2. The Science of Cognition and Human-Computer Interaction in Health Care
Vimla L. Patel, PhD, DSc
Cognitive Studies in Medicine:
Centre for Medical Education
McGill University
Montreal, Canada

3. Evolution of Medical Informatics: Importance of Interface Design
Technology is “good enough”
User Experience Dominates
Technology Matched to Increasing User Sophistication
Unfilled Technology Need
High
Sophistication of Typical User
System Complexity Masked from Lay User ?
Technology Delivers Basic Need
Current
Trend
Risk of Excess Functionality
Low
Low
High
Sophistication of Technology

4. What is the User Interface?
Those aspects of the system with which the human user comes in contact
Examples: ATM, point-and-click navigation in web browsers...
Output
Language
Communication
Protocol
Input
Language

5. Ways to Deal with User Interface Design
Focus on the way the user has traditionally functioned
Focus on optimizing the interface and requiring user adaptation
Study Interaction to see:
what aspects of the user can change
what aspects of the system can change

6. Dimensions of Human Computer Interaction (HCI)
Technological
Hardware and Software Advances
Cognitive
Representation
Knowledge Organization
Reasoning and Strategies

7. Cognitive Aspects of Interface Design
Human Processes:
Comprehension
Navigation
Communication
These Affect:
Learning
Performance

8. Comprehension and Levels of Meaning
Text-based Model - Surface
Syntax / Semantics
Generation of Local Inferences
Propositions
Situation Model - Deep
Representation of Events, Actions and Persons in Context
Conceptual
Generation of High-Level Inferences
Pragmatics: Context-Bound Inferences

9. Comprehension Processes
CLINICAL DATA
Data-driven Process
Representation of
Clinical Information
Interpretation
(in Context)
Comprehension
NEW KNOWLEDGE
Generalization
PRIOR
KNOWLEDGE
Instantiation
Conceptually
Driven Process

10. Semantic Representation of Natural Language Analysis
Semantic Network
(relationships between
propositions)
CONCEPTUAL
REPRESENTATION
situational model
PROPOSITIONAL
REPRESENTATION
text-based model
Propositional Analysis
(a form of representation of a
semantic network in memory)
NATURAL LANGUAGE
expressed through
THOUGHTS
AND IDEAS

11. Structure of Medical Knowledge in Problem Solving
SYSTEMS
LEVEL C1 C2
DIAGNOSTIC
LEVEL D1 D2 D3
INTERMEDIATE
CONSTRUCT
FA1
FA2
FA3
FA4
FA5
FINDING
LEVEL F1 F2 F3 F4 F5 F6 F7 F8 F9 + + + O1 O2 O3 O4 O5 O6
OBSERVATION
LEVEL O7 O8 O9
O10
O11
O12

12. Cognition and HCI Learning to use technology “Effects with” technology
Learning beyond technology
“Effects of” technology

13. Human-Machine Cognition
Human experts skip steps during the reasoning process
Decision-support systems go through every step methodically in the logical process
This distinction can lead to cognitive dissonance between users and machines

14. Usability Usability Testing
Testing of ease of use must be theory-driven
Usability Inspection
Cognitive walkthrough as a method for gaining insight into user perceptions and misconceptions

15. Usability Inspection Use of Medline
Cognitive Walkthrough:
Assess usability for specific tasks
Usability analyst steps through system interface methodically
Question:
Can we anticipate user goals, actions and potential problems in using an information retrieval system like Medline?

16. Cognitive Walkthrough - Medline Search
TASK: To find articles from Medline related to pregnancy and diabetes
Goal 1: Find articles related to diabetes
Sub-Goal: Do a keyword search on diabetes
Action: enter key combination <control R>
System Response: “Enter a word or phrase”
Action: type in diabetes and press enter
System Response: Returns entries
Potential Problem: List too extensive
Goal 2: Find articles on pregnancy
As above (Goal 1)

17. Goal 3: Merge list of entries on diabetes and pregnancy
Potential Problem: Subject must map term “combine” to action “merge”
Sub-Goal: Combine data sets (diabetes and pregnancy)
Action: Key combination <control n>
System Response: Screen with 2 sets and instructions “Use spacebar to select at least two sets to combine and then press <Enter>”
Action: Scroll to diabetes, press space-bar
Action: Scroll to pregnancy, press space-bar
Action: Press <enter>
System Response: Combine sets screen: Choose Boolean connective “and” or “or”
Potential Problem: Choice of connectives
Action: Select “and” and press <enter> ….

18. Results Numerous potential problems To perform the task required:
12 sub-goals
22 actions
9 transitions between screens
The goal-action analysis shows the cognitive load put on the user in doing simple tasks

19. Impact of a CPR System on Human Cognition
Studies of transition from paper records to CPR and back to paper record
Impact on knowledge organization, reasoning, and learning

20. Changes in Reasoning Patterns
Paper Records
Data driven reasoning
Computer-Based Record
Problem directed reasoning
Return to Paper Record after CPR
Problem Directed Reasoning
Residual effect of CPR on learning (after CPR removed)

21. Changes in Information Management and CPR Use
% of Record
Contents

22. Diagnostic Reasoning Using Paper Record
Multiple Hypotheses
Patient Data
Diagnostic Reasoning Using CPR
Patient Data
Hypotheses

23. Information in CPR and Hand-Written Records
Category of Information
Hand-Written
Patient Record
Computer-Based
1. Chief Complaint 10 28
2. Past Medical History 13
3. Life Style 33 19
4. Psychological Profile 11
5. Family History 7 14
6. History of Present Illness 55 27
7. Review of Systems 52 8
8. Physical Examination 60
9. Diagnosis 9
10. Investigation 29 17
11. Treatment 21 24
Total Entries
304
225

24. Collaboration and Distributed Cognition
Investigate communication between humans, between machines, and between humans and machines
Understand
How people think
analog processing, imprecise
How machines think
digital processing, precise
Capitalize on individual and group expertise

25. Sociometric Analysis of Email Communication PatternsB E
Stanford
Columbia F C D G
InterMed
Central H K J
DSG
MGH I L M

26. Email Communication Patterns
Episode 1: Vocabulary Server and Guidelines
(Jan-Feb 95)
Stanford A
Columbia B E F C D
INTERMED CENTRAL K
MGH L M

27. Email Communication Patterns Episode 2: Guidelines Representation
(Feb-Mar 96)
Stanford A
Columbia B E F C D G
INTERMED CENTRAL H K J
MGH
DSG I

28. 5. On-line Questionnaire Data
Evaluation of PatCIS
1. Video Based
Usability Testing
(“think aloud”)
2. Telephone
Interviews
(audiotape)
Patient
PatCIS
4. Log File - System
Usage Database
(e.g. infobutton)
5. On-line Questionnaire Data
(from pilot study)
Interact via Internet 3.
(to evaluators
& designers)

29. Patient-System Interaction
Lay people’s conceptual model of:
Normal health and illness
How people reason about health
How technology works
Study activities of people as they interact with systems

30. Reasoning Patterns, Data Entry
Organization of knowledge differs among lay and professionals
Reasoning patterns of professionals and patients are also different

31. Patients’ Use of Concepts in Explanations of their Decisions about Diabetes
Reliance on lay knowledge
No use of biomedical knowledge
Influence of impact of decision on daily activities, effect of regimen, and fear are major contributing factors to decision making related to illness
Information from health care team is reported only 8% of the time, but has become incorporated into patients’ knowledge base and is used in that way.
Cytryn & Patel (1998)

32. Data Entered into the System Accurately by Patients and Physicians
100
Patients
Physicians 75
Mean Percent Frequency 50
Errors occur throughout
Patients and physicians both about equally accurate in entering glucose level and diet, activity, and stress levels
Physicians more accurate in entering Insulin Dose and Hypoglycemic Reaction
Different emphasis 25
Blood Glucose
Diet, Activity,
Insulin
Hypoglycemic
Level
Stress
Reaction
Cytryn & Patel (1998)
Type of Information

33. Reasoning Pattern: Patient
Instruction:
Explain decision
Response:
“Well let’s say if I’ll eat a pound of candies my sugar will go sky high and it might even cause a stroke!”
Complications
Increased
Carbohydrate Intake
Increased Blood
Glucose Level
COND
(lead to)
Relates symptoms (data) to intermediate factors to underlying pathophysiology
Decision - will impact on weight
Cytryn & Patel (1998)

34. Reasoning Pattern: Physician
Instruction:
Interpret Scenario
Response:
“A fasting blood sugar of 160 pretty much says she’s diabetic. The postprandial is 200 also tells us she’s diabetic. She’s overweight, that’s already got me thinking that if she’s insulin resistant, then I might be able to improve that by having her lose weight.”
Overweight
Fasting Blood Glucose
Postprandial Blood Glucose
Insulin Resistant
Diabetes
COND
Relates symptoms (data) to intermediate factors to underlying pathophysiology
Decision - will impact on weight
Cytryn & Patel (1998)

35. Reprise: Reasoning Patterns
Patient
Increased
Carbohydrate Intake
COND
Increased Blood
Glucose Level
COND
(lead to)
Complications
Physician
COND
Fasting Blood Glucose
COND
Diabetes
Relates symptoms (data) to intermediate factors to underlying pathophysiology
Decision - will impact on weight
COND
Postprandial Blood Glucose
Insulin Resistant
COND
Overweight
Cytryn & Patel (1998)

36. Lay Reasoning about Childhood Nutrition
COND:
Photograph of Child
INDIGESTION
CAU:
Throwing up of food
Marasmus
Malnourishment X
CAU:
Food sits in the stomach
No food for arms and legs
Enlarged stomach,
thin arms and legs
Kwashiorkar
Sivaramakrishnan & Patel (1993)

37. Summing Up: What’s Next?
Cognitive and social sciences are mandatory for design and implementation of UI’s
Must have deep understanding of what people do and how they reason about it before developing systems that support those activities
Must have better understanding of how people think about the systems they use (acknowledging different expert and lay models)
Recognize the movement towards distributed systems, with human beings as heuristically-driven, analogue thinkers partnering with precise digital systems.

38. More Future Challenges
We will move to greater complexity in systems, but need to consider smaller, manageable tasks and thus less complex systems as experienced by users
Recognize relation between education and training
Training is not enough
Must educate to support training

39. Education to Support Training
Linked to other Information Systems
Easily & Quickly Accessible
User Interface
Decision Making
Typical
(Certain)
Guidelines
Action
Training
(Action)
Update
Ambiguous
(incomplete,
inconsistent)
Discussion/Dialogue
Seminars,
Community-based groups
Education
(Knowledge)

40. The Key Role of Theory in Generating Design Recommendations
Interpretation
Analysis and Results
Theory
Guides
Data
Generalize X
Recommendations
Customization to Case