3/5/09
1
Communication in clinical organisations
E Coiera [email protected] www.chi.unsw.edu.au
A few problems to focus the mind
• Modern health systems still struggle to improve quality and safety despite genuine motivation and resource allocation
• While hope springs eternal, Health ICT does have a history of repeated large scale implementation failure
• Why after so many years is all this still so hard in health, when other sectors like finance seem to have moved to fully digital work processes?
3/5/09
2
Four levels of system analysis
1. Algorithms 2. Computer Programs
3. Human Computer Interaction
4. Socio- technical systems
How many possible conversations can happen in a health service?
nurse
doctor patient
nurse
doctor patient
GP Lab n! r!(n-r)!
m =
3
10
3/5/09
3
The communication space is large
• Covell et al. (1985): 50% info requests are to colleagues, 26% personal notes
• Tang et al (1996): 60% of clinic is talk
• Safran et al. (1998): ~50% information transactions face to face, EMR ~10%, remainder was e/v-mail and paper
3/5/09
4
What happens in the communication space?
• Wilson et al. (1995): communication errors cause 17% of system problems, 84% potentially preventable
• Donchin et al. (1995): doctor nurse communication in ICU is 2% of work, but figures in 37% of errors
• Bhasale et al. (1998): communication contributes to ~50% adverse events in primary care
The communication space
• is the largest part of the health system’s information space
• contains a substantial proportion of the health system information ‘pathology’
• is largely ignored in our informatics thinking • is where most information is acquired and
presented
3/5/09
5
Consequences of interaction complexity
• Many tasks, Many teams, task heterogeneity, parallelism, lead to …
• Breakdowns at interfaces: incomplete, inaccurate, delayed or failed message transmission
• Multitasking: Concurrent execution of two or more different tasks. Individual has control of sequencing etc
• Interruption: Forced multitasking. Individual suspends current task with variable warning.
Communication breakdowns at interfaces of care
3/5/09
6
Transitions at the boundaries of care
• Communication breakdowns often occur at the interfaces or transitions between care
• E.g.: sign-off, hand-off, handover, shift changes, sign out of patient from ED
• Involve the transfer of rights, duties and obligations for care of a patient
• Inverse relationship between shift length and n times care is transferred
• Patients admitted by one resident and transferred to another next day have more tests and longer stay (J Gen Intern Med 1990;5:501-5)
Communication breakdowns
• 2007 Review of 444 US surgical malpractice claims: – 13% involved 81 communication ‘breakdowns’ – 73% breakdowns verbal, and 64% involved just 2 people – Commonly associated factors with breakdowns:
• Status asymmetry (74%) • Ambiguity about responsibilities (73%)
– Most common events: • Resident failing to notify attending surgeon of critical events • Attending to attending handoffs
– 43% breakdowns associated with patient handoffs – 39% breakdowns associated with transfer in patient location
(J Am Coll Surg 2007 204(4);533-40)
3/5/09
7
Managing talk at the boundaries
• 89.5% of US EDs report no formal policy on patient sign out, 50% sign out only verbally, and 43% ‘rarely’ documented transfer of attending responsibility (Acad Emerg Med 2007;14(2):192-6)
• Strategies: – Communication triggers e.g. ‘two challenge rule’ if unsafe situation
not dealt with – Read-backs e.g. confirm understanding at handover – Standardised sign-out templates, which include critical fields such
as resuscitation (“code”) status – Computerised rounding and sign out - can halve n patients missed
at resident rounds and improve allocation of resident time to seeing patients pre-round (J Am Coll Surg 2005;200(4):538-45).
Multitasking
3/5/09
8
Multitasking in the primary care consultation room
• Doctor’s use of a desktop resulted in: – Shorten responses to patient – Delayed responses to patient – Dr looked less at patient because looking at screen – Dr not hearing patient comments – Patients tried to judge when to talk based upon Drs
interactions with the computer
– (Greatbatch et al., 1993; Booth at al., 2001)
Impact of task switching
• Switch costs: responses take longer to initiate cf repetitive tasks 200 vs 500 ms; higher error rates
• Preparation costs: advanced knowledge of a switch and time to prepare reduces switch cost
• Residual costs: Even with preparation (600 ms or more) can’t avoid some baseline switch cost
• Mixing costs: Performance recovery after switch always slower for mixed cf single task repetition
(Trends Cog Sci 2003:7;134-140)
3/5/09
9
"If a teenager is trying to have a conversation on an e-mail chat line while doing algebra, she'll suffer a decrease in efficiency, compared to if she just thought about algebra until she was done. People may think otherwise, but it's a myth. With such complicated tasks [you] will never, ever be able to overcome the inherent limitations in the brain for processing information during multitasking."
Interruptions
3/5/09
10
Communication in the Emergency Department
• Face-to-face conversation 89.6%. • 30% of communication events were interruptions, rate of
11.2 per hour • 10% of communication time involved two or more
concurrent conversations (multitasking). • 12.7% of all events involved formal information sources
like the medical record.
(MJA, 2002;176:415-8)
3/5/09
11
Task and Role effects
ED interruption rates vary according to:
• Task loads measured by ‘shift intensity’ or time to see patients
(Isr J Emerg Med 2005;5:35-42)
• Clinical role (and presumably task) – 15 interruptions/hr average – Registrars - 23.5/hr, 35% time – Nurse shift co-ordinators - 24.9/hr
• Most interruptions f2f and related to patient management (Ann Emerg Med 2004;44:268-273)
ICU Ward rounds - conversation interrupted
• 75% time in communication • Conversation initiating interruptions
– 14/hr, 37% communication time • Turn-taking interruptions
– 20/hr, 5.3% communication time
(IJMI, 2005;74:791-6)
3/5/09
12
Why the Interruptions?
• Poor asynchronous channels (email, voice mail), and reliance on synchronous ones (face to face, phone) and pager.
• Synchronous bias amongst staff – pressure of work and ‘ticking-off’ the list – need for acknowledgement – face to face is high bandwidth – selfish local, not global reasoning
• A multitasking environment (requirement for parallel task execution) (BMJ, 1998;316:673-677)
Communication Policies
• UK Medical staff generated 2x as many interruptions as they received (43 vs 23)
• Policies are tactics to filter and prioritise messages (e.g. secretary).
• “I want to always be available…” • “...but I don't want to be interrupted.” • Receivers tried to assess urgency, caller, task based upon
poor information • Callers had no information about availability and so either
interrupted or failed to contact
3/5/09
13
Interruption and Error
• WM = those mechanisms involved in control, regulation and active maintenance of task information
• Interruptions challenge working memory (WM) capacity • New tasks given during an interruption may interfere with
existing tasks leading to disruption of WM processes: – forgetting of tasks.
– believing events have occurred, repeating forgotten events – decreased performance of original task
Resumption lag
• Time to restart a task after interruption double for an interrupted vs initiated task resumption (1.9 vs 3.8 s)
• Cue availability prior to interrupt reduces resumption lag • Suggests preparatory cognitive processes to mitigate
interruption, similar to multitasking • “interruption lag” - brief period prior to interrupt provides
opportunity to prepare to resume and encode retrieval cues to facilitate resumption of primary task
(Altman, Trafton Cog Sci 2004)
3/5/09
14
Reducing interruptions
• Shift from synchronous to asynchronous – Training clinicians to understand impact of interruptions and costs
of multitasking – More voice and e-mail + acknowledgment eg asynchronous
notification of lab results BUT unintended consequence is that if poorly designed, can result in more interruptions
– How? requests: e-directories - locally maintainable, rapidly updated, clinically oriented, personally annotated
– Who? Requests: Role-based call forwarding via programmable switch
– Making work ‘visible’ - whiteboards, active signs -> reduce memory load, recover from memory disruptions.
The Sacred and the Profane
Sacred (classic) making • The computer • The EMR • Terminologies • System architectures • Intelligent decision
support technologies
Profane (in the wild) • Paper • Communication • To-do lists • System implementation • System failures • Local customisation
3/5/09
15
Design challenges for health services
1. Clinicians operate with scarce cognitive resources, multiple team interfaces, multitasking and interrupting, leading to inefficiency and error.
2. We need to understand the current cognitive science literature about communication and task handling e.g. multitasking and interruption
3. We need to study communication and reasoning “in the wild” to discover just how it impacts clinical work
4. We need to start designing clinical environments which minimise boundaries, multitasking and interruption, and support clinicians becoming more effective at task management and resumption
Thank You
3/5/09
16
Cape Town International Convention Centre www.medinfo2010.org