Medical Records and Electronic Documents : a Proposal

1
Medical Records and Electronic Documents a
Proposal

• F. Laforest, A. Flory
• LISI (Information Systems Lab)
• frederique.laforest,flory_at_insa-lyon.fr

2
Information Capture vs Data Computing

• Computerised medical records
• main barrier information capture slowness and
  rigidity compared to computing facilities
• Structured databases
• are necessary to query and compute data
• are associated to capture forms
• Electronic documents
• a universal way of communication due to the Web
  expansion
• provides simplicity and flexibility
• data units are difficult to extract

3
Our Proposal DRUID

• The DRUID system we are proposing
• can be used in the specific case of medical
  records
• offers freedom end-users would fill in
  documents with the information they want to store
  in a convenient order and in a more free way
• ensures efficiency it allows to fill in a
  database quasi-automatically from document
  paragraphs

4
Documents captured in DRUID

• Semi-structured documents
• XML language
• More precisely Weakly-structured documents
• The end-user builds documents by including free
  pieces of text quoted by tags
• Strongly-structured documents are used for
  system-to-system communication
• A Document Type Definition (DTD) provides the
  list of tags allowed

5
Strongly-structured vs Weakly-structured
ltpatient id245gt ltnamegt Dupont lt/namegt
ltfirst_namegt Henri lt/first_namegt ltprescriptiongt
Give ltdosegt 3 lt/dosegt ltdose_unitgt pills
lt/dose_unitgt of ltmedication id12gt aspirin
lt/medicationgt ltfrequencygt 3 lt/frequencygt times a
day during ltdurationgt 10 lt/durationgt
ltduration_unitgt days lt/ duration_unit gt
lt/prescriptiongtlt/patientgt
ltpatient id245gt ltnamegt Dupont lt/namegt
ltfirst_namegt Henri lt/first_namegt ltprescriptiongt
Give 3 pills of aspirin 3 times a day during 10
days lt/prescriptiongtlt/patientgt
6
Prototype Weakly-Structured Documents
7
Prototype Strongly-Structured Documents
8
A Document Analyzer

• To go automatically and transparently
• from the Weakly-structured document to the
  Strongly-structured document
• from the Strongly-structured document to data in
  the Database
• Using mapping rules that tell
• what to find in which type of paragraph
• how to build the strongly structured document
• how to fill in the database
• Mapping rules are hand-written by a designer

9
Mapping rules - Introduction

• Based on
• standard ways of writing gt patterns to recognize
• thesauri and classifications of the application
  domain
• Light example
• duration of a prescription search patterns lt
  during x days gt or lt during x months gt
• The pattern lt during (number) (duration
  units)gt

10
Mapping rules - 3 levels in a rule

• Rule level
• for a paragraph, set of segments to find in it
• PrescriptionRule doseSegment?,
  durationSegment?, drugSegment
• Segment level
• lists the different ways to write a segment
• durationSegment durationExpr1durationExpr2
• Expression level
• pattern itself and database filling
• durationExpr1 duringWordsList,
  NumberListPosology.duration,
  UnitsListPosology.durUnit

11
Mapping rules - Format

• rule
• ruleName segmentName option? , segmentName
  option?
• segment
• segmentName expressionName
  expressionName
• expression
• expressionName (thesaurustable.attribut)
  thesaurus , (thesaurustable.attribut)thesau
  rus
• option ? //for optional segments

12
Architecture of our System
13
Conclusion and future works

• DRUID
• weakly-structured documents for information
  capture
• data stored automatically into a database
• Prototype
• in Java and Enterprise Java Beans
• analyzer for prescriptions done
• user interface done
• To go further
• improve analyzer
• test on other paragraph types