1. Rapid Deployment and Adoption of Health Information Technology for Real Time Biosurveillance Primary support: NCI, NLM, CDC, and the DF/HCC

2. Outline  Overview of the SPIN Architecture  SPIN National Demonstration for Cancer Research  Lessons Learned for Rapid National Deployment  SPIN enabled Biosurveillance  Summary

3. SPIN addresses (3) pervasive issues 1. Linking existing patient care systems 2. Protecting patient privacy 3. Ensuring that hospitals remain in control … and has been deployed for: Cancer Research Requiring Human Specimens Public Health Biosurveillance With the potential for other clinical applications

4. How SPIN works 1. Link existing databases Extract from existing hospital systems Transform patient encounters into HIPAA-safe vocabulary Load into hospital controlled “SPIN peer” 2. Protect Patient Privacy per HIPAA De-Identified: Statistical Level query Limited: When authorized for individual cases PHI is rarely used, and only with permission from each IRB 3. Hospital Control No central governing body Remain in control over disclosures at all times

5. (1) Linking routine care systems Extract from routine care delivery systems  Databases or XML Transform free text reports  “Scrub” patient identifiers (per HIPAA)  Autocode into controlled vocabularies such as UMLS Load into the hospital controlled PEER database  Assign a randomly generated ID to each case

6. (2) Protecting Patient Privacy Increasing levels of investigator access commensurate with investigator need and hospital policy. SPIN has enabled Statistical QueriesLimited DatasetPHI Cancer Research Feasibility Studies Case SelectionSpecimens Biosurveillance Automated AnalysisInvestigationEmergencies Potential Applications Clinical AggregatesQA/QCInformed Care

7. (3) Hospitals remain in control Each hospital (Peer) chooses who to share with And what to share (Clinical Reports, ED feeds,.. )

8. SPIN Applications & Timeline  (2001 – Present) Translational Research (cancer)  (2006 – Present) Biosurveillance  (2007 - Present) Harvard CTSA

9. SPIN enabled Translational Research  Motivation: Vast collections of human specimens and clinical data exist all over the country, yet are infrequently shared for cancer research.

10. SPIN enabled Translational Research (1) Link Existing Pathology Databases  Extract Pathology Reports (coded XML) from each site  Scrub HIPAA identifiers & Autocode diagnosis for UMLS  Generate random ID and load into local SPIN peer (2) Patient Privacy: Increasing levels of investigator access  Statistical query (With UMLS and keywords)  Individual case query (De-Identified Path Reports)  Specimen request (Remap case UUID to accession #) (3) Hospital Control  De-Identified reports for Harvard researchers (ecommons)  PHI has never been released

11. SPIN enabled Translational Research

12.  Motivation: Vast collections of human specimens and clinical data exist all over the country, yet are infrequently shared for cancer research.  Results: National prototype including 14 sites Virtual Specimen Locator (all HMS hospitals) UMLS/scrubber adopted by caBIG (caTIES) Direct influence on Markle’s CFH Common Framework

13. Sites Participating in the National Demonstration 1.Brigham & Women's Hospital* 2.Beth Israel Deaconess Medical Center* 3.Cedars-Sinai Medical Center 4.Dana-Farber Cancer Institute* 5.Children's Hospital Boston* 6.Harvard Medical School* 7.Massachusetts General Hospital* 8.National Institutes of Health 9.National Cancer Institute 10.Olive View Medical Center 11.Regenstrief Institute 12.University of California at Los Angeles Medical Center 13.University of Pittsburgh Medical Center 14.VA Greater LA Healthcare System * Participate in ongoing “Virtual Specimen Locator” collaboration

14. Lessons Learned for Rapid National Deployment We have the principles, now here are the challenges:  For each participant and for each type of data exchange, we need to map heterogeneous databases  Building agreement to share: IRBs and the political will

15. Lessons Learned for Rapid National Deployment mapping heterogeneous DBs VS Start SMALL : Grow the number of common terms!

16. Lessons Learned for Rapid National Deployment Applying lessons learned: mapping heterogeneous DBs 1. Request for Capabilities (What is available?) 2. Availability limits scope of the vocabulary 3. What big questions can be asked with only a few data elements? Pathology: age, gender, collection, free text “diagnosis” Public Health: age, gender, location, free text “complaint” CTSA: age, gender, …………, free text mining 4. Parallel tracks: autocoding and standard vocabulary approach  Different low hanging fruit: diagnosis vs MRN 5. Quick End-To-End lifecyles  Question, development, research, new question

17. Lessons Learned for Rapid National Deployment Agreement to share: IRBs and political will  SPIN has addressed the “Distributed IRB” issue  Statistical level queries easy are OK by IRBs  Difficulty arises going to the next step HIPAA limited data set PHI

19. 2006 to present: SPIN enabled Biosurveillance  Motivation: Detect infectious disease outbreaks Track the spread of influenza Provide early warning signs of bioterrorism Re-identify patients as fast as possible during public health emergency

20. AEGIS: Automated Epidemiologic Geotemporal Integrated Surveillance

21. 2006 to present: SPIN enabled Biosurveillance (1) Link Routine Care Delivery Systems  Extract Emergency Department visits from each site Write a simple SQL statement & set routine extraction time  Transform Autocode free text of the “Chief Complaint” Anonymize (blur) home addresses & preserve spatial clusters  Load into local SPIN peer Generate the random linked identifier Peer database is encrypted ( key does NOT live on disk! )

22. 2006 to present: SPIN enabled Biosurveillance Preserving both spatial clusters & patient privacy  Calculate latitude/longitude for each patient address.  Skew the latitude/longitude with respect to the underlying population density.  Return only the anonymized coordinates during routine analysis  Return real address during emergencies

23. 2006 to present: SPIN enabled Biosurveillance (1) Link Routine Care Delivery Systems  Extract Emergency Department visits from each site  Anonymize patient addresses & Autocode “Chief complaint”  Generate random identifier and load into local SPIN peer (2) Patient Privacy: Increasing levels of investigator access  Statistical query (Automated routine analysis)  Limited Disclosure(Alarm Investigation )  Patient Re-Identification(Emergency Investigation) (3) Hospital Control  Which public health agencies do I trust? (CDC, DPH)  What do I want to allow in each investigation scenario?

24. 2006 to present: SPIN enabled Biosurveillance DPH/CDC is responsible for assigning Identity  Role Each Hospital is responsible for assigning Role  Policy

25. 2006 to present: SPIN enabled Biosurveillance One Example of Hospital Authorization Policies DPH/CDC is responsible for assigning Identity  Role Each Hospital is responsible for assigning Role  Policy

26. 2006 to present: SPIN enabled Biosurveillance

27. 2006 to present: Biosurveillance using SPIN & AEGIS  Motivation: Detect infectious disease outbreaks Track the spread of influenza Provide early warning signs of bioterrorism Re-identify patients as fast as possible during public health emergency  Results: Fulfills AHIC Biosurveillance Use Case One of four federally funded NHIN architectures Enables our existing biosurveillance application (aegis.chip.org)

28. Summary  SPIN addresses 3 pervasive issues  Linking routine care systems  Protecting patient privacy  Ensuring that hospitals remain in control  SPIN enables increasing levels of access:  Automated real-time biosurveillance  Alarm Investigation  Emergency Investigation  SPIN is decentralized and builds agreement for timely national adoption