Title: TRANSFORMING BONE MARROW TRANSPLANTATION DATA INTO A CASEBASED REASONING SYSTEM FOR DECISION SUPPORT
1TRANSFORMING BONE MARROW TRANSPLANTATION
DATAINTO A CASE-BASED REASONING SYSTEMFOR
DECISION SUPPORT
- Hadi H.K. Kharrazi
- Grace Paterson
- Tuyet Thieu
- Zhihong Wang
- Feb 2004
2BONE MARROW TRANSPLANTATION OVERVIEW
3Bone Marrow Transplantation Overview Contd.
4Bone Marrow Transplantation Overview Contd.
5- Clinical Conceptualization and Reasoning
- Physicians preferences for colleagues as
information resources may be due to the way
clinical concepts can be conveyed - Medical educators focus on learning in context
- Case-specific knowledge, e.g., why a medication
was given to a specific patient - Knowledge abstraction and general rules
6Unified Medical Language System (UMLS)
Hematopoietic development begins with a single
bone marrow "stem cell. Hematopoietic
(Hematopoietic System) Body System
CUIC0162791 development lt2gt (development
aspects) Physiologic Function
CUIC0011900 Beginning Temporal Concept
CUIC0015663 Single lt2gt (Singular) Quantitative
Concept CUIC0023693 bone marrow lt1gt Body Part,
Organ, or Organ Component CUIC0005802 Bone
(skeletal bone lt2gt) Fully Formed Anatomical
Structure CUIC0001234 Marrow lt1gt Body Part,
Organ, or Organ Component CUIC0005678 Stem cell
(Stem Cells) Cell CUIC0005767
- Metathesaurus
- Semantic Network
- MetaMap Transfer process text for concepts
- Coding knowledge from ill-structured clinical
discourse - Natural language processing
- Filter for pertinence
7Core concepts for problem, solution, outcomes
Generalized Rules
Indexing-Vocabulary For Patient Cases
Disease model and operational knowledge
8Creating the Case Base
- Knowledge acquisition is a bottleneck
- Implicit knowledge stored in Patient Records
- Case Acquisition and Transcription Info-Structure
Methodology transforms EMRs into Operable
Clinical Cases - Similarity between cases
- Retrieve best-matching past-cases to formulate a
solution for case at hand
9Data Presentation
- What is an Operable Clinical Case?
- Numerical Standardization Translation
- Pattern Presentation for an OCC
10What is an OCC?
- An operable clinical case is a real life case
that can be a new case or an on procedure case.
It composes a set of standardized
attribute-value pairs (defined as a set of
situation-problem), solution (treatment,
medication), and diagnostics. - OCC Component
- Problem Attributes Age, Gender, Symptoms, Vital
Signs, Lab Report. - Solution Attributes Diagnostics / Prognostics
- Outcome Attributes Assessment, Rehabilitation,
Progress Status. - Standardization enables Indexing
- Indexing is a vector of attributes
11Numerical Standardization Translation Methodology
- Numerical Content Equivalence (NCE) is a process
to translate continuous numeric values of
attributes in an EMR to discrete equivalence-
values in an OCC. - There are two possible algorithms for the
conversion of the attribute-values in an EMR
12- Algorithm 1
- Defining an interval of values and assigning a
category value to the interval. - If the numeric value of age or lab result is
among the range of values, then mapping the value
of age into the appropriate category. - For example The value range above 1000 per
µL of platelet count attribute is mapped to 0 and
below 1000 per µL is mapped to 1. - Algorithm 2
- Defining a distance threshold to design a pattern
rather than a number or a category. - For example During the first 6 weeks, until the
new bone marrow starts making white blood cells
(engraftment), the complication of
Graft-Versus-Host Disease (GVHD) could happen to
a patient. A pattern of symptoms can be used to
prevent GVHD during the engraftment period.
13ECOG Performance Status
14Pattern Presentation for an OCC
- Pattern is the combination of signs, symptoms,
and terms used to describe the patient in the
PROBLEM. - Each case has a SOLUTION described as a pattern
of activities (e.g., medication) - The OUTCOME is the discharge status is the status
of the patient who has received the BMT
treatment. The Eastern Cooperative Oncology
Group (ECOG) scores are used to measure a
performance status of a patient to assess how a
patient's disease is progressing, assess how the
disease affects the daily living abilities of the
patient, and determine appropriate treatment and
prognosis.
15Transforming clinical records into EMR
- Input our cases are from a single resource and
in MS Excel format - Standardization IBMTR HL7 CDA-compliant EMR
schema - Transformation implement structural equivalence
between the original Medical Record (MR) and EMR - References
- IBMTR HL7 CDA-compliant EMR schema
- Core Insert form
- Identify and extract meaningful terms (such as
temperature, chemotherapy type etc.) in MR to map
conceptually equivalent predefined EMR-attributes - Classify EMR-attributes for problem specification
(Admitting diagnosis, pre-conditioning,
transplant procedure) - Localization
16Clinical Document Architecture
- Health Level Seven (HL7) is one of these
standards that allow communication and
integration among health related computer
information systems and applications - An HL7 specification for the BMT community was
submitted in 2001 - Document Architecture (CDA)
- Refers to HL7 Reference Information Model
- Specify the format of the information and the
method for communicating the information - An application of the CDA standard (IBMTR) has
been proposed to the HL7 community of practice as
an international BMT patients information
exchange standard - Microsoft Office InfoPath software supports the
design of the set of CDA documents to represent
information flow
17Solution Architecture
Syed Sibte Raza Abidi, Knowledge Management
lecture, January 2004
18Solution Flow Diagram