Transcript TCRN Site Reports

TIES Cancer Research Network
Y3 Face to Face Meeting
U24 CA 180921
Session 1 Project Status Updates
October 9th, 2015
University of Pennsylvania
University of Pennsylvania
Abramson Cancer Center
TCRN Update
Michael Feldman, MD, PhD
2
Penn Team
• Michael Feldman, Assoc Prof Path & Lab Med – Site PI
• Joellen Weaver – Director SOM Biobank – Regulatory and
Admin
• Nate Digiorgio – Sys Admin and DBA
• Kevin Lux – Interface
• Tara McSherry – End user training and support, QA
• Fred Valdivieso – Tissue bank manager – honest broker
Database
• Latest code set running 5.3.1
• Records coded ~750K
– Active coding pipeline via daily data load from
Medview (home grown web portal for lab
results)
– Catching all Surg path cases from 3 hospitals
downtown hospitals at Penn Medicine
(65K/yr) – PAH was not online in 2014
Penn TIES Website
•
•
•
•
•
•
Penn site is up and running
User enrollment and authorization
Rules of the road
Documents AUA, MTA’s…
Link to instructional videos
Information related to TIES and our
instance
QA
• Review of QA results identified
– Automated search identified 300 records
– Manual scrub of records
• Missed name (Full name, Lname,
Fname) most common
• Reports were manually scrubbed and
returned into data set
Roll Out
• Started rolling out
– Surgical pathologists (5 to date)
– Remaining pathologists this year (25)
– Presentation to Cancer leadership very positive
• Deployed to Cancer center IT lead
• Start deploying to translational investigators
• SOM Biobank as honest broker
• Presentation to Pathology leadership
– Departmental division directors
– Anatomic Pathology Directors
– Pathology staff
Value
TIES study in prep to research
• Breast cancer researcher
– Looking for invasive breast cancer by
year since 2010
– Studying metabolism in cancer wanted
to make sure we had enough cases at
Penn to address questions using
retrospective case control study design
Value Radiology
Retrospective cohort study
• Find all cases of in situ lobular breast
cancer diagnosed on core biopsies who did
not have DCIS or invasive carcinoma
– What did follow-up resection show?
– Correlate with radiology finding
– Question is what group of LCIS needs
further surgical treatment when only
LCIS is diagnosed on a core biopsy
Structured + NLP
• How can we link structured data with TIES?
– Pathology
• Biobank – LIMS is labvantage, same at RPCI
– Oncology – tumor registry Metriq similar to
Pitt and RPCI
Virtual slides at Penn
• Behind firewall
– Aperio image server
– Aperio inmage scanner
– Moving to Philips platform
• Work with Pitt to develop tools to integrate
vSlides into TCRN TIES deployment
How can TIES be expanded
(more corpuses)
• Radiology
– Oncology
• Diagnosis
• Response Criteria (RECIST) – Lesion
changes, Lymph node size, bone lesion,
PET…
• Heterogeneity – structural (CT and MRI) and
genomic (PET)
• Integrated process research – precision
medicine
TIES at ‘Georgia Regents
University’ – F2F
Roni Bollag, MD, PhD
Nita Maihle, PhD
Jennifer Irons Carrick, BS
Sameera Qureshi, BS
Rahil Khan, BS
Who is GRU?
• MCG (1833–2011)
• GHSU (2011–2013)
• 2013: Georgia Regents University
= GHSU + ASU
one of only four public comprehensive research institutions
in the state of Georgia
Now AU
• Founded in 1828
– 13th -oldest medical school
– 8th -largest medical school
• Nine Colleges
• 9000 students (across campuses 240 medical
students/year)
• Supporting GR Health System
– 478-bed Georgia Regents Medical Center
– 154-bed Children's Hospital of Georgia
GRU Cancer Center
• The GRU Cancer Center has three expressions, one
laboratory-based, one clinic-based and one
education-based. Each of the GRU Cancer Center
facilities are designed with a purpose. Using a
modern, scientific approach known as evidencebased design along with the expert advice of cancer
patient advisors, our treatment facilities bring in
nature and sunlight and puts patients in charge in as
many ways possible. Our research facilities celebrate
innovation, combining form with function to offer
our scientists an environment for collaboration and
discovery.
GRUCC Areas of Research Focus
Cancer Immunology, Inflammation and Tolerance
Cancer Prevention & Control
Molecular Oncology & Biomarkers
Tumor Signaling & Angiogenesis
Our Team
Repository Side:
•Jinni Carrick – Biorepository Registrar
•Sameera Qureshi – Biorepository Lab Associate
•Rahil Khan – Biorepository Lab Assistant
•Denise Harper – Biorepository Admin. Assistant
•Roni Bollag –Biorepository Director
•Samir Khleif – Cancer Center Director
•Nita Maihle – Cancer Center Liaison for TIES
Our Team (con’t)
IT Side:
•Mia Jolly – Business Analyst/Project Coordinator
•Pankhil Patel – DBA
•Mike Hyrman – Encore consultant for data download
•Colleen Cain – Director Enterprise Application Systems
•Charles Busbee – Manager Database & Application
Administration
•Michael Casdorph – Associate VP Academic & Research
Technology
The GRU Biorepository:
 Provide MCG researchers access to high quality,
annotated specimens, to support basic and
clinical cancer research
 Shared resource for the Cancer Center,
important for its NCI designation process (P30
grant)
BRAG-Onc:
 Serves as coordinating hub and central
repository for the statewide repository network
BRAG-Onc
Mission
The goal of the Biorepository Alliance of
Georgia-Oncology Program (BRAG-Onc) is to establish a
state-wide process for the banking of tumor tissue and
blood samples from a population that reflects the
diversity of the cancer patient population of the state to
support basic and clinical research.
Sample Inventory* for the GRU/BRAG-Onc Biorepository:
GRU Biorepository Inventory
Collection Site
Total Donors Consented
Total Specimens
Collection Site
Total Donors Consented
Total Specimens
ALL (Acute Lymphocytic Leukemia)
28
215
LUNG (Lung)
AML (Acute Myelogenous Leukemia)
207
2736
LYMPH (Lymphomas: Hodgkin, Non-Hodgkin)
358
266
3851
1907
ANAL (Anal)
13
79
MCL (Mantle Cell Lymphoma)
8
62
BLAD (Bladder)
318
3841
MELA (Melanoma)
56
534
BONE (Sarcomas of Bone)
4
35
MESO (Mesothelioma)
0
0
BRST (Breast)
892
8062
MM (Multiple Myeloma)
313
3313
CERV (Cervical)
29
142
CLL (Chronic Lymphocyctic Leukemia)
84
1049
NASL (Head and Neck: Nasal Cavity, Sinuses, Nasopharynx, Oral Cavity,
Oropharynx)
153
1486
OPTH (Ophthalmic Sites)
3
16
OTH (Other (Spleen))
77
1133
OVAR (Ovarian, Fallopian, or Peritoneal)
210
1174
PANC (Pancreatic)
146
1392
PEDT (Pediatric Tumors)
2
24
PROS (Prostate)
421
4651
CML (Chronic Myelogenous Leukemia)
64
833
CNS (Central Nervous System: Brain and Peripheral NS)
347
3106
COLN (Colorectal)
375
3990
CTCL (Cutaneous T-Cell Lymphoma)
3
31
EDMT (Endometrial)
33
195
ENDO (Endocrine: Parathyroid, Adrenal, Thyroid)
545
5844
ESPH (Esophageal)
21
265
GALL (Gallbladder)
15
126
GAST (Gastric)
89
843
GDM (Gross Metastatic Disease)
0
0
GYO (Gynocologic: omentum, other)
274
1669
HEME (Hematopoietic disorders, MGUS, MDS, other)
290
2039
187
2416
HNO (Head and Neck: Other and unknown)
602
5619
INC (Incomplete, open)
0
0
INTS (Small Intestine)
31
208
HEPT (Hepatic, Biliary Tree)
LEUK (Leukemias)
51
696
LNO (Lymph Node, node biopsies)
149
1021
LRNX (Head and Neck: Larynx and Hypopharynx)
96
897
RENL (Renal and Ureter)
477
5857
SALV (Head and Neck - Salivary Glands, Paraglangliomas)
128
987
SARC (Soft Tissue Sarcoma)
44
473
SKIN (Skin - Non Melanomas)
160
1092
TBD (To be determined)
33
215
TEST (Testicular)
25
251
THOR (Thoracic, other)
46
451
THYM (Thymus)
17
195
UNK (Unknown, unknown orgin)
129
801
URTH (Urethral, Penile)
17
157
UTER (Uterine)
241
1195
VUL (Vulva, Vagina)
26
121
All
8103
77295
TCRN Challenges
• EHR = Cerner Millenium / Pathnet
(switched from Copath in 2005)
• Biorepository Database: TissueMetrix
– Possibly soon to convert to OnCore/
LabVantage / LabMatrix…
• Mostly non-functional for translational
research applications
GRU Pathology
• 12 - 14,000 surgical specimens/year
• Since 2000 (target for retrospective data
capture) – 255,000 reports
• MCG surgical emphases: endocrine oncology,
urologic oncology, gynecological oncology
Our Progress
• GRU Biorepository
Team added to caTIES
•
Added to TIES-TCRN
study
Compliance at GRU
• obtained IRB sanction of TIES consortium
aggregate data to be designated as nonhuman subjects research
• IRB renewal for TIES project
IRB Approval
IT support at GRU
• engaged the services of Mike Hyrman at
Encore Health Resource, LLC to develop
code to export retrospective surgical
pathology reports from Cerner Pathnet into
firewall-protected honest brokerage
archive
Governance at GRU
• Creation of GRU TIES project approval
committee to evaluate external and
internal TIES project submissions, in
conformance with TCRN governance
mandate.
– Natasha Savage
– Nita Maihle
– Roni Bollag
Safe Harbor at GRU
• DeID software package licensed (Y2) from
DeIData systems implemented to generate
over 140,000 reports with scrubbed
identifiers
• Phase I and II complete for DeID QA audit.
Phase III nearly complete
Total Reports to-date
148,430 reports
Report Types
• Autopsy
• Surgical
• Bone marrow
• Cytology (likely superfluous)
Phase II – bin by year
Phase II – bin by age
Phase II – bin by gender (?)
Pilot Project Approval
Summary
• GRU node is up and running…
• Compliance is current from institutional
perspective
• Gaps in data upload – needs to be tweaked
• QA is in progress (I and II complete)
• Logistics in pilot project implementation:
– So far, so good…
Looking Forward to the Future!
TCRN F2F Meeting 2015
Roswell Park Cancer Institute
•
•
•
•
•
•
•
Independent, free standing cancer institute since 1898
NCI designated comprehensive cancer center since 1974
Buffalo, New York
27 acre main campus and three affiliated sites
133 hospital beds
3,282 employees, including 308 faculty members, 606
nurses, and over 100 senior scientists
20 scientific shared resources including biospecimens and
data banks
Roswell Park Cancer Institute
• 5,284 hospital admissions, 205,000 outpatient
visits, 31,000 patients under active care/year
• Cancer Registry Cases In 2014:
•
•
•
3,425 ( analytic class 00-22)
486 (non-analytic class 32—1st course elsewhere, subsequent
treatment at RPCI)
1,670 (non-analytic class 30-43, including consults, e.g., path)
• Population catchment area:
•
•
•
8 county regions of Western New York, 1.5 M (86%) of
RPCI patients
48 states
13 foreign countries
RPCI TIES Current Statistics
•
•
•
•
•
Data from 1995 to 2013
Cancer Patients: 72,376
Cancer Pathology Cases: 156,555
2014-2015 to be loaded
Beyond 2015, live HL7 feed every six
months
• Number of registered users: 14
IT Tasks Completed since last F2F
• Executed RPCI HIPAA Checker scripts to check
for Patient Names, MRNs, and Accession
Numbers missed by De-ID and the TIES autoscrubber (by design)
• RPCI TIES Web Site is completed and opened to
the public (https://www.roswellpark.edu/ties)
• Set up automated live feeds of pathology reports
to a dedicated folder on the TIES private server
for 2015 onward
• Completed Phase 1, 2, and 3 QA of reports
loading, De-I, and querying accuracy
IT Tasks Completed since last F2F
• Executed RPCI HIPAA Checker scripts to check
for Patient Names, MRNs, and Accession
Numbers missed by De-ID and the TIES autoscrubber (by design)
• RPCI TIES Web Site is completed and opened to
the public (https://www.roswellpark.edu/ties)
• Set up automated live feeds of pathology reports
to a dedicated folder on the TIES private server
for 2015 onward
• Completed Phase 1, 2, and 3 QA of reports
loading, De-ID, and querying accuracy
Next IT Tasks
• TIES – LIMS Interface
• Search for corresponding tissue samples
available in LIMS.
• To accomplish this we need a way to get the
Case Set data from TIES central hub.
• Create an API
• Write the case set information into the RPCI-TIES
database besides the TIES Hub
• Load 2014 and 2015 pathology reports
Query Quality Assurance
Process
1.
2.
3.
4.
Used seven test queries to analyze accuracy
Designed and executed an initial query
Manually checked each report to assess accuracy
Based on results, we refined the query
1. For some queries, multiple sub-queries were executed and the results
combined
2. As one example, to look for patients with invasive ductal carcinoma of
the breast, we searched for “invasive ductal carcinoma of the breast”
AND “invasive ductal carcinoma, breast”
5. In addition, to compare results, we conducted a
corresponding query against the Cancer Registry database
Query Quality Assurance
Results
Terms present in the text, but not as the current diagnosis, was the most
common error. The pathologist would mention the search terms in the
“Diagnosis” section but in a different context, e.g., “Patient has a history of
Hodgkin’s lymphoma”
One Actual Example of Using TIES in
Combination with Cancer Registry for
Patient Cohort Selection
Identify patients who have had a surgery at RPCI for Serous Adenocarcinoma
of the Ovary AND a recurrence
1. First, we searched for patients who had the same diagnosis multiple
times (recurrence)
2. Designed two queries: “Serous Ovarian Adenocarcinoma” and “Serous
Adenocarcinoma, Ovary”
3. Created one case set of all results and exported to excel (almost 1,000
total reports)
4. Identified patients that had multiple reports and manually reviewed the
reports (71 reports, 35 patients)
5. Used CR data for the clinical selection criteria
6. Used final list to search for tissue
De-ID QA
Preliminary Results
• All pathology reports for CY 1997 and 2002 were loaded into TIES and run through
the TIES De-ID software.
• Using the TIES built-in randomization function, a subset of reports where chosen for
each year of reporting and manually reviewed for under or over scrubbing of PHI
data.
• Preliminary results using De-ID only without added scripts:
De-ID QA
Final Results
• Run RPCI HIPAA Checker for all loaded reports to
identify PHI missed by De-ID software
• Of the 156,555 reports, 7,549 were flagged as
missed PHI and quarantined for manual review and
scrubbing
• 3,920 (52%) found to have actual missed PHI
• Types of PHI missed by De-ID software
• Patient Names
• Patient MRNs
• Pathology report numbers
De-ID QA
Final Results
After all scripts were run and manual checks were completed,
• 7,549 (4.82%) reports were manually checked and scrubbed during the De-ID
validation
• 3,920 (51.9%) reports contained PHI that was not scrubbed by the TIES De-ID
Software
• Overall, assuming the accuracy RPCI HIPAA checker, our findings indicate that the DeID and TIES Auto-Scrubber missed PHI in 2.5% of the total RPCI reports.
De-ID QA Final Results
RPCI Checker MRN
1.
2.
3.
4.
5.
Designed a script to search for the different formats of a MRN present in path reports
Manually reviewed reports with missed MRNs
Run TIES autoscrubber to reduce the number of MRNs to be checked manually
Reran the script
Flagged reports were quarantined and then manually reviewed
RPCI Researchers Qs at TIES
Introductory Presentation
•
•
•
•
What other clinical data is accessible?
What type of reports are searchable?
Would you have molecular results?
Could you also search radiology reports? (Clinical
staging)
• Can you search other centers data without IRB
approval?
• Could RPCI investigators search radiology reports
loaded in other TIES nodes?
RPCI-TIES/TCRN
Team and Roles
• Dr. Carmelo Gaudioso, RPCI CoI., member of
executive TCRN Committee
• Monica Lopez Murphy, TIES Regulatory
Administrator and member of the TCRN P & P
subcommittee
• Kelly Duncan, TIES Regulatory Administrator, user
educator
• Mayur Sakthivel, TIES IT Administrator, responsible
for RPCI TIES IT development
Progress Toward Implementation
• TIES is live as 09/09/2015
• 14 active user accounts
• Dissemination program
• Presentations and TIES Demo
• DSRGs
• Scientific Programs
• Fellows
• Significant investigator interest
• Website, fliers, Mission Magazine article
• Scheduled user training sessions
Q&A
Thanks!
Questions-Comments?