001:
Vast collections of human specimens and relevant
clinical data exist all over the country, yet are
infrequently
shared
for
cancer
research.
Challenges:



How to link existing pathology systems for cancer research?
How to ensure patient privacy in accordance with HIPAA?
How to encourage hospital participation?
Availability
Millions of Paraffin Embedded Tissues
Smaller
Collections
of
Fresh
/
Frozen
Tissues
for Translational Research Requiring Human Specimens
Shared Pathology Informatics Network






National prototype including HMS, UCLA, Indiana, UPMC, …
Live Production instance at HMS including 4 hospitals
Created Open Source Tools
caBIG adopted caTIES from SPIN
Influenced Markle’s Common Framework federated query
TMA construction using specimens from four sites
http://spin.chip.org
for Translational Research Requiring Human Specimens
for Translational Research Requiring Human Specimens
For Population Health Surveillance
For translational research requiring human specimens
For Population Health Surveillance




Geotemporal cancer disease incidence rates
Seasonal infectious diseases such as influenza
Disease flares such as Irritable Bowel Disease (IBD)
Other use cases exist, these are the ones under concentrated study
For Population Health Surveillance:
disease outbreaks
For Population Health Surveillance:
seasonal influenza
http://aegis.chip.org/flu
For Population Health Surveillance: pharmacovigilance
http://www.plosone.org/article/info:doi%2F10.1371%2Fjournal.pone.0000840
SHRINE: Intended Investigation Use Cases
For translational research requiring human specimens
For population health surveillance
For Observational Studies of Genetic Variants*




High throughput genotyping
+
High throughput phenotyping
+
High throughput sample acquisition
=
Orders of magnitude
Faster to obtain huge populations for genomic studies
Cheaper
*Courtesy of Zak Kohane
For observational studies of genetic variants
High throughput sample acquisition

CRIMSON
High throughput genotyping

CRIMSON samples  SNP arrays
High throughput phenotyping

Natural language processing “smoking status”
Orders of magnitude



Faster to obtain huge populations for genomic studies
Cheaper
“disruptive technology”
Lynn Bry, MD, PHD et all
Summary of topics covered
Overcome statistical noise and reproducibility with large
patient populations
Policy

History of cross-institutional IRB agreements
Technology



Architecture
Current status and roadmap
Development Challenges and Opportunities
Intended future translational research scenarios



for Translational Research Requiring Human Specimens
for Population Health Surveillance
for Observational Studies of Genetic Variants
Acknowledgements: Core SHRINE team
Zak Kohane
Griffin Weber
Shawn Murphy
Dan Nigrin
Ken Mandl
Sussane Churchill
Doug Macfadden
Matvey Palchuck
Andrew McMurry
(SHRINE Lead / HMS)
(HMS CTO / bidmc)
(I2B2 CRC / partners)
(Children’s CIO)
(Public Health Use Cases/ CHIP IHL)
(I2B2 Executive director)
(HMS CBMI IT Director)
(Ontology Lead / HMS)
(Architect / HMS)
Could give an entire talk on all the collaborators, multi-institutional
effort. Asking forgiveness from those not listed
Acknowledgements: Core SPIN team
Zak Kohane
(SPIN PI / HMS)
Frank Kuo
(PSL Program Director / BWH)
(PSL Pathologist / MGH)
(PSL Pathologist / BIDMC)
(PSL Pathologist / Children’s)
(PSL Developer / BWH )
(Biosurviellance PI/ Children’s)
(Biosurviellance Dev Lead / Children’s)
(SPIN Developer/ HMS)
(SPIN Developer / NCI at HMS)
(SPIN Developer / HMS
John Gilbertson
Mark Boguski
Antonio Perez
Mike Banos
Ken Mandl
Clint Gilbert
Greg Polumbo
Ricardo Delima
Britt Fitch
http://spin.chip.org/community.html
Acknowledgements: Core I2b2 team
https://www.i2b2.org/about/structure.html
Thank You
 http://catalyst.harvard.edu/shrine
 Andrew_McMurry
(@) hms.harvard.edu