Tailoring Health Information on the Web to the Needs of the Individual

1
Tailoring Health Information on the Web to the
Needs of the Individual

• The Role of Knowledge Representations and Formal
  Vocabularies

2

• Josette Jones, RNc, Licentiate Nursing,
  Licentiate MIS
• Gyda Bjornsdottir, RN, MSN
• Patricia F. Brennan, RN, Ph.D, FAAN. FACMI
• University of Wisconsin Madison

3
Purpose of the Study

• Report on
• Indexing patient-specific web pages
• Implication for information searching by patients

4
Background and Significance

• Capitalizing on the interactive technology and
  health-oriented web sites to extend discharge
  teaching
• Issues in information searching

5
Entities Involved in Information Searching

• Set of indexed web documents
• An information need formulated in a query
  language
• Matching mechanism

6
How are documents indexed?How do we search for
information?

• Keyword indexing and searching
• Controlled Vocabulary
• Concept indexing and searching

7
Use of Knowledge Representation Systems in
Indexing and Searching

• Concept terms are derived from a Knowledge
  Representation System (KRS)
• A KRS has
• An underlying terminology,
• A semantic (i.e the meaning of the term),
• A restricted syntax (set of reasoning rules)
• Health Care / Nursing KRS

8
How are Web Documents in HeartCare Indexed?
9

• Nurse-clinicians tagged web documents with
• Selected concepts from Medical Language Subject
  Heading (MeSH)
• Supplemented with keywords reflecting local
  clinical practice

10
Examples of Index Terms Used

• MeSH concepts angina, cardiovascular diseases,
  diabetes, hyperlypidemia, patient-physician
  relation, etc.
• Keywords lifting, driving, eating out, menu
  planning, recipes, etc.

11
Reliability of the Judgment

• Sample
• 54 patient-specific web pages
• 3 independent judges
• Results
• Same terms 69 (37 pages)
• Exact same terms 35
• Additional terms by 1 or more 34
• Same term by 2 judges 15 (8 pages)
• Different terms 16 (9 pages)

12
Search Function for Patients in HeartCare
13
Design of Search Function

• Use of controlled vocabulary (list of index terms
  used)
• Reformulation of the title of the web document to
  better reflect the subject matter

14
(No Transcript)
15
(No Transcript)
16
Use of Search Function

• Sample
• Patients enrolled in HeartCare
• Time frameJuly 1, 1999 February 11, 2000 (218
  days)

17
Use of Search Function

• Results
• Search sessions observed N137
• Patients involved in searches N 37Total
  searches initiated per patient range from 1 to 21

18
Use of Search Function

• Total searches N 137
• Accessed 65 (47.4)
• Selection of keywords only 37 (27.0)
• Browsing and/or retrieval of web pages 35
  (25.5)

19
Use of Search Function

• Total terms used in searches 41/71
• Most used terms appetite, driving, breast pain,
  behavior change, Aspirin, diet, incisions
• Frequency of term selected during a search ranged
  from 1 term (57 times) to 9 terms (1 time)
• Number of web documents viewed per selected
  keyword ranged from 0 (63 times ) to 14 (1 time)

20
Discussion

• Issues related to indexing
• MeSH concepts not granular enough to describe the
  content of all web documents
• Ambiguity of language (e.g. synonymy, polysemy)
• Issues related to searching
• Terms used by nurses may not be
  significant/useful to patients

21

• The results discussed in this presentation were
  collected from the experiences in HeartCare
  (NLM/NINR Grant LM06249).
• The support of the HeartCare team is gratefully
  acknowledged.

22
(No Transcript)