AGBT Precision Medicine 2016 Cohort Indentification
1. 9
DATE-SHIFT
P A T I E N T P R O X Y I D M A P P I N G
ProxyID
MRN
Name
Gender
Date of birth
Social security
Address
etc.
ProxyID
Assigned Date-shift
( 1-365 )
Researcher
makes request
Researcher receives
approved patient data
Central Institutional
Review Board
approves
request
NYGC
identifies
cohort Select 1 Doctor
or Hospital per
patient
Contact Patient
Patient consents
ProxyID
De-identified MR:
Demographics
Visits
Observations
Diagnosis
Procedures
Medications
(prescribed/filled)
Medicaid
ProxyID → CDRN-ID
Patient-match records
across sites
Data quality assurance
Example:
Researcher wants to find all
ALS patients in NY under 40ProxyID
Site Code
Healthix
eMPI
HISTORICALDATA
1
2
3 Patient–ProxyID Map
5 Data Transformation
6 Data Loading
DATE-SHIFTED
EHR
4
Patient-Matched
Data
Solving the Problem of
CohortIdentificationforPrecisionMedicineResearch
1 New York Genome Center, 2 Columbia University Medical College
3 Montefiore Hospital, 4 Weil Cornell College of Medicine,
5 Biomedical Research Alliance of New York (BRANY IRB),
6 Clinical Directors Network, 7 Healthix
Toby Bloom1
, George Hripsak2
, Parsa Mirhaji3
, Alex Low4
, Cristyn Kells1
, Dorian Leary1
,
Karen Hackett1
, Clint Beilman1
, Fred Criscuolo1
, Nina Lapchyk1
, Shailu Gargeya1
,
Thomas Check7
, Jonathan Tobin6
, Curtis Cole4
, Raffaella Hart5
, Rainu Kaushal4
One of the limiting factors in successfully performing large precision medicine genomics studies is
identifying enough participants for a study. The problem is particularly acute for rare conditions.
In New York City, a number of large hospitals have joined forces to build a centralized, deidentified
medical records repository as part of the NYC CDRN (Clinical Data Research Network). A patient’s
records are merged across multiple providers and patients are made re-contactable. After a
DATA COLLECTION COHORT SELECTION
1
2
5
6
7
CDRN
IRB
D
a
ta
b
a
se
7 Million
Patients
8
RESPONSE 4
3 REQUEST
Collect consent
Notify all hospitals of patient consent
10
Hospitals deliver consented data
H
ospitals
CUMCDOCTORS
MONTEFIORE
WCMC
NYPCUMC
MSHSNYPWCMC
NYULMC
LFHC
11
Partially funded by the Patient Centered Outcomes Research Institute
cohort has been identified, the CDRN will obtain approval from a central IRB, arrange to
recontact the patients and obtain consent for the proposed study, and provide the needed
information to the PI. As part of the Patient-Centered Outcomes Research Network, it is possi-
ble to identify cohorts across 11 such CDRNs nationally, though each has its own identification
process. We show here how the NYC-CDRN manages the process: