doi:10.1111/jpc.14193
ORIGINAL ARTICLE
Medication error trends and effects of person-related,
environment-related and communication-related factors on
medication errors in a paediatric hospital
Elizabeth Manias ,1,2,3 Noel Cranswick,4,5 Fiona Newall,3,5,6 Ellie Rosenfeld,3 Carlye Weiner,3 Allison Williams,7
Ian CK Wong,8 Narelle Borrott,3 Jerry Lai9 and Sharon Kinney3,6
1
School of Nursing and Midwifery, Centre for Quality and Patient Safety Research, Deakin University, Sydney, New South Wales, 2Department of Medicine,
Royal Melbourne Hospital, 3Melbourne School of Health Sciences, and 6Department of Paediatrics, Royal Children’s Hospital, University of Melbourne,
4
Department of Medicine, Clinical Pharmacology Unit, and 5Murdoch Childrens Research Institute, Royal Children’s Hospital, 7School of Nursing and
Midwifery, Monash University, Melbourne, 9eResearch, Deakin University, Geelong, Victoria, Australia and 8Research Department of Practice and Policy,
School of Pharmacy, University College London, London, United Kingdom
Aim: This study aimed to examine reported medication error trends in an Australian paediatric hospital over a 5-year period and to determine
the effects of person-related, environment-related and communication-related factors on the severity of medication outcomes. In particular, the
focus was on the influence of changes to a hospital site and structure on the severity of medication errors.
Methods: A retrospective clinical audit was undertaken over a 5-year period of paediatric medication errors submitted to an online voluntary
reporting system of an Australian, tertiary, public teaching paediatric hospital. All medication errors submitted to the online system between
1 July 2010 and 30 June 2015 were included.
Results: A total of 3340 medication errors was reported, which corresponded to 0.56% medication errors per combined admissions and presentations or 5.73 medication errors per 1000 bed days. The most common patient outcomes related to errors requiring monitoring or an intervention to
ensure no harm occurred (n = 1631, 48.8%). A new hospital site and structure had 0.354 reduced odds of producing medication errors causing possible
or probable harm (95% confidence interval 0.298–0.421, P < 0.0001). Patient and family involvement had 1.270 increased odds of identifying medication
errors associated with possible or probable harm compared with those causing no harm (95% confidence interval 1.028–1.568, P = 0.027). Interrupted
time series analyses showed that moving to a new hospital site and structure was associated with a reduction in reported medication errors.
Conclusion: Encouraging child and family involvement, facilitating hospital redesign and improving communication could help to reduce the
harm associated with medication errors.
Key words: family involvement; hospitalised child; interdisciplinary communication; medication error; patient involvement.
What is already known on this topic
What this paper adds
1 Children are at high risk of experiencing medication errors.
2 Past research has been conducted on identifying the prevalence
of medication errors in children; however, little is known about
the causes, human factors, contributing factors and patient outcomes of medication errors, as well as the involvement of
patients and family members in detecting medication errors.
3 Past research shows that common types of medication errors
include overdose or under-dose, dose omission, mistakes in administration techniques, wrong medication, wrong person and inattention
to children’s changing needs, including alterations in body weight.
1 Hospitalised children or family members alerted health professionals
about the occurrence of medication errors in 15% of occasions.
2 Patient and family involvement was associated with increased odds
of identifying medication errors causing possible or probable harm,
while a new hospital site and structure was associated with reduced
odds of medication errors causing possible or probable harm.
3 Interrupted time series results showed that a new hospital site
and structure was associated with a reduced percentage of
medication errors per combined admissions and presentations
and reduced medication errors per 1000 bed days.
Correspondence: Professor Elizabeth Manias, School of Nursing and
Midwifery, Centre for Quality and Patient Safety Research, Deakin University, 221 Burwood Highway, Burwood, Vic. 3125, Australia; email:
[email protected]
Conflict of interest: None declared.
Accepted for publication 22 July 2018.
320
Children are at high risk of experiencing medication errors.
They are vulnerable due to their limited ability to communicate,
especially those who are preverbal and have difficulties in
expressing their needs. Insufficient availability of paediatric formulations contributes to the need for increased medication
Journal of Paediatrics and Child Health 55 (2019) 320–326
© 2018 The Authors Journal of Paediatrics and Child Health published by John Wiley & Sons Australia, Ltd on behalf of Paediatrics and Child Health Division (The Royal Australasian
College of Physicians)
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
E Manias et al.
compounding, and a lack of information from manufacturers
about preparing these medications can also lead to medication
errors.1
Extensive research has been conducted on identifying the
prevalence of medication errors in children. Medication error
rates differ considerably due to varying contexts of care and
methods used for analysis. Variations have ranged between 0.51
per 1000 bed days2 and 82.9 medication errors per 1000 bed
days.3
In attempting to reduce medication errors, increased focus
has been placed on person-related factors, such as addressing
how nurses, doctors and pharmacists contribute to these errors.
While past research has shown that health professionals can
influence medication errors,4 less is known about how child
and family involvement may prevent and detect medication
errors. A retrospective clinical audit conducted in Australia
demonstrated that, in 8.0% of reported medication errors
(n = 219), children and family members had alerted health professionals that a medication error had occurred.5 Further work
is needed to examine child and family involvement in influencing medication errors.
Major causes of medication errors involve communicationrelated factors,5,6 with research showing that communication
problems account for over half of all causes associated with medication errors.7 Difficulties with written communication feature
predominantly, involving illegible and misinterpreted medication
orders. Conversely, details about how communication processes,
such as ward rounds and handovers, affect medication errors are
largely missing.
Environment-related factors can have an enormous impact on
patient safety. A case study approach was undertaken in the
USA8 where extensive consultation was sought from architects,
patients, family members, managers and clinicians about what
they felt was important in designing a new hospital. Through
these consultations, a comprehensive checklist was created to
guide the development of the new hospital and enable a culture
of patient safety. In this study, there was no information about
how environment-related factors affected medication errors. The
hospital targeted was an acute care facility with no details about
specific provisions for paediatric care.
Examining the complexities of diverse factors associated with
medication errors can help identify how these factors influence
the severity of harm and ascertain possible strategies for
improved patient safety. The aims of the study were to examine reported medication error trends in an Australian paediatric hospital over a 5-year period and to determine the effects
of person-related, environment-related and communicationrelated factors on the severity of medication outcomes. In particular, the key intent was to focus on the influence of
changes to a hospital site and structure on the severity of
medication errors.
Methods
A retrospective clinical audit was undertaken over a 5-year
period of paediatric medication errors submitted to an online voluntary reporting system of an Australian, tertiary, public, teaching paediatric hospital. The hospital, comprising 334 beds,
provides a comprehensive range of clinical services, tertiary care,
Medication error trends and effects
health promotion and prevention programmes. It contains an
emergency department, an intensive care unit and a state-wide
trauma service. The human research ethics committee of the hospital approved the conduct of the study. A medication error was
defined as any preventable event that may have caused or led to
inappropriate medication use or patient harm while the medication was in the control of the health professional, patient or family member.9
On 26 October 2011, during the study period, the hospital had
moved from an old site to a new building complex. On that day,
all patients were moved from the old hospital site to the new
hospital site, including all health professionals involved in providing their care. Alongside the move, there were structural changes
associated with the new hospital complex. In the new hospital,
most hospitalised children were situated in single rooms, with
dedicated spaces for family members. These dedicated spaces
were created to facilitate family-centred care and engagement
with clinicians. Dedicated medication rooms were developed for
all wards in the new hospital (Fig. 1). In the old hospital, only
one medical ward had a dedicated medication room, whereas in
all other wards, the medication room was combined with the
patient treatment room (Fig. 2). In the old hospital, when the
medication room was combined with the patient treatment room,
there was extensive activity, with children undergoing procedures such as peripheral intravenous cannula insertion, venepuncture and bladder catheterisation, which sometimes led to
distractions from excessive noise and interruptions. In both forms
of the medication room, whether combined with the treatment
room in the old hospital or as a stand-alone facility in the new
hospital, pharmacists dispensed medications, physicians ordered
medications and nurses prepared and administered medications.
In the new hospital, the nurse–patient ratio was the same, each
unit had the same staff, and guidelines and policies were
the same.
Data collection and procedure
All medication errors submitted to the online system between
1 July 2010 and 30 June 2015 were included. Children’s medical
records were accessed. For many years, hospital management
adopted a no-blame policy, actively encouraging clinicians to submit medication errors as they occurred. As a result, reported
medication errors contained detailed descriptions of medications
errors in the form of free text as well as comprehensive responses
to close-ended questions. A medication safety committee developed the summary data of medication errors, and those with
severe outcomes were discussed and analysed as they occurred.
Feedback was incorporated in online medication error reports on
patient outcomes and in strategies adopted to address recurring
problems and causes.
The National Coordinating Council for Medication Error
Reporting and Prevention tool was used as the data collection
instrument.9 This tool provides a structured and standardised
approach in categorising medication errors. The tool enabled
person-, environment- and communication-related factors to be
systematically classified.
Journal of Paediatrics and Child Health 55 (2019) 320–326
© 2018 The Authors
Journal of Paediatrics and Child Health published by John Wiley & Sons Australia, Ltd on behalf of Paediatrics and Child Health Division (The Royal Australasian College of
Physicians)
321
Medication error trends and effects
E Manias et al.
Fig. 1 Medication room in the new
hospital.
Fig. 2 Medication room combined
with the patient treatment room in
the old hospital.
Data analysis
Descriptive and inferential data analysis was undertaken using
SPSS, version 23 (IBM Corp., Chicago, IL, USA). Frequency
counts and percentage counts were calculated for all variables of
interest. Calculations of medication errors per combined admissions and presentations and medication errors per 1000 bed days
were obtained from resourcing the statistical data of the hospital’s
annual reports. These annual reports contained data on the number of patient admissions and presentations and the number of
bed days over the 5-year period and were used for data collection
and analysis.
Univariate associations with severity of medication errors were
examined for all person-related, environment-related and
communication-related factors using cross tabulations and chi
square tests. The outcome or dependent variable was the patient
322
outcome measure (scores of 1–3 relating to no patient harm and
scores 4–8 relating to possible or probable patient harm).9
Person-related factors included the individual responsible for the
error (nurse/doctor/pharmacist/patient/family member) and
patient and family involvement in identifying a medication error
(yes/no). Environment-related factors included: the hospital site
and structure (new/old) and clinical site of the medication errors.
Communication-related factors included: informal bedside communication problem (yes/no), clinical handover problem
(yes/no) and medical record documentation problem (yes/no).
These communication-related factors were selected because they
were the most commonly occurring data. Use of the new hospital
site and structure commenced on 26 October 2011. Thus, medication errors taking place before 26 October 2011 occurred in the
old hospital site and structure, while those taking place from
26 October 2011 onwards occurred in the new hospital site and
Journal of Paediatrics and Child Health 55 (2019) 320–326
© 2018 The Authors
Journal of Paediatrics and Child Health published by John Wiley & Sons Australia, Ltd on behalf of Paediatrics and Child Health Division (The Royal Australasian College of
Physicians)
E Manias et al.
Medication error trends and effects
structure. Binary multiple logistic regression modelling was subsequently undertaken, with factors demonstrating univariate
associations of less than or equal to P = 0.25. Use of a less stringent P value of 0.25, compared with a P value of 0.05, was
Table 1 Clinical settings in which medication errors occurred
(n = 3340)
Clinical setting
Intensive care unit
Medical unit
Surgical unit
Neonatal intensive care unit
Cardiac unit
Psychiatry unit
Oncology unit
Neuroscience unit
Adolescent unit
Pharmacy department
Emergency department
Perioperative area
Immunology department
Hospital-in-the-home
Medical short stay
Surgical short stay
Radiology
Outpatient units
Table 2
n
%
462
441
362
354
269
255
192
158
142
139
128
124
106
79
70
44
10
5
13.8
13.2
10.8
10.6
8.1
7.6
5.7
4.7
4.3
4.2
3.8
3.7
3.2
2.4
2.1
1.3
0.3
0.1
arbitrarily chosen to safeguard against excluding potentially
important factors at the univariate level when determining what
factors to test in the binary logistic modelling level.10 Binary
logistic regression modelling involved calculating the ability of
various explanatory variables to increase or decrease the odds of
a binary categorical outcome, which can only have two values. In
this case, the values of the binary outcome were the presence of
medication errors causing possible or probable harm and those
causing no harm. The level of significance used for logistic modelling was α = 0.05.
The impact of the new hospital site and structure was initially
explored through an interrupted time series analysis that
included both the step-change (hospital site, old = 0; new = 1)
and slope-change (time) components and the interaction term.
However, because the contribution of the slope-change and
interaction components was statistically non-significant, our
final model retained only the step-change component.11 Due to
over-dispersion, all analyses were conducted using the quasiPoisson model. There was no clear sign of dependencies
(i.e. autocorrelation) or seasonality; thus, no further adjustments were made.
Results
During the study period, a total of 3340 medication errors was
reported, which corresponded to 0.56% per combined admissions
and presentations or 5.73 medication errors per 1000 bed days.
In addition, there were 6680 human factors and 4008 causes
Patient outcomes of medication errors relating to eight categories of harm (n = 3340 medication errors)
Patient outcome
n
%
1 Circumstances had the capacity to cause error
2 Error occurred, but the error did not reach the patient
14
618
0.4
18.5
3 Error occurred that reached the patient but did not
cause patient harm
979
29.3
1631
48.8
5 Error occurred that resulted in temporary harm and
required intervention
89
2.7
6 Error occurred that resulted in temporary harm and
required prolonged hospitalisation
8
0.2
7 Error occurred that resulted in permanent patient
harm
8 Error occurred that required intervention necessary to
sustain life
0
0
1
0.1
4 Error occurred that reached the patient and required
monitoring or intervention to confirm no harm
Examples of patient outcome
Warfarin dose not signed on drug chart
600 mcg of naloxone was prescribed on the medication
chart instead of 60 mcg – 10× overdose. Patient did
not receive the dose of naloxone
Weight recorded as 23 kg when weighed in day surgery
admission; on inspection, the patient obviously
weighed less than this, but issue was not detected until
patient was anaesthetised and after premedication
given comprising midazolam and paracetamol
(acetaminophen)
Intravenous order for intravenous ticarcillin + clavulanic
acid in emergency department but not handed over.
Intravenous antibiotics commenced about 14 h late
Noradrenaline (norepinephrine) found to be
disconnected following severe hypotension and need
of volume and increase of inotrope requirements
following a bed turn
Patient’s intravenous antibiotic, flucloxacillin, changed to
oral antibiotic but continued to be given as intravenous
form, delaying patient discharge
Not applicable
Patient given overdose of arginine. Patient was
transferred to paediatric intensive care unit for
hemofiltration and treatment of the overdose
Journal of Paediatrics and Child Health 55 (2019) 320–326
© 2018 The Authors
Journal of Paediatrics and Child Health published by John Wiley & Sons Australia, Ltd on behalf of Paediatrics and Child Health Division (The Royal Australasian College of
Physicians)
323
Medication error trends and effects
Table 3
E Manias et al.
Causes associated with medication errors (n = 4008 causes)
Cause of medication error
n
Communication-related factors, n = 3340 (83.3%)
Medical record documentation
717
Clinical handover
571
Informal bedside communication
542
Misinterpretation of order
518
Misread or unread order
481
Illegible handwriting
235
Decimal point
107
Units of measurement
91
Telephone communication
56
Ward round
22
Confusion with medication name, n = 229 (5.7%)
Generic name confusion
203
Brand name confusion
26
Confusion with labelling and reference material, n = 439 (11.0%)
Label of dispensed product is wrong
161
Problems with reference material
142
Container of manufacturer similar or
136
confusing
%
17.9
14.2
13.5
12.9
12.0
5.9
2.7
2.3
1.4
0.5
5.1
0.6
4.0
3.5
3.4
identified in the database during the study period. The median
age of children was 4.3 years (interquartile range 4 months,
13.5 years), and 30.4% were children aged 12 months or younger. Of the reported medication errors, 43.5% occurred in female
children. Table 1 shows the clinical settings in which medication
errors occurred.
Most medication errors occurred during the daytime between
0700 and 1530 (n = 1736, 52.0%). However, considerable medication errors also occurred after business hours and overnight.
Medication errors took place predominantly during weekdays,
but 20.6% of medication errors occurred during weekend days.
The most common patient outcomes related to errors requiring
monitoring or an intervention to ensure that no harm occurred
(n = 1631, 48.8%), followed by errors reaching patients that did
not cause patient harm (n = 979, 29.3%) (Table 2).
Table 3 shows communication-related factors of medication
errors, details about generic and trade name confusion and concerns with labelling and reference materials. A total of 4008
causes were attributed to medication errors. The most common
communication-related factor was problems with medical record
documentation (n = 717, 17.9%), which included health professionals not checking for documented allergies and confusion
about children’s actual recorded weight. Labels of dispensed
products sometimes had incorrect details documented, but the
contents of the products were correct (e.g. identification of vehicles on intravenous fluid bags with additives). At other times, the
labels and the contents were incorrect. Problems with reference
materials included limitations associated with available medication resources, leading to delays in administration as clinicians
attempted to obtain further information (n = 142, 3.5%). Examples included availability of injectable guidelines for powder volumes of methylprednisolone for established brands but not for
new brands and the lack of administration instructions for cyclophosphamide and mesna.
324
Table 4 Human factors and contributing factors (n = 6680)
Type of factors
Human factors, n = 3340
Performance deficit
Knowledge deficit
Inadequate screening of patient
Miscalculation of dose or infusion rate
Failure to activate delivery system properly
Wrong amount of active medication used
Error in stocking
Wrong diluent used for infusion
Stress
Wrong amount of diluent used
Wrong medication added to infusion
Intimidating behaviour
Contributing factors, n = 3340
Policies and procedures
Frequent interruptions and distractions
Communication relating to patient
movements
Lack of available trained health
professionals
Inadequate training provided
Insufficient or incorrect counselling offered
to patients or parents
Floor stock
n
%
1436
504
445
341
173
126
108
69
55
48
27
8
21.5
7.5
6.7
5.1
2.6
1.9
1.6
1.0
0.8
0.7
0.4
0.1
1619
639
461
24.2
9.6
6.9
216
3.2
155
126
2.3
1.9
124
1.9
Commonly occurring human factors included performance
deficits (n = 1436, 21.5%) and knowledge deficits (n = 504,
7.5%). Contributing factors that frequently occurred included clinicians not adequately following policies and procedures
(n = 1619, 24%). Problems with patients’ movements across
transitions, such as transfers involving emergency department
and operating rooms, were also evident (n = 461, 6.9%)
(Table 4).
Bivariate logistic regression modelling identified 10 factors that
significantly predicted associations with children experiencing
possible or probable harm versus no harm from medication errors
(Table 5). In situations where doctors or pharmacists were
responsible for the medication error, there were reduced odds of
a harmful medication error occurring compared to when nurses
were responsible. Involvement of children and families in identifying medication errors had 1.270 increased odds of harmful
medication errors occurring compared with those who were not
involved. Problems with clinical handover, informal bedside communication and medical record documentation all increased the
odds of harmful medication errors. The new hospital site and
structure showed 0.354 reduced odds in being associated with
harmful medication errors compared with the old hospital site
and structure. The intensive care unit, oncology unit and emergency department all demonstrated increased odds of having
harmful medication errors compared to medical units.
Figure 3 shows the interrupted time series trends for the effect
of the new hospital site and structure on reported medication
errors. The risk of percentage medication errors per combined
admission and presentation was reduced by 35.4% (relative risk
Journal of Paediatrics and Child Health 55 (2019) 320–326
© 2018 The Authors
Journal of Paediatrics and Child Health published by John Wiley & Sons Australia, Ltd on behalf of Paediatrics and Child Health Division (The Royal Australasian College of
Physicians)
E Manias et al.
Table 5
Medication error trends and effects
Binary logistic regression model for explanatory factors associated with medication errors causing possible or probable harm (n = 3340)
Variable
Person-related factors
Person responsible for the medication error
Person responsible for the medication error
Patient and family involvement in detecting an
incident
Environment-related factors
New hospital site and structure
Clinical setting where error occurred
Clinical setting where error occurred
Clinical setting where error occurred
Communication-related factors
Problem with informal bedside communication
Problem with medical record documentation
Problem with clinical handover
Reference level
Comparator level
Odds ratio
95% confidence
intervals
P value
Nurse
Nurse
No
Doctor
Pharmacist
Yes
0.571
0.410
1.270
0.483–0.674
0.285–0.591
1.028–1.568

Purchase answer to see full
attachment