B RIS The Biomedical Translational Research Information System (BTRIS) James J. Cimino Chief, Laboratory for Informatics Development National Institutes of Health Clinical Center Senior Scientist, Lister Hill Center for Biomedical Communications National Library of Medicine Adjunct Professor, Department of Biomedical Informatics Columbia University College of Physicians and Surgeons September 12, 2008

B RIS Learning Objectives 1.An introduction to the compilation and integration of clinical and research data from multiple disparate sources 2.Understanding the authorization issues related to reuse of patient clinical data 3.Understanding the terminology issues related to the reuse of coded clinical and research data 4.Familiarity with the approach being taken at the National Institutes of Health to collect, integrate, and code clinical and research data into a single repository, for authorized resuse in biomedical research.

B RIS No Relevant Financial Relationships with Commercial Interests

B RIS The Clinical Center, 1981

B RIS What Was the Problem? Patients in the NIH Clinical Center were on research protocols, but… …their data were in a clinical information system Needed to extract individual data for analysis Needed cross-patient queries for additional analysis

B RIS Patient Information Graphics System (PIGS) Integration of a desktop computer (pre-PC) to mainframe clinical information system (TDS) Download vital sign and drug administration data Various graphical displays SCAMC, 1981

B RIS

The Columbia University Medical Center Clinical Data Warehouse Clinical data repository to collect data from ancillary systems Coded with the Medical Entities Dictionary

B RIS MED Structure Medical Entity Laboratory Procedure CHEM-7 Plasma Glucose Test Laboratory Specimen Plasma Specimen Substance Sampled Part of Has Specimen Event Laboratory Test Diagnostic Procedure Substance Measured Glucose Plasma Anatomic Substance Bioactive Substance Chemical Carbohydrate

B RIS Using the MED for Summary Reporting Plasma Glucose Test Serum Glucose TestFingerstick Glucose Test Lab Test Intravascular Glucose Test Lab Display Chem20 Display

B RIS Clinical data repository to collect data from ancillary systems Coded with the Medical Entities Dictionary The Columbia University Medical Center Clinical Data Warehouse Back end for clinical information systems (CIS, WebCIS, PatCIS, PalmCIS, QingCIS, MendonÇIS…) Reorganized for use as Clinical Data Warehouse

B RIS How often are patient with the diagnosis of myocardial infarction started on beta blockers?

B RIS Find all patients with Diagnosis in class MYOCARDIAL INFARCTION AND with Medication in class BETA BLOCKER How often are patient with the diagnosis of myocardial infarction started on beta blockers? Find all patients with Diagnosis in class MYOCARDIAL INFARCTION

B RIS The Clinical Center, 2008

B RIS The Clinical Center, 2008 Patient data stored in EHR (Eclipsys) Need to extract individual data for analysis Need cross-patient queries for additional analysis Data may require transformation: –De-identification and Re-identification –Indexing –Aggregation by time –Abstraction by classification –Conversion to relevant concepts

B RIS What is BTRIS? Biomedical data –Research data collected using clinical information systems –Clinical data collected using clinical information systems –Research data from research information systems –Non-human data Reuse of data to support translational research Hence: Biomedical Translational Research Information System

B RIS The NIH, 2008

B RIS What is BTRIS? BTRIS

B RIS What is BTRIS? BTRIS

B RIS OntologyOntology Data Acquisition Processes Coding Indexing De-Identifying Permission Setting BTRIS What is BTRIS? Data Repository Data Retrieval Functions Authorization Subject-Oriented Cross-Subject Re-Identification NLP Data Analysis Tools Subject Recruitment Hypothesis Generation Hypothesis Testing

B RIS BTRIS Issues Data sources Data model integration Research Entities Dictionary Access policies User requirements User access and user Interface Terminology-based queries

B RIS Data Sources Order entry system Ancillary systems Institute and Center (IC) systems Individual researchers’ systems Notebooks

B RIS Data Model Integration Events and details Entity-relation vs entity-attribute-value Denormalization

B RIS Research Entities Dictionary Apelon’s Terminology Development Editor NCI/caBIG Thesaurus

B RIS Research Entities Dictionary (RED)

B RIS Apelon’s Terminology Development Editor NCI/caBIG Thesaurus Research Entities Dictionary Content Organization

B RIS Access Policies Privacy Act, not HIPAA Policy Working Group Intellectual property vs. public domain Identifiers Unlinked, coded data

B RIS BTRIS Data Storage Policy Inactive Protocols Active Protocols 1. All clinical and research data, whether from active or inactive protocols, are stored in BTRIS in a timely manner.

B RIS BTRIS Data Use Policy Inactive Protocols Active Protocols PIs 2. PIs can see data collected on their patients as part of their active protocols

B RIS BTRIS Data Use Policy Inactive Protocols Active Protocols PIs 3. PIs may indicate that some of their data should be blocked to avoid unmasking their study subject assignment. Blocked By PI

B RIS Identifiers Coded Data Active Protocols Inactive Protocols BTRIS Data Use Policy 4. It is helpful to consider data to consist of identifiers vs. other (“coded”) data linked to those identifiers. PIs can get to coded data in their own active protocols, but not those in inactive protocols. PIs Blocked By PI

B RIS Identifiers Coded Results Coded Events Active Protocols Inactive Protocols BTRIS Data Use Policy 5. It is also helpful to consider that coded data can be further divided into recorded events (such as a medication order or a lab result report) and results (such as the medication ordered or the value of the test result).

B RIS Identifiers Coded Results Coded Events Active Protocols Inactive Protocols BTRIS Data Use Policy PI Only PI Only Restricted By PI Sharable Through PI 6. Identifiers in active protocols are restricted to PIs but other data are generally be sharable with some PI control. The PI may designate some events, and perhaps their data, as completely restricted while the protocol is active.

B RIS Identifiers Active Protocols Inactive Protocols Unencumbered By PI BTRIS Data Use Policy Coded Results Coded Events PI Only PI Only Restricted By PI Sharable Through PI 7. When a protocol becomes inactive, the identifying information is inaccessible, but the other (coded) data (events and results) become generally available. No Access

B RIS Identifiers Active Protocols Inactive Protocols Collaborative with PI Unencumbered By PI BTRIS Data Use Policy Coded Results Coded Events PI Only PI Only Restricted By PI Sharable Through PI 8. A PI may restrict use of some data from inactive protocols, for example requiring other users of the data to work in collaboration with the PI. No Access

B RIS Identifiers Active Protocols Inactive Protocols Collaborative with PI Unencumbered By PI BTRIS Data Use Policy Coded Results Coded Events PI Only PI Only Restricted By PI Sharable Through PI 9. Data may move back to be considered active if a PI (not necessarily the original PI), gets approval to use the data, including (as per the IRB), identifying information. No Access

B RIS Identifiers Active Protocols Inactive Protocols Collaborative with PI Unencumbered By PI BTRIS Data Use Policy Coded Results Coded Events PI Only PI Only Restricted By PI Sharable Through PI 10. To summarize, PIs have access to data in their own active protocols, subject to any blocking that may be required. No Access PIs

B RIS Identifiers Active Protocols Inactive Protocols Collaborative with PI Unencumbered By PI BTRIS Data Use Policy Coded Results Coded Events PI Only PI Only Restricted By PI Sharable Through PI 11. The NIH community will have access to all data, except identifiers, with restrictions on data (active or inactive) by PIs. Restrictions may differ for events versus results. No Access NIH Community

B RIS User Requirements 2006 study of user requirements

B RIS User Requirements Provide Medication lists at time of patient encounters. Include drug diaries for inpatient, outpatient and in-between encounters in the patient medical record. Include all chemo and non-chemo drugs from CRIS and IC systems Provide Medication administration documentation (drug diaries) with times as part of patient record Provide ability to compare patient results, Medication Administration Records between dates and/or encounters Provide drug randomization info, compliance records and drug accountability info for all investigational, study and prescription drugs Provide all Clinical Center lab results with times of specimen draws Provide external lab results Provide archival images Provide demographics data including age, BMI, race, gender, contact info, etc Provide access to genomics and bio-markers data Provide cumulative blood volumes, research drugs and radiation for subject over a given period Provide ability for ICs to feed expanded diagnosis/problem lists Provide searching and filtering patients' data by all diagnosis, tests, procedures, protocol & protocol classifications Standardize medication and lab test codes. Provide integration of adverse events data in the data warehouse, Provide integration of protocol census, status and subject accrual tracking data from Protrak in the data warehouse Provide ability to attribute different events to protocols, viz., consent signed, protocol activated, orders, observations, adverse events, etc Provide integration of staff, patient and user index data across source systems in data warehouse. Provide original informed consent, and updated consents for re-contact of patients for research for all protocols. Provide searchable consents and image of consents in database. Provide answers to: –Can tissues be used for cancer/genetic research, other research, germ line testing –Can patient be re-contacted for questions? Provide single patient amendments Provide access and track biological specimen data Provide access to Appointment Data Provide “Review of Systems” info for each patient visit. Provide patient de-identification services Standardize Units Of Measure

B RIS 2006 study of user requirements User Requirements Investigator User Group Demos: –To user group –One-on-One –Town hall meeting

B RIS User Access/User Interface Canned reports (e.g., recruitment for IRB) Reports with prompts (e.g., sets of lab results) Ad hoc queries (whatever you want!) Selected Business Objects

B RIS Terminology-Based Queries Hripcsak G, Allen B, Cimino JJ, Lee R. Access to Data: Comparing AccessMed to Query by Review. Journal of the American Medical Informatics Association; 1996;3(4):

B RIS

Hripcsak G, Allen B, Cimino JJ, Lee R. Access to Data: Comparing AccessMed to Query by Review. Journal of the American Medical Informatics Association; 1996;3(4): Terminology-Based Queries Business Objects prompt with hierarchy

B RIS

Hripcsak G, Allen B, Cimino JJ, Lee R. Access to Data: Comparing AccessMed to Query by Review. Journal of the American Medical Informatics Association; 1996;3(4): Business Objects prompt with hierarchy Terminology-Based Queries Ancestor-descendant join with filtering

B RIS Ancestor-Descendant Join with Filtering Start with parent-child table Transitive closure produces ancestor- descendant table Designate subset of “searchable” classes (ancestors) Display flat list (no tree walk) Let user select desired classes Query selects ancestor-descendant rows and joins data tables based on descendants

B RIS Status Modeling source data –Clinical Research Information System –Clinical Data Repository –Crimson (NIAID)

B RIS

Modeling source data –Clinical Research Information System –Clinical Data Repository –Crimson (NIAID) Status Unified BTRIS data model

B RIS

MRN BTRIS ID Identifiers 1234 B00001 John… B00002 Mary… B00003 Umberto... BTRIS ID Other Data B00001 Sodium 132 B00002 Sodium 143 B00003 Sodium 138 BTRIS Identified Data MRN Identifiers Other Data 1234 John…. Sodium Mary…. Sodium Umberto... Sodium 138 BTRIS ID Other Data B00001 Sodium 132 B00002 Sodium 143 B00003 Sodium 138 Deidentified Data Gender Avg Data Male Sodium 135 Female Sodium 143 Other Sodium 218 PIs NIH Researchers Patient Recruitment Others

B RIS

Modeling source data –Clinical Research Information System –Clinical Data Repository –Crimson (NIAID) Unified BTRIS data model Status Database versions –Specific for principal investigator –Scrambled data for general use Business Objects demonstration

B RIS

BTRIS in the Research Process Hypothesis Generation Write Protocol Scientific Review Analysis of Resources Required Recruitment Patient Accrual Data Gathering And Analysis Publications IRB Approval B RIS

Next Steps Rebuild database in Oracle Total historical data sets from CDR, CRIS, CRIMSON Live interfaces with Lab, Radiology, CRIS and CRIMSON Continue to collect user feedback Develop report sets BTRIS July 2009

B RIS More Issues Patient ownership of de-identified data If genomic data are included, can we truly de-identify? Is protocol number a patient identifier? How important is protocol attribution? What data should not be shown to PIs? What is the extent of data control by PIs?

B RIS Moving Forward Despite centuries of clinical research, data are still fragmented Sophisticated terminology is the key to reuse of disparate data Policy issues are bigger than technical issues

B RIS BTRIS Will: Be the preferred system to analyze NIH clinical and non-clinical data Aggregate and standardize disparate and isolated data sets Automate and streamline processes that are traditionally manual and cumbersome Prioritize data sources and functionality based on needs of user community Future of BTRIS: today the NIH, tomorrow the world?

B RIS btris.nih.gov