Title: Strategic Management of Anatomic Pathology Laboratory Information Systems for Enhancement of Transla
1Strategic Management of Anatomic Pathology
Laboratory Information Systems for Enhancement of
Translational Research Efforts.
- Ashokkumar A. Patel, MD
- Case Comprehensive Cancer Center
- Case Western Reserve University, Department of
Pathology - ashok.patel_at_case.edu
- APIII 2008
2Objectives
- Understand the advantages and limitations of the
AP LIS and AP LIS data for translational
research. - How to provide large, clean, structurally sound
and de-identified data set from the AP LIS - Integrate this data set with other data resources
to support translational research. - How to take advantage of your APLIS to enhance
biorepository efforts? - A case study from Case Western Reserve University
3Overview Example of facilities and locations
that interact with the AP-LIS.
Data sharing with Cancer Center
Paraffin tissue repository
Outcomes studies
Data requirements for new standards (DICOM
supp122, HL7 CDA)
Clinical Research projects
Billing/Finance systems
Need Direct feeds to Cancer Registry
Residency training studies
McPherson Pincus Henry's Clinical Diagnosis
and Management by Laboratory Methods, 21st ed.
4Anatomic Pathology Report
5General Comments
- The AP report is a signed document, not a
database to database transaction. - It has been designed for a specific purpose (even
though it is used for many others) - It is a communication between physicians
- It is an identified document in the header and
in the text
LIS
EHR
6Evolution of Pathology LIS
- Anatomic Pathology
- Paper reports
- Text based electronic reports
- Billing/transaction APLIS
- Template synoptic reporting
- Structured synoptic reporting
- Structured meta-data encoded reports
7Structured Data Entry
- Various forms of structured data entry are
becoming popular in pathology - Rule based dictations and intelligent
transcription - Dictation templates and Macros
- Structured Data Entry Programs
- Low level structured data programs enhance
parsing and auto coding - High level programs enforce consistency,
completeness and can hard code coding
8Structured Data Entry Tools
9PATHOLOGY REPORTSTODAY
10Pathology Reports After Synoptic Reporting
University Hospitals
7/24/2007
11A Partially Structured Document
12Send specimens to pathology lab
Surgery
Accession/ Log specimen
Gross
Order blocks/slides
Pathology Data Flow
AP Lab Info System
Preliminary Dx./ Ancillary testing
Diagnosis
Final Report
13Partially Structure Document
- A partially structured document
- Large Narrative Sections (Fields) such as Final
Diagnosis, Gross Description represent
database elements - If the case involves more than one clinical
specimen (or Part), the narrative sections
divided by the author into sub-sections (one sub
section for each clinical specimen) - The information model does not match the real
world situation!
Patient (surgical event)
Left Arm Nevus B
Right Arm Nevus A
Case S07-100 Report Final Diagnosis A Dermal
Nevus. B Compound Nevus
14Large Narrative Fields that are Structured to the
Author
15Suppressed data within AP Lab Info System
16- Suppressed data
- Processing, coding and other data generated by
the department - Many not be interesting to the surgeon (but may
be interesting to others)
17- Pathology descriptions and diagnosis
-
- histology processing and inventory data at the
part, block and slide level - LIS that can act as a massive tissue banking
system, with data associated directly with
specimens.
Histology data can be found in discreet data
fields for basic information identifying blocks
and slides.
Individual slide information can be useful for
DICOM whole slide imaging.
18CASE STUDY at
19Information derived from our APLIS For 1995-2007
routine accessions and GYN-cytology. gt 1
million reports processed 1 Facility
UH-CMC Grouped by Sub-specialty pathology
groups/ Organ Systems.
20(No Transcript)
21(No Transcript)
22(No Transcript)
23(No Transcript)
24(No Transcript)
25Data Integration
- Cancer Registry
- Clinical Trails
- Biorepository inventory system
- Family Medicine registry
- Epidemiology/Genetics/Hereditary research
databases - Links to whole slide images (in
process)
TissueBank
Pathology LIS
Cancer Registry
Datawarehouse
Data Mart
Data Mart
Data Mart
Data Mart
Disease Specific Registries (Researchers)
26Links to Frozen Tissue Bank Specimens
27Method
- Identify all variations of specimen and part
identifiers in the data set (2005 reports) - Identify all variations in person identifiers in
the data set - Parse identifiers from text and text sections
from each other - Store in an XML file(s)
Find Problems
Study Dataset
Write/Run Parsers
Find Problems
Write XML Files
28Identification of objects and information
29(No Transcript)
30(No Transcript)
31(No Transcript)
32Identification of objects and information Curren
t limitations with AP-LIS data
33What happens to specimen A?
Dr. I.M. Hacker (surgeon) sends 5 containers from
patient Xs surgery removing tumor from left
lung.
Left upper lobe
A
UPPER DIVISION LEFT UPPER APICAL POSTERIOR
ANTERIOR SEGMENTS
B
C
AP WINDOW 5 POSTERIOR
ANTERIOR AP WINDOW
D
12L
E
8
F
34This is three database lines of text representing
a single paragraph. Full hard returns create new
paragraphs within a single database field.
Free text blob This is a single database field
representing multiple paragraphs. The narrative
text is mixed with the identifier.
35What does Left Upper lobe or 12L
represent? Links to database dictionary must be
specific and accurately linked to narrative
text. (i.e. Part type/specimen received).
Part and Block Identifier are part of the
narrative text.
36Part Labels -Shows up on Final Report -Free
text value -Represents what surgeon wrote on
specimen container -Vocabulary not controlled by
Pathology
Part Types -Suppressed database value -Does not
show up on Final Report -Discreet values and
Vocabulary controlled by Pathology -Avoid using
generic terms to define data dictionary.
37Example Case 2
ltprimary-attribute name"part_id" content"A"
line-nbr"0" auto-create"false"
master-pattern"( A-Za-z(-,.
A-Za-z),. )" preferred-pattern"(A-Z)
"gt
Colon is used
Identifiers are hard to find not consistently
implemented. This makes it very hard to address
information about Specimen A
Dash - is used
38Only Letters used
"( A-Za-z(-,. A-Za-z),. )"
Colon used
Period used
39Large free text fields with no separation
of Identifiers from descriptions
The author is structuring data within a field.
If it is not identified, it not useable!
40Identification of objects and information
allows-Reorganization of AP LIS data for
Researcher friendly views
41Which block has tumor for specimen A?
Research Request Tissue sections on
Adenocarcinoma of the Lung with Adjacent normal
sections.
1
2
3
PartA
Left upper lobe
4
5
6
7
8
42(No Transcript)
43Creation of XML files
44XML
- A file that acts like a database
- Identifies and separates all data elements
- Makes each element addressable
- Allows easy reconfiguration of the file
- Create child files that that contain a subset of
the parent file in any structure you want - Coding and Structured Data can be included in any
XML based report - One can include links to images or links to lab
manuals
45(No Transcript)
46(No Transcript)
47Original String
Parse String
48(No Transcript)
49(No Transcript)
50Parse text link part (specimen)
51Embedded Identifiers
52- Identifiers embedded in the body of the report
- Pathologists Names
- Patient Names
- Hospitals
- Clinical History
- Specimens (Slide IDs)
- There are good clinical and legal reasons for
this - But requires complex, quirky and high
maintenance de-id parsing for use in the
non-clinical space
Imbedded ID
53A subset of available information from your APLIS
54Discreet Data elements
Represents a subset of the useful information in
the LIS
Free Text Sections
55Histology data can be found in discreet data
fields of basic information for identifying
blocks and slides.
Individual slide information can be useful for
DICOM whole slide imaging.
56Discreet data for billing codes, SNOMED codes,
and link to other specimens from same patient.
57A subset of available informationHOWEVER
58Specimen processing information (poorly linked to
narrative data)
Left Upper Lobe Entire Mass
59Coding information (poorly linked to narrative
data)
60Auto coding errors
ltprimary-attribute name"part_id" content"A"
line-nbr"0" auto-create"false"
master-pattern"( A-Za-z(-,.
A-Za-z),. )" preferred-pattern"(
A-Z,. )"gt
61Auto coding errors
Example Case 3
62Who is to be blamed?
- Pathologists
- Transcriptionists
- Residents
- Gross/accessioning techs
- Anatomic Pathology Laboratory Information system
- Auto coder
63Summary
- With a great deal of effort, one can parse a
pathology report so that the data components are
addressable - Such a data set will allow useful things such as
reports on single part or reports that include
just gross description and specimen processing - This will be unique data set - no one else has
such a data set - There is no substitute for clear and consistent
labeling when the data is collected and entered
into your system - Limitations to APLIS exists, because we accept
those limitations
64Best Practice
- Proper training of staff
- Be consistent with transcription of free text
fields. - Punctuation
- Labeling key section(s)/part(s)
- Defining data dictionary
- Specific and descriptive as possible
- (Bx, other / Misc. Vs. Bx, prostate)
- Understanding how text data effects other parts
of the system. - Clinicians reading report
- Auto coding/Billing
- Future research needs
65Summary
- Structured data will lead to opportunities
- Tissue Banking
- Measure Quality and Efficiency
- Outcomes Research
- Grant Funding
- DICOM/Imaging
- Molecular
- Many others.
66Acknowledgement
- Rajnish Gupta System Architect
- Bob Lanese DBA/webmaster
67Thank You.
If everyone is thinking alike, then someone
isnt thinking. George S. Patton, Jr.
- Ashokkumar A. Patel, MD
- ashok.patel_at_case.edu
- 216-368-5106
68- A LEFT UPPER LOBE, WEDGE RESECTIONCRLF--MOD
ERATELY DIFFERENTIATED ADENOCARCINOMA, SEE
NOTE.CRLF--EMPHYSEMATOUS CHANGES.CRLFCR
LFNote The carcinoma measures 1.2 x 1.1 cm in
largest dimension. It is subpleural, but does
not penetrate into the pleura (elastic stain on
sections A5 and A6). Angiolymphatic invasion is
not identified. The margins of this wedge
resection are uninvolved.CRLFCRLFB
UPPER DIVISION, LEFT UPPER APICAL POSTERIOR AND
ANTERIOR SEGMENTS, RESECTIONCRLF--EMPHYSEMATO
US CHANGES.CRLF--RECENT HEMORRHAGE,
CONSISTENT WITH RESECTION SITE (SPECIMEN
A).CRLF--BRONCHIAL MARGIN, NO EVIDENCE OF
MALIGNANCY.CRLF--VASCULAR MARGIN, NO EVIDENCE
OF MALIGNANCY.CRLF--NO EVIDENCE OF MALIGNANCY
IN ONE PERIBRONCHIAL LYMPH NODE.CRLFCRLFC
AP WINDOW 5 POSTERIOR LYMPH NODES,
BIOPSYCRLF--NO EVIDENCE OF MALIGNANCY IN
THREE LYMPH NODES.CRLFCRLFD ANTERIOR AP
WINDOW, LYMPH NODES, BIOPSYCRLF--NO EVIDENCE
OF MALIGNANCY IN TWO LYMPH NODES.CRLFCRLFE
12 L LYMPH NODES, BIOPSYCRLF--NO EVIDENCE
OF MALIGNANCY IN ONE LYMPH NODECRLFCRLFF
LYMPH NODES DESIGNATED 8, BIOPSYCRLF--NO
EVIDENCE OF MALIGNANCY IN SEVEN LYMPH
NODES.CRLFCRLF
69(No Transcript)