1Department of Anaesthesia, South Manchester University Hospitals NHS Trust, Withington Hospital, Nell Lane, Manchester M20 8LR, UK. 2Division of Imaging Science and Biomedical Engineering, The Stopford Building, The University of Manchester, Oxford Road, Manchester M12 9PT, UK
Accepted for publication: June 26, 2000
Abstract
A survey into the attitudes of anaesthetists to features in monitoring instruments, particularly the design of alarms, visual warnings, alarm limits and the general instrument interface is reported. Questions in the survey had short introductions outlining a clinical scenario followed by items that proposed alternative design features that an instrument might have. Participants were asked to grade their responses to these alternatives on a scale of 1 (strongly disagree) to 5 (strongly agree). The results suggest that anaesthetists would welcome the use of more advanced technology in instrument design. They prefer context-specific messages and alarms. They reject overt control systems for delivering anaesthesia, except for use in exceptional circumstances. Generally, the preferences of anaesthetists are consistent with known principles of safe, ergonomic design.
Br J Anaesth 2000; 85: 7814
Keywords: anaesthetists, attitude to computers; equipment, safety; equipment, alarms; monitoring, computerized
This communication reports the results of a nationwide survey into the attitudes of anaesthetists to features in monitoring instruments, particularly the design of alarms, visual warnings, alarm limits and the general instrument interface. We set out to answer three questions. First, what sorts of design features are preferred by anaesthetists? Secondly, are these preferences consistent with the principles of safe, ergonomic design? Thirdly, do attitudes indicate ways to improve instrument design?
Methods and results
Survey design
The survey had three sections and
was based on a previously validated questionnaire.1 2 The first section consisted of a series of
questions about the features of monitoring instruments. The questions were
usually prefaced by a short clinical scenario. Responses were then invited
to proposed alternative instrument design features relevant to the
scenario. Participants were asked to grade their responses to each of these
items on a scale of 1 (strongly disagree) to 5 (strongly agree). The
wording of the questions and the items is shown in Table 1.
|
The questionnaire was sent to 1500 anaesthetists selected randomly from the British Medical Association membership list. Return was by Freepost through an enclosed pre-addressed envelope. Non-returners were followed up after 1 month with a new copy of the questionnaire and after 2 months by telephone to obtain information which was taken to be representative of non-responders.
Statistical
analysis
Statistical analysis followed a standard form3 using SPSS for Windows version
8.0. Responses were placed in order of popularity using the Friedman rank
(highest, most popular). The mode was used to indicate the extent of
agreement. After adjustment for item score skewness, exploratory factor
analysis was performed using principal components analysis, which produces
the most parsimonious description of factors. Used in this way, factor
analysis seeks to find underlying groupings concordant with an identifiable
underlying attitude.
Responses
In total, 504 valid
replies were received, of which 382 were returned initially and 122 as a
result of reminders. This return rate of 33.6% is disappointing but
compares favourably with the best group in the preliminary study.1 2 No significant differences
between the background of responders and non-responders were
demonstrated.
Question 1: how much control should a
decision support system have?
The design options in the items for
this question concerned the instrument having direct control of treatment
and methods of presenting alarms and warnings. Three factors were
identified. The first was about giving control of treatment to the machine,
and was characterized by low ranks and modes for machine-centred
approaches. The only acceptable situation identified as a candidate for
machine control was when there were distinct signs of danger.
The second and third factors were about how warnings might be given. Public warnings (i.e. warnings that are visible/audible to other staff) characterized the second factor, whereas in the third factor the warnings were private. We conclude that anaesthetists see private warnings as serving functions different from public warnings.
Question 2: how should a computer decision
support system communicate with the clinician?
In this scenario both
warnings/alarms and advice were to be given by the instrument. The
items contrasted flow diagrams and/or text with structured
sounds4 5 and spoken alarms. Advice was
contrasted with warning in the items under this question. Three factors
were identified: factor 1, all graphical; factor 2, all sounds; and factor
3, flow diagrams best used off-line. The anaesthetists showed a
preference for thinking of audible systems only in terms of warnings.
Audible advice had a low rank and mode.
Question 3: what
is the best design of visual warnings on individual monitors?
In the
situation of limited display space, idiographic symbols were contrasted
with different text-based systems, and ranged from simple short
messages to displays that had to be interpreted from an error code. Again,
three factors were identified. The first concerned text-based options,
in which the ranks, modes and factor weights were good. Factors 2 and 3
were non-homogeneous in design type. Idiographic designs appeared in
both, mixed with text-based approaches.
Question 4:
what is the optimal way of displaying alarm limits?
The final
question concerned the design and positioning of alarm limit displays.
Novel limit displays such as polygons6 were included, along with more conventional
colour-coded, numerical and graphical methods. Position options were
also included. Two factors emerged from the analysis, differentiated by
whether the alarm limits are permanently visible near to the main display
or further away. Close association of limits with the raw measurement data
was preferred.
Comment
Bias
The
biggest potential source of bias was that the responders were a
self-selected group with a non-representative interest in
instrument design. This sort of bias would explain the low return rate and
is consistent with the tone of some of the comments made in the third
section. We conclude that it is very hard, if not impossible, to avoid such
self-selection. However, this may not be significant since these
anaesthetists were most likely to have had an interest in equipment and
were representative of those we most need to sample if improvements are to
be made.
What sorts of design features are preferred by
anaesthetists?
The results reported do not suggest that
anaesthetists fear the use of more advanced technology in instrument
design. In fact, many of the responses gave a cautious approval to more
radical approaches than those incorporated in current instruments.
Context-specific visual messages were preferred, as were
sound-based designs with greater complexity, such as structured
sounds. New measurements and more sophisticated displays were preferred,
but detailed advice should be available off-line. Designs of visual
warnings and alarm limits were considered immaterial provided they were
easy to understand and closely associated with the main display of an
instrument. The anaesthetists did not like systems that control the
delivery of anaesthesia.
Are these preferences consistent
with known principles of safe, ergonomic design?
Best practice in
instrument design7
emphasizes consistency, simplicity, redundancy and visibility as being good
design principles. Most of the preferences of the anaesthetists were
consistent with this approach. For example, the anaesthetists preferred
condition-specific alarms in a hierarchy of alarms going from
non-specific towards specific. However, offering more physiological
measurements and more complex displays in an already overcrowded
environment is not consistent with good ergonomics.
Can
attitude research guide future safer designs?
The direct testing of
instrument human factors performance in normal practice is virtually
impossible because critical incidents are too rare. Therefore, testing
requires simulation and measurements of performance which are surrogates of
performance in the field (e.g. reaction time). 8 9 Some design features preferred by
anaesthetists might be tested in this way. The results of such testing will
demonstrate the validity of this attitude-based approach to detecting
key features in design. If it is valid, the type of attitude survey
reported here may prove useful in any
environment.
Acknowledgements
We are happy to acknowledge the financial support of this study by Medical Industrial Equipment Ltd., Exeter.
Footnotes
References
1 Beatty PCW. A preliminary survey into clinical attitudes to computer based decision support systems. Br J Anaesth 1997; 78: 131P
2 Beatty PCW. User attitudes to computer-based decision support in anesthesia and critical care: a preliminary survey. Internet J Anesthesiol 1999; 3
3 Tabachnick BG, Fidell LS. Using Multivariate Statistics, 3rd Edn. New York: Harper Collins, 1996
4 Patterson RD. Guidelines for Auditory Warning Systems on Civil Aircraft. London: Civil Aviation Authority, 1982
5 Patterson RD, Edworthy J, Shailer MJ, Lower MC, Wheeler PD. Alarm Sounds for Medical Equipment in Intensive Care Areas and Operating Theatres. Southampton: Institute of Sound and Vibration Research, 1986
6 Green CA, Logie RH, Gilhooly KJ. Aberdeen polygons: computer displays of physiological profiles for intensive care. Ergonomics 1996; 39: 41228[ISI][Medline]
7 AAMI Human Factors Committee. Human factors engineering guidelines and preferred practices for the design of medical devices. American National Standards Institute, 1993
8 Morris RW, Montano SR. Response times to visual and auditory alarms during anaesthesia. Anaesth Intens Care 1996; 24: 6824[ISI][Medline]
9 Westenskow DR, Orr JA, Simon FH, Ing D, Bender H-J, Frankenburger H. Intelligent alarms reduce anesthesiologists response time to critical faults. Anesthesiology 1992; 77: 10749[ISI][Medline]