Transcript OWL in CDC`s Tuberculosis Surveillance/Response

HCLS
Semantic Web in Healthcare
A view of where we are and where
we need to go in health care
semantics
Cecil O. Lynch, MD , MS
[email protected]
TM
Validating Content for Incoming Case Notifications:
CDC’s Tuberculosis Reasoner Tool
Authors
Participating Organizations
Division of Information Shared
Services
Division of Integrated Surveillance
System and Services
Craig Cunningham, OntoReason LLC
John P. Abellera, MPH DISSS
Sandy Price, PMP Northrop Grumman
Division of Tuberculosis Elimination
Northrop Grumman
OntoReason LLC
Avenida Information Technology LLC
GB Kesarinath, MS DISS
www.ontoreason.com
[email protected]
www.avenidait.com
TM
Project Charter
PHIN message guide, in the form of HL7 v2.5 for Tuberculosis (TB) report of Verified Case of
TB (RVCT) to CDC.
Several states plan to use their own NEDSS compliant state-based disease surveillance
system to capture and send case information of TB.
However, these systems may not include all the validations to minimize the amount content
error, and in most cases, do not include functionality to alert for drug resistant cases.
Project to develop a message content validation application to implement the business rules
provided by the retired Tuberculosis Information Management System (TIMS)
•
Over 300 business rules, which ensure data collection and reporting quality.
•
Tag messages with content validation errors
•
Alert messages to accelerate program notification of MDR/XDR case reports
TM
Message Processing Integration
• The message processor could be integrated in a number of
alternative ways
• In line with the message processing system
• Post processing message components from a database
• As a Web Service remote procedure call
TM
Deployment Architecture
Figure 4. Data flow diagram of incoming TB case notification message within CDC’s DMB.
TM
Message Content Validation Architecture
Visualization Tools
Semantic Knowledge Store
Outbound
Results
Queue
DMB
JMS Interface
Knowledge Registry
Inbound MSG
Queue
OTR Reasoning Framework
Message Parsing
Reasoner
Message
Validation
Reasoner
Alert Reasoner
Public Health Ontology Extraction for TB Msg.
TM
Types of problems that could be solved by
extending the TB framework
• The application component has great flexibility, and
can be used in a number of ways
•
•
•
•
Processing of case reports for consistency and accuracy
Identification of trends in reporting
Review of historical data against existing standards
Feedback to situational awareness processes
TM
The use of an OWL ontology
• The TB Ontology represents a machine processable version of
the implementation guide
• Contains message element structures with requirements as it
relates to data types, value set usage, field requirements
• Definitions of validation rules
• Contains value set extraction from VADS data
• Correlation between PHIN field identity and HL/7 v2 message
elements.
TM
HL7 Message Artifact Taxonomy
•
•
•
Structured taxonomy
Related to codeSystem
Related to TB Question
•
Carries the attribute of code
TM
Question Detail
Divided into semantic type
Related to:
• HL7 Segment
• Errors that can be associated
• TIMS question it replaces
• HL7 data type of the answer
•
CDC program usage (optional or required)
TM
Rule Processing
• Because of the nature of data validation, a means for expressing
the validation requirements which can generate the desired error
messages is required
• There are a number of different types of rules defined
• Static rules which are not based on the ontology which
enforce certain consistencies or perform logically complex
operations.
• Code validation rules which are generated from the ontology
by looking at the requirements defined for elements and the
associated value set
• Data validation rules which are defined within the ontology
and expressed through the use of rule templates
TM
Message Content Validation Rule
Implementation
TIMS Validation Rules Disposition
Deleted - Message Structure
19
Deleted - TIMS Specific Validation
91
Deleted - Not a single message check
3
Deleted - Obsolete
2
Deleted - Reciprocal
62
CVR Rule
125
Vocabulary
56
Grand Total
•
Count
358
During knowledge engineering process we were able to implement a
more efficient rule set
TM
Message Content Validation Rules
•
Rules executed in Java Expert System Shell
•
•
•
Efficiency of reasoning model
Dynamic configuration from ontology
Enables complex rule definition
•
Ability to deal with HL/7 representations of data, and time based
comparisons things such as the differences between dates
•
Message components are broken down into defined facts and asserted
into the engine
•
Rules infer additional facts based on content such as the existence of a
data validation error or alert situation
•
Ontology extractions
• Message definition
• Vocabulary standards
• Rule definitions
• Error and alert result details
TM
Message Content Validation Results View
Human
readable
TB
message
processed with validation error
The results from the message content validation
processing are represented in human and machine
processable formats and added to the TB message for
communication and storage. The results are then used
to alert program staff of issues via email.
TB message processed with MDR
Alert results. Above is the xml
representation, to the left, is the text
representation of the same result.
TM
Processing Results
No.
# Messages
Processing Time
@Rhapsody (Rhapdy
In Q to CVR
INBOUND Q)
Processing time
@CVR (CVR
INBOUND Q to CVR
OUTBOUND Q)
1
2
3
1670
820
620
37 Sec
16 Sec
10 Sec
7 Min 56 Sec
4 Min 14 Sec
2 Min 49 Sec
• Average time to process a message -> 3.5 sec
• Capacity to process 300,000 messages a day
TM
Conclusions
The content validation tool is intended for use by the CDC in its TB program
•
The baseline for functionality that will be made available to state and local TB programs
•
Ultimately expansion of the message validation methodology to other CDC Program Areas.
The functionality to generate alerts show promise for epidemiologists and can help profile for
MDR or XDR drug treatment.
The results from the message validation/testing and integration within DMB are indicative of
the effectiveness of the alerting functions and the potential for expansion of validation tool
to provide additional feedback to the reporting jurisdictions in a timely manner.
Overall, the TB message validation tool was effective in identifying all tested errors and drug
resistance, and creating alerts with little impact to the overall message processing
•
Successful completion of this project will be dependent on full integration within CDC’s data
warehouse scheduled for 3 quarter 2008
TM