HumanCentred Design of Medical Decision Support Tools

1
Human-Centred Design of Medical Decision Support
Tools

• Peter Jagodzinski, Mo Harris, Keith Greene
• Human-Centred Systems Design Research Group
• University Of Plymouth

2
Overview

• Problem adverse events in obstetrics
• Solution (?) expert systems engineering
• Reality of the work system
• Human-centred redesign
• Revised functional model of the DSS
• Conclusions human activity systems need
  human-centred design

3
Problem

• The need for a 25 reduction in the incidence of
  adverse events in obstetrics and gynaecology
• (Chief Medical Officer in Organisation with a
  Memory, DoH 2001)

• Communication problems contribute to twice as
  many errors as skill-based problems
• (Wilson et al., The Quality of Care in
  Australian Health Care Study, The Medical Journal
  of Australia, 1995)

4
A decision support system for obstetrics teams
a case study
5
Software engineering approach

• top-down design objectives pre-determined by
  technical experts
• technically-led optimises on technical
  considerations
• reductionistic problem is decomposed into its
  technical components
• Deterministic assumes everything in the system
  is predictable

6
Solution expert system

• Problem solving, rule orientated
• Single intervention
• Didactic advice
• Single user
• Patient excluded
• Individual expert cognition, finite state
  machine model
• It works in experimental testing as good as
  experts , better than average clinical practice

7
But, the reality of the work system

• Multiple players
• Distributed decision-making
• Ill-structured problems
• Shifting, competing goals
• Context of organisational objectives
• Time stress
• Action/feedback loops, not single-point decisions
• High stakes

8
Human-centred systems (re-)design approach

• Methods from Sociology and Social Psychology
• Ethnographic study of obstetrics work system
• Making work visible
• Staff apparently operating with different mental
  models of events
• Parents ill-informed on progress, highly anxious
• Fathers present but excluded
• Record-keeping takes priority for attendant
  midwives insufficient psycho-social support for
  parents
• Lack of up-to-date information to central
  Delivery Ward management

9
Time midwives spent out of delivery room and
record keeping for recorded 1st stage of labour
111 hrs
30 hrs
21 hrs
10
Record keeping frequently impaired supportive care
11
Issues of accountability. The buck passed. Who is
responsible?
Senior House Officer (SHO) asked to review
cardiotocogram Midwife It just flipped up. We
tried left side right side, its OK now. We plan
to sort of reassess in an hour for that lip to go
and then an hour wait no more than that. Its
very very thin meconium, its not thick at all,
it was just there. As I say it often happens in
labour. Do you just want to write that down? Ill
get the little stamp for you. SHO Actually I
wouldnt bother with the stamp cause that is
going to be quite hard to put in a category.
12
Rethinking the decision support system

• New functional model
• Problem representation for shared mental models
• Informed narrative, not didactic advice
• Longitudinal process, not single-point decisions
• Mapping path to successful completion a roadmap
• Team orientated
• Patient-centred, parents included
• Contextual domain model instead of individual
  cognition model
• ie a Socio-Technical system working practices
  technology

13
Record keeping with mother
14
Revised decision support system

• supports bedside record keeping
• acts as a vehicle to support communication
• between clinicians and parents
• between clinicians and clinicians
• provides support and reassurance to clinicians
  and parents

15
Conclusions human activity systems

• too rich to be described satisfactorily by
  software engineering systems analysis and
  design
• dont behave deterministically. They require
  soft, probabilistic methods for design
• methods have to comprehend the qualitative
  features
• methods may not conform with the traditional,
  quantitative expectations of software engineers
• probabilistic methods can be rigorous but in a
  different way
• The work was supported by the Medical Research
  Council, Grant No G9721800