Content uploaded by Varuna Prakash
Author content
All content in this area was uploaded by Varuna Prakash on Sep 29, 2014
Content may be subject to copyright.
Mitigating errors caused
by interruptions during medication
verification and administration:
interventions in a simulated
ambulatory chemotherapy setting
Varuna Prakash,
1,2
Christine Koczmara,
3
Pamela Savage,
4
Katherine Trip,
5
Janice Stewart,
6
Tara McCurdie,
2
Joseph A Cafazzo,
1,2
Patricia Trbovich
1,7
▸Additional material is
published online only. To view
please visit the journal online
(http://dx.doi.org/10.1136/bmjqs-
2013-002484).
For numbered affiliations see
end of article.
Correspondence to
Varuna Prakash, Healthcare
Human Factors, Techna Institute,
University Health Network, 190
Elizabeth Street RFE 4th Floor,
Toronto General Hospital,
Toronto, Ontario, Canada,
M5G 2C4;
varuna.prakash@utoronto.ca
Received 11 September 2013
Revised 20 May 2014
Accepted 22 May 2014
To cite: Prakash V,
Koczmara C, Savage P, et al.
BMJ Qual Saf Published Online
First: [please include Day
Month Year] doi:10.1136/
bmjqs-2013-002484
ABSTRACT
Background Nurses are frequently interrupted
during medication verification and
administration; however, few interventions exist
to mitigate resulting errors, and the impact of
these interventions on medication safety is poorly
understood.
Objective The study objectives were to (A)
assess the effects of interruptions on medication
verification and administration errors, and (B)
design and test the effectiveness of targeted
interventions at reducing these errors.
Methods The study focused on medication
verification and administration in an ambulatory
chemotherapy setting. A simulation laboratory
experiment was conducted to determine
interruption-related error rates during specific
medication verification and administration tasks.
Interventions to reduce these errors were
developed through a participatory design
process, and their error reduction effectiveness
was assessed through a postintervention
experiment.
Results Significantly more nurses committed
medication errors when interrupted than when
uninterrupted. With use of interventions when
interrupted, significantly fewer nurses made
errors in verifying medication volumes contained
in syringes (16/18; 89% preintervention error
rate vs 11/19; 58% postintervention error rate;
p=0.038; Fisher’s exact test) and programmed in
ambulatory pumps (17/18; 94% preintervention
vs 11/19; 58% postintervention; p=0.012). The
rate of error commission significantly decreased
with use of interventions when interrupted
during intravenous push (16/18; 89%
preintervention vs 6/19; 32% postintervention;
p=0.017) and pump programming (7/18; 39%
preintervention vs 1/19; 5% postintervention;
p=0.017). No statistically significant differences
were observed for other medication verification
tasks.
Conclusions Interruptions can lead to
medication verification and administration errors.
Interventions were highly effective at reducing
unanticipated errors of commission in medication
administration tasks, but showed mixed
effectiveness at reducing predictable errors of
detection in medication verification tasks. These
findings can be generalised and adapted to
mitigate interruption-related errors in other
settings where medication verification and
administration are required.
INTRODUCTION
Several reports, including the Institute of
Medicine’sTo Err is Human
1
and the
Agency for Healthcare Research and
Quality’sThe Effect of Health Care
Working Conditions on Patient Safety
2
have identified interruptions and distrac-
tions as factors contributing to medical
errors. Distractions were cited as causal
factors in nearly half of all medication
error reports submitted to the United
States national error-reporting database,
and were the most frequently reported
factor contributing to patient harm.
3
Although interruptions may occur at
any stage of the medication process, the
medication administration stage is of par-
ticular interest because it represents the
last opportunity for an error to be inter-
cepted before reaching the patient.
4
Nurses have cited interruptions and dis-
tractions as a top cause of errors during
medication administration,
5
and such
ORIGINAL RESEARCH
Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484 1
BMJ Quality & Safety Online First, published on 6 June 2014 as 10.1136/bmjqs-2013-002484
Copyright Article author (or their employer) 2014. Produced by BMJ Publishing Group Ltd under licence.
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
interruptions are significantly associated with a variety
of medication administration errors (eg, administering
wrong medication, dose, infusion rate).
6
Thus, there is
a strong need to develop interventions that can reduce
interruption-related errors during medication adminis-
tration. To date, a variety of interventions have been
proposed, including: prohibition of non-essential con-
versation, phone calls and pages
78
; use of ‘Do Not
Disturb’vests and signage
910
; use of a medication
administration checklist
910
; and use of a clearly
demarcated ‘No Interruption Zone’
11
or physical
barrier
12
in medication preparation areas. Notably,
most of the above interventions were designed to
reduce the number of interruptions occurring during
medication administration, with limited evaluation of
the resulting impact on medication administration
error rates. Indeed, a recent review suggests that there
is only weak evidence regarding the effectiveness of
such interventions in reducing interruptions and
resulting medication errors.
13
Thus, there is a need to
develop effective interventions for interruption-related
errors, and to assess the impact of these interventions
on medication error rates.
In a previous ethnographic study
14
in an ambula-
tory chemotherapy unit at a large cancer centre in
Toronto, we identified two broad categories of
safety-critical tasks prone to interruptions (ie, medi-
cation verification tasks and medication administra-
tion tasks) that could lead to errors. Medication
verification tasks consisted of checking the five rights
of medication administration (ie, right patient, right
medication, right dose, right route, right time), and
were found to be primarily susceptible to errors of
detection (eg, failing to notice a discrepancy between
the medication order and medication label). In con-
trast, medication administration tasks such as admin-
istering medication via infusion pumps or
intravenous push were found to be susceptible to
errors of commission (eg, setting the wrong infusion
rate). In the current study we aimed to (A) investigate
the association, if any, between interruptions and
medication verification and administration errors, (B)
design interventions to reduce such errors in the
presence of interruptions, and (C) assess the effect-
iveness of these interventions in reducing the identi-
fied medication verification and administration
errors arising from interruptions. We conducted a
simulation laboratory experiment to assess the effect-
iveness of interventions as a prerequisite to live clin-
ical implementation.
Methodology
The current work was conducted in three phases over
a time period of 6 months. An overview of the three
phases is shown in figure 1. Details of each phase are
described in the following sections.
Phases A and C: preintervention and postintervention
experiments
Study setting
Experiments conducted in phases A and C took place
in a high-fidelity simulation laboratory, where nurses
were asked to carry out medication verification and
administration tasks within a highly realistic but con-
trolled setting. This experimental design was chosen
as it allows test administrators to make detailed obser-
vations of the impact of interruptions and interven-
tions in a manner that would be impractical and
unduly disruptive in a live clinical environment.
The simulation laboratory was equipped with
theatre-style rooms, one-way glass and cameras (see
online supplementary figures A1 and A2 in appendix
1)that allowed realistic simulation of an ambulatory
chemotherapy unit, including patient beds, chairs,
computerised physician order entry (CPOE) system,
intravenous infusion equipment and paperwork.
Manikins were used instead of patients. All medica-
tion bags, syringes, intravenous tubing sets, paper
medication orders, medication labels and compu-
terised medication order screens were identical to
those used in the institution’s regular practice.
Coloured water or saline was used in place of real
medications. An audio recording of a busy hospital
unit was played throughout the experiment to provide
realistic ambient noise. An actor-facilitator playing the
role of a charge nurse guided participants through
each scenario. To further recreate the busy,
interruption-filled environment, actors played the
roles of patients, family members and fellow nurses.
Four actors participated in this study, playing the roles
of a charge nurse, a family member and two patients.
A fifth person, whose primary role was to assist the
investigator in the observation room, also played the
interjectory role of a physician. Additionally, three
realistic patient manikins were placed in beds and
chairs, thereby bringing the total number of mock
patients to five. Thus, the simulated environment
mimicked the cognitive load experienced by nurses
working in the chemotherapy unit. Further details
regarding the simulation setting are provided in online
supplementary appendix 1.
Study design
An initial preintervention experiment was conducted
to understand whether or not interruptions were asso-
ciated with medication errors. Nurses were asked to
perform medication verification and administration
tasks under two conditions: uninterrupted (Condition
1) and interrupted (Condition 2). Thus, the experi-
ment was a 2 (interruption condition)×7 (task type)
within-subjects (repeated measures) design. The order
of interruption and non-interruption tasks was coun-
terbalanced to avoid carryover effects.
Results emanating from the preintervention
Condition 2 were used as a baseline (control) for the
Original research
2Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
postintervention experiment. In other words, the
postintervention experiment compared Condition 2
(where nurses were interrupted, with no interven-
tions) to Condition 3 (interrupted, with interventions)
using a between-subjects design. To permit compar-
ability across the three conditions, equivalent scen-
arios, planted errors, and type/timing of interruptions
(where applicable) were used in all conditions, as
listed in table 1. The postintervention experiment
took place approximately 2 months following the pre-
intervention experiment, as the time in between was
used to develop interventions (ie, Phase B).
Table 1 describes all tasks (some of which contained
planted errors), interruptions and performance
metrics pertinent to the simulation experiment. The
tasks, planted errors and interruptions were designed
based on extensive ethnographic observations gath-
ered during a prior study in this care area.
14
Specifically, interruptions were selected based on the
frequency with which they occurred during each task,
as observed during the ethnographic study. To further
ensure that the experiment accurately reflected partici-
pants’real-world practice, the tasks were presented to
participants in realistic scenarios. Participants encoun-
tered each planted error only once per experiment,
even if they performed that task in multiple scenarios.
For example, a participant may have been asked to
verify medication names in five scenarios in
Condition 1, but only one of the five scenarios con-
tained a planted error in the medication name. Each
scenario contained a maximum of one planted error.
Further details regarding the scenarios are presented
in online supplementary appendix 1.
Participants
Nurses from the ambulatory chemotherapy unit were
recruited via a sign-up sheet located in the unit, and
were eligible to participate if they worked in the unit
and routinely administered chemotherapy at the time
of the study. In accordance with institutional ethics
protocols, nurses provided informed consent and
were remunerated for their participation with an
amount commensurate with their hourly wages.
Participant characteristics are summarised in table 2.A
χ
2
test of homogeneity revealed no significant demo-
graphic differences between the two participant
cohorts.
Experimental procedure
At the start of the study, the investigator introduced
the participant to the lab environment and briefly
described the process of simulation testing. In the pre-
intervention condition, participants were asked to
start carrying out the medication verification and
administration tasks. In the postintervention condi-
tion, the participant received 30 min of training on
the interventions prior to carrying out the medication
tasks. Specifically, in the training session, the investiga-
tor explained each applicable intervention and how to
use it. The participant was then asked to practice
using each intervention and resolve any doubts before
starting the experiment. Actors playing the roles of
family members and patients also assisted in the train-
ing process by providing interruptions during the par-
ticipant’s practice with interventions. The training
process was concluded once the participant had
demonstrated his/her ability to correctly use each
intervention by successfully completing each practice
task and using each intervention when applicable. The
actor playing the role of the charge nurse then pro-
ceeded to start the experiment by directing the partici-
pant towards the first scenario.
Data collection
Two trained observers collected live data from an
observation room located behind one-way glass while
the experiment was in session. Specifically, observers
documented errors (ie, Pass, Fail) on an Excel work-
sheet containing a list of all tasks. If there was an
intervention for which compliance was dependent on
Figure 1 An overview of the three phases, Phase A: Preintervention Experiment, Phase B: Intervention Design, Phase C:
Postintervention Experiment.
Original research
Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484 3
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
Table 1 Description of tasks, interruptions, planted errors, performance metrics and applicable interventions in simulation experiments
Task Number and timing of interruptions* Planted error Performance metric†
Applicable
intervention(s)‡
Medication verification tasks (assessment of error detection)
1. Verifying medication name
Participant was required to verify medication
name on label against electronic and paper
medication orders
During medication verification, participant was
interrupted by:
2 requests from nursing colleague; 1 question
from patient
Medication name on label did not
match name on medication orders.
Sound-alike, look-alike medications
were chosen (eg, Carboplatin vs
Cisplatin)
Task was coded as ‘fail’if participant did not
detect the planted error
Verification booth,
standardised workflow
2. Verifying medication dosage
Participant was required to verify medication
dosage on label against electronic and paper
medication orders
During medication verification, participant was
interrupted by:
1 question from patient; 1 work-related call; 1
request from physician; 1 background infusion
pump alarm
Dosage on the medication label did
not match that in the medication
order
‘Fail’if participant did not detect the planted error Verification booth,
standardised workflow
3. Verifying medication volume in syringe
Participant was required to verify medication
volume on syringe label against electronic and
paper medication orders
During medication verification, participant was
interrupted by:
1 request from patient’s family
The syringe contained an incorrect
volume of medication (underfilled
by 5 mL—a clinically significant
amount)
‘Fail’if participant did not detect the planted error Verification booth,
standardised workflow
4. Verifying medication volume in ambulatory
infusion pump (AIP)
Participant was required to verify medication
volume programmed in AIP against medication
order and medication label
During medication verification, participant was
interrupted by: 1 question from nursing
colleague, 1 question from patient’s family
The medication volume
programmed in the AIP did not
match that on the medication order
‘Fail’if participant did not detect the planted error Verification booth,
standardised workflow
5. Verifying patient identification (ID)
Participant was required to verify patient name
on medication label against the patient’s
armband
During patient armband verification,
participant was interrupted by: 1 question
from patient, 1 request from nursing colleague
The name on the medication label
did not match that on the patient’s
armband. Sound-alike, look-alike
names were chosen (eg, Pamela
Chan vs Patricia Chan)
‘Fail’if participant did not detect the planted error Speaking aloud
Medication administration tasks (assessment of error commission)
6. Intravenous push
Participant was required to administer a
chemotherapeutic agent to a patient via manual
intravenous push over the pharmacy-prescribed
timeline of 6–10 min
During the intravenous push, the participant
was interrupted by: conversations from patient
and family, 1 request from nursing colleague,
1 question from patient, repeated background
infusion pump alarms
No error was planted in this task ‘Fail’if participant did not administer medication
within pharmacy-prescribed timeline (eg, 6–10 min
for vinorelbine)
Visual timers
7. Pump programming and infusion initiation§
Participant was required to administer medication
by correctly hanging the medication bag, closing
the clamp of previously hanging medication
tubing set, opening the clamp for medication to
be delivered, and programming an infusion pump
with the prescribed administration rate and
volume of medication
During pump programming and infusion
initiation, the participant was interrupted by:
requests from nursing colleague, requests from
patient, conversations with patient’s family,
and background infusion pump alarms
No error was planted in this task ‘Fail’if participant programmed pump with
incorrect rate or volume, forgot to open/close
appropriate clamps, hung medication bags at
incorrect heights such that the wrong medication
was being infused, or forgot to start the infusion
entirely
No interruption zones with
motion-activated indicators,
speaking aloud, reminder
signage
*Applicable to Conditions 2 and 3 only. The number and timing of interruptions was kept consistent between the two conditions to permit comparability.
†Participants were instructed to report detected errors to the charge nurse (played by an actor-facilitator).
‡Applicable to postintervention condition (ie, Condition 3) only.
§As described in online supplementary appendix 1, the pump programming task occurred in four of the five scenarios. Therefore, Pass/Fail performance was determined using collective criterion; that is, participants had to
correctly program the pump in all four scenarios to receive a ‘Pass’.
Original research
4Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
the participant (eg, speaking aloud), observers add-
itionally documented whether or not the intervention
was used at each instance where an opportunity for
use was present. Observers compared notes after each
session to ensure consensus. Any discrepancies
between observer notes were resolved by consulting
video recordings of the session.
Data analysis
Data emanating from the experiment were coded
according the criteria described in table 1
(‘Performance Metrics’column). McNemar’sχ
2
test
was used to assess differences in error rates between
Conditions 1 and 2 in the preintervention experiment.
Fisher’s exact test was used to assess differences in
error rates between Conditions 2 and 3 following the
postintervention experiment. These comparisons were
justified because all tasks, interruptions and scenarios
were kept equivalent between the two experiments.
An αof 0.05 was used for all statistical tests. All data
were analysed using SPSS V.18.0 for Mac.
Phase B: intervention development
To ensure a participatory design approach (ie, an
approach where key stakeholders and end-users are
involved in intervention design), nine nurses from the
chemotherapy unit who had participated in previous
phases of the study were recruited to take part in
focus groups, where they brainstormed potential error
mitigation strategies and iterated upon the design of
interventions. When appropriate, designs for interven-
tions were sketched on paper. Qualitative input
regarding nurses’impressions of the potential effect-
iveness, uptake and feasibility of implementation of
each solution was gathered during each discussion.
Focus group data therefore served as a form of
requirements gathering (supplemented by prior obser-
vational studies) to inform intervention design.
The resulting interventions are described below.
With the exception of the patient ID verification task,
all other tasks employed multiple applicable interven-
tions at a time (ie, interventions were employed as a
system, as shown in table 1).
Interventions for medication verification tasks (errors of
detection)
1. Verification Booth: Results of previous ethnography
revealed that nurses were interrupted 57% of the time
while verifying medication label information against the
CPOE system.
14
With this in mind, a ‘Verification Booth’
(figure 2A) was developed to provide nurses with a phys-
ically distinct quiet space to conduct verifications at com-
puter stations. The booth was a transparent enclosure
fitting around computer stations that allowed nurses to
monitor and access their patients in case of medical
emergency.
15
Strategic signage was placed on the booth
to remind passers-by of the criticality of tasks taking
place within.
2. Standardised Workflow: During preceding phases of the
study,
14
it was observed that nurses rarely followed a
standardised workflow for verifying medications prior to
reaching the patient. When interrupted, nurses often
omitted verification of medications against the CPOE,
paper order or patient’s armband. The dual paper/elec-
tronic order system used in the unit exacerbated the
potential for such omissions.
To mitigate errors resulting from these omissions,
nurses’workflow was standardised through training,
Information Technology (IT) cues, and making use of
physical space. Nurses were requested to pick up medica-
tions from the pharmacy area, and proceed directly to the
Verification Booth rather than approaching the patient
first. Nurses would then check each medication label
against the electronic order, followed by the paper order,
and would document on screen and paper that the medi-
cations had been checked. A redesigned prototype of the
CPOE software interface was created
16
that accommo-
dated a forced verification check process, and displayed
visual indicators of the status of verification of each medi-
cation. Any discrepancies would therefore be resolved
before the medications reached the point of care and had
the potential to cause harm.
3. Speaking Aloud: Nurses were asked to use a ‘Speak
Aloud’protocol when verifying medication labels against
the patient’s armband.
15
This required the nurse to ver-
balise identifying information (eg, patient’s name, date
of birth and medical record number) during verification.
Table 2 Characteristics of participants in preintervention and
postintervention experiments
Characteristic
Participants in
Phase A:
Preintervention
experiment (n=18)
Participants in
Phase C:
Postintervention
experiment (n=19)
Age
18–29 years 5 3
30–39 years 8 7
40–49 years 3 5
50–65 years 1 3
>65 years 1 1
Sex
Male 3 2
Female 15 17
Years of nursing experience
<1 year 0 0
1–10 years 11 7
11–20 years 6 8
21–30 years 0 1
>31 years 1 3
Frequency of administering chemotherapy via infusion pumps
<Once a week 1 5
1–5 times per week 12 7
2–3 times per day 0 0
>3 times per day 5 7
Original research
Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484 5
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
It was hypothesised that this action of speaking aloud
would alert patients and coworkers of the critical task at
hand, and help increase nurses’focus on the numerical
matching task. An analogous scenario would be a bank
teller counting money out loud before customers; in the
medication administration environment, the action of
speaking aloud cues patients and coworkers to wait until
the critical task is complete before asking questions or
otherwise engaging the nurses’attention.
Interventions for medication administration tasks (errors
of commission)
The following interventions were proposed for medi-
cation administration tasks
15
:
1. Visual timers for intravenous pushes: Results of a preced-
ing phase revealed that nurses lost track of time when
they were interrupted during administration of intraven-
ous push medications. This resulted in medications being
administered too quickly or too slowly, both of which
can have severe physiological consequences for
patients.
17
To mitigate such errors, it was proposed that
a visual timer (figure 2C) be attached to each intravenous
pole with the infusion pump. Rather than a numerical
stopwatch-like function, the timer counted down by pro-
portionally reducing the visual coloured indicator, with
no audible alarms or distractions. Nurses would start the
timer prior to commencing manual intravenous pushes.
2. No interruption zones with motion-activated indicators:
The immediate area surrounding infusion pump poles
was visually demarcated as a ‘No Interruption Zone’
(figure 2B). A motion-activated ‘busy’indicator was
mounted on top of the intravenous pole, and would
light up when nurses stepped in front of an intravenous
pole to hang bags, adjust tubing or program infusion
pumps. This served as an automatic indicator to
passers-by that the nurse was conducting a critical task
and should not be interrupted.
3. Speaking aloud: For the reasons listed previously (see
point 3 under Interventions for Medication Verification),
nurses were also asked to speak aloud when program-
ming infusion pumps. For instance, a nurse would say,
‘I’m programming a volume of 250 mL at a rate of
500 mL/h.’
4. Reminder signage: To aid nurses in recovering from inter-
ruptions during pump programming, and to assist them
Figure 2 Photographs depicting, (A) Verification Booth, (B) No Interruption Zones with Motion-activated Indicator, (C) Visual Timers,
(D) Reminder Signage.
Original research
6Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
in programming infusion parameters correctly even after
being interrupted, strategic signage was placed on and
near infusion pumps (figure 2D). The signage reminded
nurses to check infusion parameters, clamps and tubing
connections. The prominent presence of this signage dir-
ectly on the intravenous pole served as a visual cue,
reminding nurses to double-check infusion parameters
prior to administration.
RESULTS
Intervention utilisation
The use of some interventions (such as the
Verification Booth, No Interruption Zone,
Standardised Workflow and CPOE enhancements)
was forced upon the participant according to the
design of the physical environment. For interventions
that required active use by participants, the rate of
utilisation was as follows: Visual Timers: 100% util-
isation; Speaking Aloud during Pump Programming:
53%; Speaking Aloud during Patient Identification
Verification: 74%.
Error rates in medication verification and administration
Error rates for medication verification and administra-
tion tasks under all three experimental conditions are
shown in table 3. The results show that interruptions
were associated with a significant increase in error
rates for the following four tasks: verifying volume in
a syringe, verifying volume in an ambulatory infusion
pump, intravenous push and infusion pump program-
ming. The number of nurses committing errors in
these four tasks significantly decreased in the postin-
tervention condition. However, use of interventions
did not significantly decrease error rates for other
medication verification tasks.
DISCUSSION
To our knowledge, this is the first study to make use
of controlled high-fidelity simulation to explicitly
examine the relationship between interruptions, error
rates and the effect of interventions on medication
error rates. We identified that nurses committed sig-
nificantly more errors in infusion pump programming
and intravenous push delivery, and failed to detect
errors in several critical parameters of medication
verification when interrupted. These findings provide
important insight into understanding the contribution
of work interruptions to medication errors. More sig-
nificantly, we identified characteristics of interventions
that were effective at mitigating these error types.
Intravenous push delivery errors were significantly
reduced through use of a simple visual timer that
allowed nurses to temporally monitor the push
without requiring them to perform mental calcula-
tions of elapsed time or remember numerical starting
time values. Nurses commented that the timer display
provided an easy visual reference without detracting
from their ability to teach, monitor and care for
patients throughout the duration of the push. Nurses
were extremely eager to use the timers in their own
care environments, which is an encouraging finding
given the simple implementation and low-cost nature
of this intervention.
Similarly, pump programming errors were signifi-
cantly reduced through a combination of No
Interruption Zones, motion-activated indicators,
speak-aloud protocols and infusion pump signage.
Because our study design tested these interventions as
a system rather than individually, it is difficult to con-
clusively identify the specific mechanisms that led to
this result. Speaking aloud may have helped improve
Table 3 Error rates in medication verification and administration tasks, under all three conditions
Task
Number of nurses committing error (%)
Preintervention experiment Postintervention experiment
Condition 1:
uninterrupted (n=18)
Condition 2:
interrupted (n=18)
Significance
(Condition 1 vs 2)*
Condition 3:
interrupted (n=19)
Significance
(Condition 2 vs 3)†
Medication verification tasks (assessment of error detection)
1. Verifying medication
name
3 (17%) 6 (33%) No (p=0.160) 4 (21%) No (p=0.319)
2. Verifying medication
dosage
4 (22%) 4 (22%) No (p=0.595) 1 (5%) No (p=0.153)
3. Verifying medication
volume in syringe
9 (50%) 16 (89%) Yes (p=0.003) 11 (58%) Yes (p=0.038)
4. Verifying medication
volume in AIP
10 (56%) 17 (94%) Yes (p=0.002) 11 (58%) Yes ( p=0.012)
5. Verifying patient ID 7 (39%) 6 (33%) No (p=0.591) 6 (32%) No (p=0.593)
Medication administration tasks (assessment of error commission)
6. Intravenous push 8 (44%) 16 (89%) Yes (p=0.02) 6 (32%) Yes (p=0.001)
7. Pump programming and
infusion initiation
0 (0%) 7 (39%) Yes (p=0.03) 1 (5%) Yes (p=0.017)
*McNemar’sχ
2
test (within-subjects analysis).
†Fisher’s exact test (between-subjects analysis).
Original research
Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484 7
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
nurses’focus on pump programming parameters by
increasing the distinctiveness of the information being
verbalised,
18
and the presence of the No Interruption
Zones and associated signage may have acted as final
visual cues for nurses, reminding them to conduct one
last check of pump parameters prior to administra-
tion. Thus, a combination of environmental modifica-
tions and simple speak-aloud interventions may
provide a low-cost method of mitigating pump pro-
gramming and infusion initiation errors caused by
interruptions.
Interestingly, the speak-aloud intervention was not
effective when applied to patient identification verifi-
cation tasks. We suggest that this differential effect
may be due to the very different nature of medication
verification vs medication administration. In contrast
to the unpredictable and constantly evolving nature of
medication administration, medication verification is a
highly mechanistic and predictable task
19
that may be
more prone to habituation, confirmation bias and
complacency effects. Thus, reliance on a ‘people-
dependent’intervention such as speaking aloud may
be less effective at reducing errors because it is ultim-
ately reliant on human memory, vigilance and adher-
ence to rules.
20 21
After the experiment, some nurses
commented that they may not remember to consist-
ently speak out loud when interrupted in the real
environment, suggesting that there is a ‘ceiling effect’
to the effectiveness of this intervention. Studies
suggest that technological solutions that automate
tasks (eg, bar code medication administration
systems), force functions and relieve the memory
burden placed on humans may be more effective at
reducing adverse events,
20 21
and this automation may
be particularly well-suited to tasks that involve mech-
anistic comparison or routine checking of informa-
tion.
19 22
The real value of the speak-aloud
intervention might be in deterring people from inter-
rupting nurses. However, we were not able to evaluate
this hypothesis because all interruptions were held
constant in our experiments.
For other tasks involving mechanistic verification of
information, interventions such as the Verification
Booth and standardised workflow with CPOE
enhancements were effective at reducing wrong
volume errors in syringes and AIPs. We suggest that
our enhancements to the CPOE system (ie, forced
checks of all medication parameters and clearly visible
verification status) acted as a cueing function that
encouraged task resumption by reminding nurses of
outstanding verification items after being interrupted.
This finding is in line with research suggesting that
use of cueing functions on clinical IT systems can
encourage task resumption by reminding the user of
the task at hand.
23–25
Interestingly, the same interven-
tion was not effective at mitigating wrong medication
name and wrong dose errors. We attribute this finding
to two reasons. First, the preintervention error rate
for these two tasks was already relatively low, indicat-
ing that there was less room for improvement com-
pared with the other verification tasks. This may be the
result of nurses being more vigilant in verifying medi-
cation name and dosage compared with other medica-
tion information. Second, the limited nature of the
CPOE enhancements may have had an effect: while
the prototype incorporated layout changes and visual
cues, it did not incorporate interventions such as
TALLman lettering (eg, CARBOplatin vs CISplatin)
that specifically targeted ‘look alike, sound alike’medi-
cations. This further highlights the need for more spe-
cificity in automated interventions to reduce nurses’
reliance on vigilance and memory for error detection.
Limitations of the study
We acknowledge that there are limitations to this
study. First, participants were aware that they were
being observed during the high-fidelity simulation
experiment. It is possible that their behaviour may
have been altered as a consequence (ie, the
Hawthorne effect), though post-test debriefs suggested
that this was not a significant problem given the high
fidelity of the simulation. Second, the number of
errors planted in the simulation experiment was artifi-
cially high compared with real life, and may have
caused participants to become more vigilant for errors
as the experiment progressed. However, the order of
presentation of task types was counterbalanced to
limit this effect. Lastly, we were able to assess the
effectiveness of interventions when they were grouped
together as a system, but our study design did not
allow us to definitively assess the effectiveness of each
individual intervention. We also did not assess the lon-
gitudinal impact of interventions. Conducting these
additional assessments is a goal of future research.
CONCLUSIONS
The present research identifies that interruptions
increase the chances of nurses committing safety-
critical errors when delivering high-risk medications.
Our study adds to the literature by providing exam-
ples of low-cost interventions (eg, visual timers) that
can enhance patient safety by reducing medication
administration errors. We found that our proposed
interventions were effective at reducing errors of com-
mission in medication administration tasks, but less
effective at reducing errors of detection in medication
verification tasks. We suggest that routine, predictable
errors of detection cannot be successfully mitigated
through ‘people-dependent’interventions alone, but
would likely benefit from interventions that are more
automated and less reliant on human memory and
vigilance. Identifying and testing the effectiveness of
such interventions is a potential avenue of future
work. Because interruptions represent a highly
complex sociotechnical phenomenon
26
with poten-
tially different effects on different task types, no
Original research
8Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
single intervention is sufficient to achieve a reduction
in error. Rather, mitigation efforts must be designed
with a thorough understanding of task and error types
to be effective.
Author affiliations
1
Faculty of Medicine, Institute for Biomaterials and Biomedical
Engineering, University of Toronto, Toronto, Ontario, Canada
2
Healthcare Human Factors, Techna Institute, University Health
Network, Toronto, Ontario, Canada
3
Institute for Safe Medication Practices Canada, Toronto,
Ontario, Canada
4
Princess Margaret Cancer Centre, University Health Network,
Toronto, Ontario, Canada
5
Lawrence S. Bloomberg Faculty of Nursing, University of
Toronto, Toronto, Ontario, Canada
6
Odette Cancer Program, Sunnybrook Health Sciences Centre,
Toronto, Ontario, Canada
7
HumanEra, Techna Institute, University Health Network,
Toronto, Ontario, Canada
Acknowledgements The authors are grateful to the oncology
nurses who participated in all phases of this study. We also
thank Karin Ayanian, Michelle Dowling, Archana Gopal,
Melissa Griffin, Diane Kostka and Ilia Makedonov for their
assistance in conducting the simulation experiment.
Contributors All authors contributed to this work. VP executed
the design and testing of interventions, analysed data, and
prepared the manuscript. CK, PS, KT, JS, TM and JAC
participated in intervention development and analysis activities.
PT conceived, designed and executed the overall research study.
All authors contributed to and approved the final manuscript.
Funding This research was funded by the Canadian Patient
Safety Institute. The opinions in the present paper are those of
the authors and do not necessarily reflect the sponsor’s official
position.
Competing interests None.
Ethics approval Research Ethics Board approval for this study
was obtained from the University Health Network (Reference:
#08-0306-BE) and the University of Toronto (Reference:
#24457).
Provenance and peer review Not commissioned; externally
peer reviewed.
Open Access This is an Open Access article distributed in
accordance with the terms of the Creative Commons
Attribution (CC BY 3.0) license, which permits others to
distribute, remix, adapt and build upon this work, for
commercial use, provided the original work is properly cited.
See: http://creativecommons.org/licenses/by/3.0/
REFERENCES
1 Kohn LT, Corrigan JM, Donaldson MS. To Err is Human:
Building a Safer Healthcare System. Washington, DC: Institute
of Medicine. National Academy Press, 2000.
2 Hickam DH, Severance SS, Feldstein A, et al.The Effect of
Health Care Working Conditions on Patient Safety. Evidence
Report/Technology Assessment Number 74. Rockville, MD:
Agency for Healthcare Research and Quality (AHRQ
Publication No. 03-E031), 2003.
3 Santell JP, Hicks RW, McMeekin J, et al. Medication errors:
experience of the United States Pharmacopeia (USP)
MEDMARX reporting system. J Clin Pharmacol
2003;43:760–7.
4 Leape LL, Bates DW, Cullen DJ, et al. Systems Analysis of
Adverse Drug Events. JAMA 1995;274:35–43.
5 Cohen H, Robinson ES, Mandrack M. Getting to the root of
medication errors: Survey results. Nursing 2003;33:36–46.
6 Westbrook J, Woods A, Rob M, et al. Association of
interruptions with an increased risk and severity of medication
administration errors. Arch Intern Med 2010;170:683–90.
7 Clifton-Koeppel R. What Nurses Can Do Right Now to reduce
Medication Errors in the Neonatal Intensive Care Unit.
Newborn and Infant Nursing Reviews 2008;8:72–82.
8 Hohenhaus SM, Powell SM. Distractions and Interruptions:
Development of a Healthcare Sterile Cockpit. Newborn and
Infant Nursing Reviews 2008;8:108–10.
9 Pape TM, Guerra DM, Muzquiz M. Innovative Approaches to
Reducing Nurses’Distractions During Medication
Administration. The Journal of Continuing Education in
Nursing 2005;36:108–16.
10 Relihan E, O’Brien V,O’Hara S, et al. The impact of a set of
interventions to reduce interruptions and distractions to nurses
during medication administration. BMJ Qual & Saf 2010;19:
e52–e52.
11 Anthony K, Wiencek C, Bauer C, et al. No Interruptions Please:
Impact of a No Interruption Zone on Medication Safety in
Intensive Care Units. Critical Care Nurse 2010;30:21–9.
12 Colligan L, Guerlain S, Steck SE, et al. Designing for
distractions: a human factors approach to decreasing
interruptions at a centralised medication station. BMJ Qual &
Saf 2012;21:939–47.
13 Raban MZ, Westbrook JI. Are interventions to reduce
interruptions and errors during medication administration
effective?: a systematic review. BMJ Quality & Safety Published
Online First: 26 August 2013. doi:10.1136/bmjqs-2013-002118
14 Trbovich P, Prakash V, Stewart J, et al. Interruptions During the
Delivery of High-Risk Medications. Journal of Nursing
Administration, JONA 2010;40:211–18.
15 Prakash V, Trbovich P. Mitigating the Effects of Interruptions
During High-Risk Medication Administration. In: Poster
session: International Forum on Quality and Safety in
Healthcare. Nice, 2010.
16 Prakash V, Trbovich P. Designing Health Care IT for
Interruption-Filled Environments: Redesigning Interface
Elements of a Chemotherapy Medication System. In:
Proceedings of the 2012 Symposium on Human Factors and
Ergonomics in Healthcare: Bridging the Gap. Baltimore, MD:
Human Factors and Ergonomics Society, 2012.
17 Paparella S. Avoiding Disastrous Outcomes with Rapid
Intravenous Push Medications. J Emergency Nurs
2004;30:478–80.
18 MacLeod CM, Gopie N, Hourihan KL, et al. The production
effect: Delineation of a phenomenon. J Experiment Psychol
2010;36:671–85.
19 White RE, Trbovich PL, Easty AC, et al. Checking it twice: an
evaluation of checklists for detecting medication errors at the
bedside using a chemotherapy model. BMJ Quality & Safety
2010;19:562–7.
20 ISMP Medication Safety Alert!: Medication error prevention
‘toolbox’. 1999. http://www.ismp.org/newsletters/acutecare/
articles/19990602.asp. Institute of Safe Medication Practices
(accessed Dec 2008).
21 Taylor L, Wood SD, Chapman R. Analysis and Mitigation of
Reported Informatics Patient Safety Adverse Events at the
Veterans Health Administration. In: Proceedings of the 2012
Symposium on Human Factors and Ergonomics in Health
Care: Bridging the Gap. Baltimore, MD: Human Factors and
Ergonomics Society, 2012.
22 Parasuraman R, Sheridan TB, Wickens CD. A Model for Types
and Levels of Human Interaction with Automation. IEEE
Original research
Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484 9
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
Transactions on Systems, Man, and Cybernetics—Par t A :
Systems and Humans 2000;30:286–97.
23 Grundgeiger T, Sanderson P. Interruptions in healthcare:
Theoretical views. Int J Med Inf 2009;78:293–307.
24 Cooper R, Viller S, Burmeister J. Observations in a maternity
ward: Usability considerations in an interrupt driven
environment. In: Proceedings of the 2004 National Health
Inormatics Conference. Brisbane, Australia: Health Informatics
Society of Australia, 2004.
25 Collins S, Currie L, Bakken S, et al. Multitasking by clinicians
in the context of CPOE and CIS use. Stud Health Technol
Inform 2007;129:958–62.
26 Coiera E. The Science of Interruption. BMJ Qual & Saf
2012;21:357–60.
Original research
10 Prakash V, et al.BMJ Qual Saf 2014;0:1–10. doi:10.1136/bmjqs-2013-002484
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
doi: 10.1136/bmjqs-2013-002484
published online June 6, 2014BMJ Qual Saf
Varuna Prakash, Christine Koczmara, Pamela Savage, et al.
ambulatory chemotherapy setting
administration: interventions in a simulated
during medication verification and
Mitigating errors caused by interruptions
http://qualitysafety.bmj.com/content/early/2014/06/06/bmjqs-2013-002484.full.html
Updated information and services can be found at:
These include:
Data Supplement http://qualitysafety.bmj.com/content/suppl/2014/06/06/bmjqs-2013-002484.DC1.html
"Supplementary Data"
References http://qualitysafety.bmj.com/content/early/2014/06/06/bmjqs-2013-002484.full.html#ref-list-1
This article cites 18 articles, 5 of which can be accessed free at:
Open Access
http://creativecommons.org/licenses/by/3.0/
commercial use, provided the original work is properly cited. See:
others to distribute, remix, adapt and build upon this work, for permitsof the Creative Commons Attribution (CC BY 3.0) license, which
This is an Open Access article distributed in accordance with the terms
P<P Published online June 6, 2014 in advance of the print journal.
service
Email alerting the box at the top right corner of the online article.
Receive free email alerts when new articles cite this article. Sign up in
(DOIs) and date of initial publication.
publication. Citations to Advance online articles must include the digital object identifier
citable and establish publication priority; they are indexed by PubMed from initial
typeset, but have not not yet appeared in the paper journal. Advance online articles are
Advance online articles have been peer reviewed, accepted for publication, edited and
http://group.bmj.com/group/rights-licensing/permissions
To request permissions go to:
http://journals.bmj.com/cgi/reprintform
To order reprints go to:
http://group.bmj.com/subscribe/
To subscribe to BMJ go to:
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
Collections
Topic
(175 articles)Open access
Articles on similar topics can be found in the following collections
Notes
(DOIs) and date of initial publication.
publication. Citations to Advance online articles must include the digital object identifier
citable and establish publication priority; they are indexed by PubMed from initial
typeset, but have not not yet appeared in the paper journal. Advance online articles are
Advance online articles have been peer reviewed, accepted for publication, edited and
http://group.bmj.com/group/rights-licensing/permissions
To request permissions go to:
http://journals.bmj.com/cgi/reprintform
To order reprints go to:
http://group.bmj.com/subscribe/
To subscribe to BMJ go to:
group.bmj.com on September 29, 2014 - Published by qualitysafety.bmj.comDownloaded from
