Transcript Tailoring Health Information on the Web to the Needs of

Knowledge Representation In
Nursing and Its Impact on
Database Design
Josette Jones, RNc, Licentiate Nursing,
Licentiate MIS
Patricia F. Brennan, RN, Ph.D., FAAN,
FACMI
University of Wisconsin Madison
Purpose of the Study
Report on
• Indexing web pages using a subset of MeSH
to create a set of patient specific information
Background and Significance
•Changes in health care
• Consumerism
• Cost management
•Need for indexing health information
Methods of Indexing
• Keyword indexing
• Keywords are terms occurring in the documents itself
(title, abstract, body, URL)
• Concept indexing
• Concept terms describe the subject matter of the
document.
• Concept terms may or may not occur in the document
• Concept terms are taken from Knowledge Representation
Models
Knowledge Representation Model
• A Knowledge Representation Model has :
• An underlying knowledge representation language
(meta-language) with its vocabulary and explicit
structure
• A semantic (meaning of the expressions of the
language)
• A restricted syntax (set of reasoning rules)
• Examples of Health Care / Nursing Knowledge
Representation Models
The HeartCare Project
The HeartCare Project
• Features
• Providing health information graduated to patient’s stage of recovery
and tailored to his/her medical profile and individual needs
• Design
• Filtered set of cardiac recovery resources available on the web and
self constructed web pages are stored in an Access© database.
Web pages are described with index terms
• Index terms are also used to describe the medical profile
• Matching algorithm web page – patient
Indexing Web Pages in HeartCare
Nurse-clinicians tagged web documents with:
• Selected concepts from Medical Language
Subject Heading (MeSH), a class of
Knowledge Representation Models
for bio-medicine
• Supplemented with terms reflecting local
clinical practice
Implications for Retrieval
Issues:
(1) too many pages pulled per patient
(2) too many duplicate pages
(3) some pages were pulled that did not exactly
match the patient profile.
Examples of Total Web Pages in
Combination with Menu title
Retrieved for a Patient
Patient
Combinations
Retrieved
Unique
Combinations
patient 1
266
138
patient 2
891
647
patient 3
324
281
patient 4
584
203
Example of duplicate page retrieval for a Patient: Web Page
http://rex.nci.nih.gov/NCI_Pub_Interface/Clearing_the_Air/clearing.html,
keywords smoking and behavior changes
•
•
•
•
•
•
•
•
•
Menu Title
Taking charge of your health - Week
3-6
Taking charge of your health - Week
3-6
Taking charge of your health - Week
3-6
Beginning lifestyle changes - Week 712
Beginning lifestyle changes - Week 712
Beginning lifestyle changes - Week 712
Changing your lifestyle - Week 13-26
Changing your lifestyle - Week 13-26
Changing your lifestyle - Week 13-26
Condition
• Diabetes
• Hypertension
• Smoking
• Hypertension
• Smoking
• Diabetes
• Diabetes
• Hypertension
• Smoking
Pages Retrieved that Does not Match
the Patient’s Profile
Sample Male Patient with Risk Factors
Hypertension and Stress
• 11 different pages relating to the risk of smoking
and smoking cessation
• 8 web pages relating to risk factors for women
• 5 pages specific about being overweight and
weight loss
• 2 pages on the topic of diabetes management
Example of Indexing
http://www.women.americanheart.org/physicians/sub_content/ten.html
tagged with the terms “diet “and “weight” is pulled 4 different
times for the menu heading “Ten questions a woman should ask
her healthcare provider”
http://www.amhrt.org/Heart_and_Stroke_A_Z_Guide/calccb.html
with terms “Beta Blockers/Calcium Channel Blockers” and
“Medications” are pulled for all conditions that have the subject
heading assigned, even when not applicable
Discussion
• Flawed indexing system
• Conceptualization problem
• Lacking structure of index terms
This study is supported by NLM/NINR Grant
LM06249, Principal Investigator Dr. P.F. Brennan
The authors want to thank the members from the
HeartCare team for their advice and support.
Josette Jones
[email protected]
Patricia F. Brennan
[email protected]
University of Wisconsin Madison