International Journal of Industrial Ergonomics 67 (2018) 171–179

Contents lists available at ScienceDirect

International Journal of Industrial Ergonomics

journal homepage: www.elsevier.com/locate/ergon

Application of HFACS, fuzzy TOPSIS, and AHP for identifying important T

human error factors in emergency departments in Taiwan
Min-chih Hsieha, Eric Min-yang Wanga, Wui-chiang Leeb,∗, Lun-wen Lia, Chin-yi Hsiehc,
Weide Tsaid, Chih-ping Wangd, Ju-li Huangd, Te-chu Liud
a
Department of Industrial Engineering and Engineering Management, National Tsing Hua University, 101, Section 2, Kuang-Fu Road, Hsinchu, 30013, Taiwan, ROC
b
Department of Medical Affairs and Planning, Taipei Veterans General Hospital & National Yang-Ming University School of Medicine, 201, Section 2, Shipai Road, Beitou
District, Taipei City, 11217, Taiwan, ROC
c
Emergency Department, Hsinchu MacKay Memorial Hospital, 690, Section 2, Kuang-Fu Road, Hsinchu, 30071, Taiwan, ROC
d
Emergency Department, MacKay Memorial Hospital, 92, Section 2, Zhongshan N. Rd., Taipei City, 10449, Taiwan, ROC

A R T I C LE I N FO A B S T R A C T

Keywords: In previous studies, the frequency of error factors associated with medical adverse events seems to be the only
Human errors analysis criterion for understanding the distribution of error factors in hospitals. However, the types of error that occur
HFACS most frequently in hospitals are not necessarily the most important. Therefore, this study integrated human error
AHP analysis and fuzzy TOPSIS to reconcile this discrepancy. The purpose of the study is to identity the important
TOPSIS
human error factors in emergency departments (ED) in Taiwan. Human factors analysis and classification system
Patient safety
(HFACS) was used to analyze 35 ED adverse events to define the error factors. Multiple criteria decision making
(MCDM) methods such as analytic hierarchy process (AHP) and fuzzy Technique for Order Preference by
Similarity to Ideal Solution (TOPSIS) were applied to evaluate the importance of error factors. Results showed
that decision errors, crew resource management, inadequate supervision, and resource management were the
important human error factors related to ED adverse events. This study recommends that MCDM should be
applied to further analyze the results based on the criteria.

1. Introduction (Taiwan Patient-safety Reporting System, 2016). Based on the litera-

Human error is one of the common factors contributing to the ma-
jority of incidents and accidents occurred within complex systems
(Wenner and Drury, 2000; Liang et al., 2010), such as a medical system.
According to the report of the Institute of Medicine in 1999 (Kohn et al.,
2000), medical adverse events might be responsible for 44,000 to
98,000 deaths and more than 1 million injuries in the United States
hospitals per year. Since the estimate data is nearly three decades old,
James (2013) updated the estimation using data from several studies
published between 2008 and 2011. The results showed that, annually,
at least 210,000 deaths were associated with preventable harm in
hospitals. Similarly, a national survey conducted in France revealed
that nearly 10,000 deaths were potentially related to medical adverse
events whilst half of them could have been prevented with appropriate
care (Freund et al., 2013). Regarding to the states of medical adverse
events in Taiwan, the annual report from Taiwan Patient-safety Re-
porting system (TPR) showed that 56,297 adverse events occurred in
2015 and almost a third of the cases resulted in harm to patients
tures reviewed above, it is not surprising that medical adverse events
are terribly common in hospitals.
1.1. Medical adverse events in emergency department
Medical adverse events are common in emergence departments (ED)
and often lead to severe outcomes (Rothschild et al., 2010). Several
studies indicated that at least 3% of all adverse events occurred in the
ED (Calder et al., 2010; Stang et al., 2013). Thus, ED has been identified
as a hospital location where adverse events are highly attributable to
errors (Fordyce et al., 2003; Calder et al., 2010; Stang et al., 2013). The
issues mentioned above are mainly resulted from disrupted sleep cycles,
multiple interruptions, acute time constrains, patient acuity and com-
plexity, and high patient volume and overcrowding (Chisholm et al.,
2000; Trzeciak and Rivers, 2003; Fordyce et al., 2003; Epstein et al.,
2012; Stang et al., 2013). Therefore, many studies have analyzed ad-
verse events in ED and tried to understand the mechanism and com-
position of factors contributing to errors. Fordyce et al. (2003)

∗
Corresponding author.
E-mail addresses: g9674019@cycu.org.tw (M.-c. Hsieh), mywangeric@gmail.com (E.M.-y. Wang), wclee@vghtpe.gov.tw (W.-c. Lee), lilonwen@gmail.com (L.-w. Li),
jis876s@gmail.com (C.-y. Hsieh), weidetsai@gmail.com (W. Tsai), chihpin@ms1.mmh.org.tw (C.-p. Wang), ruby@mmh.org.tw (J.-l. Huang), cv1258.3396@mmh.org.tw (T.-c. Liu).

https://doi.org/10.1016/j.ergon.2018.05.004
Received 31 October 2017; Received in revised form 25 April 2018; Accepted 17 May 2018
0169-8141/ © 2018 Elsevier B.V. All rights reserved.
M.-c. Hsieh et al.
identified 346 errors which occurred in ED and categorized them into
diagnostic studies, administrative procedures, pharmacotherapy, doc-
umentation, communication, and environmental maintenance based on
the error types. Friedman et al. (2008) used three types of error cate-
gories, including adverse event, near miss, and medical error to classify
the errors in ED. However, these error types were the phenomena re-
sulting from the mishaps in emergency treatment process, which could
not represent the causes of the errors (Reason, 1990). Thus, in order to
reduce the ED adverse events, an overall understanding of the reasons
behind the phenomena is necessary.
There are two types of human errors, namely, active human errors
and latent human errors. Reason (1995) indicated that active human
errors lead to accidents with immediate influence. Latent human errors
result in accidents indirectly, that is; the adverse consequences may
hide within the system. The errors only become obvious when com-
bining with other factors to breach the defenses of a system. Cosby
(2003) constructed a framework for classifying the error factors in ED.
The error types of this framework involved individual, teamwork,
working environment, and management issues. In more detail, in-
dividual issues in this framework included skill-set errors, task-based
errors, and personal impairment. Teamwork issues implied teamwork
failure in ED; working environment issues included the environment in
ED and in hospitals; management issues meant administration problem
in hospital. Although the framework contains active and latent error
factors, the classifying factors were still insufficient to help researchers
to fully understand the causes of errors.
1.2. Human factors analysis and classification system
Human factors analysis and classification system (HFACS) is derived
from Reason's Swiss cheese model (Reason, 1990, 1997); it provides an
organizational framework for accident analysis (Daramola, 2014). In
HFACS, errors are divided into four categories, including unsafe acts,
preconditions for unsafe acts, unsafe supervision, and organizational
influence (Wiegmann and Shappell, 2001, 2003).
The Framework of HFACS was shown in Fig. 1. The “unsafe acts”
category is the research focusing on accidents. The behavior of opera-
tors that directly lead to and/or form active errors in accidents in
medical processes are described. Taking decision errors for example,
lack of patient information, professional medical knowledge, and ex-
perience may result in decision errors. The latent errors in accidents
lead to the errors in the category of “predictions of unsafe acts”, which
are considered as the psychological precursors of the failures in the
category of “unsafe acts”. For example, physical/mental limitations,
referring to medical operations that go beyond the operator's control,
occurs when an operator takes on an operation required more experi-
ence than he has or she is unfamiliar with its treatment.
The category of “unsafe supervision” also relates to latent errors. For
this category, the causes of an unsafe act can be traced back to the level
of frontline supervisions. For instance, inadequate supervision occurs in
the situation where a manager does not share experience, supervise,
monitor, and train team members when necessary or provides in-
sufficient support throughout. The category of “organizational influ-
ences” is considered as a latent error category in accident analysis. It is
commonly related to faulty decisions with direct impact on supervisory
practices at management level. For example, resource management
refers to decisions, made by decision makers of the highest level in an
organization, on overall distribution of assets, such as staff establish-
ment and training and purchase of equipment.
Li and Harris (2006) analyzed 523 accidents in the Republic of
China (ROC) Air Force between 1978 and 2002 through the application
of the HFACS framework. The results revealed several key relationships
between errors at the operational level and organizational inadequacies
at both the immediately adjacent level (preconditions for unsafe acts)
and higher levels in the organization (unsafe supervision and organi-
zational influences). Diller et al. (2013) applied HFACS to investigate
172
International Journal of Industrial Ergonomics 67 (2018) 171–179
105 medical adverse events, over 1700 errors occurred in operational
and organizational level were identified. The structure of HFACS has
been widely used in studies analyzing aviation accidents and other
fields (Li et al., 2008; Chauvin et al., 2013; Daramola, 2014; Chiu and
Hsieh, 2016; Madigan et al., 2016). HFACS framework contains errors
categories related to technical operation, staff management, and orga-
nization operation; consequently, the full picture of an incidence can be
revealed through analyzing human errors. The current research,
therefore, applies HFACS framework to investigate medical adverse
events in emergency departments in Taiwan.
1.3. Multi-criteria decision-making method
In the previous studies, the frequency of error factors associated
with medical adverse events seems to be the only criterion for under-
standing the distribution of error factors in hospitals. For instance,
Fordyce et al. (2003) analyzed the adverse events in ED and recorded
the frequency of error factors of six categories. Lisby et al. (2005)
identified several error types and recorded the frequency of the errors
from the medication adverse events. Freund et al. (2013) analyzed ED
adverse events and recorded the frequency of errors of five categories.
However, the types of error that occur most frequently in hospitals are
not necessarily the most important ones. Important cross-criteria error
factors such as preventability and reproducibility were absent in the
previous studies. Wang and Chou (2015) assessed the management
issue related to patient safety in hospital by multi-criteria decision-
making (MCDM) method. Chiu and Hsieh (2016) identified the human
error factors in aviation maintenance tasks with HFACS and success-
fully applied one of the MCDM method with four criteria to generate
the efficient improvement strategies.
Technique for Order Preference by Similarity to Ideal Solution
(TOPSIS) is a well-known MCDM method. It has been commonly used
to solve decision-making problems in many different research fields,
such as aviation safety, supply chain management, healthcare, chemical
engineering, and business and marketing management (Behzadian
et al., 2012; Kuo et al., 2012; Kannan et al., 2014; Chiu and Hsieh,
2016). TOPSIS, proposed by Hwang and Yoon (1981), considers the
performance of alternatives while taking multiple criteria into account
at the same time (Bai et al., 2014). Its primary concept is that the
chosen alternative should have the shortest geometric distance from the
positive ideal solution and the longest geometric distance from the
negative ideal solution. However, in a decision-making process, TOPSIS
was insufficient to deal with the vagueness or ambiguity problems
(Kannan et al., 2014; Chiu and Hsieh, 2016). Being aware of this lim-
itation, we have incorporated TOPSIS with fuzzy set theory. By doing
so, the decision-makers are able bring unquantifiable, incomplete, or
non-obtainable information and partially uncertain facts into the deci-
sion model (Kannan et al., 2014; Chiu and Hsieh, 2016).
Previous research has pointed out that the real word system cannot
be fully represented by the data of crisp numbers (Kannan et al., 2014;
Chiu and Hsieh, 2016). This insufficiency results from the vagueness,
imprecision as well as the subjective nature of human reasoning, jud-
gement, and preferences. To bridge the gap, the fuzzy set theory was
developed to model the uncertainty of human judgement. The fuzzy set
theory represents the selection of decision-makers by linguistic values.
The selections are then converted to fuzzy numbers so that the MCDM
problem is dealt with. What is more, triangular fuzzy numbers (TFN)
have been taken as an effective approach to formulating decision issues
related to subjective and imprecise information (Chang and Yeh, 2002;
Chang et al., 2007; Torlak et al., 2011). Accordingly, the fuzzy set
theory and TOPSIS were jointly applied to identify the important error
factors to improve patient safety in ED, and TFN were used to evaluate
the selections of decision-makers.
Additionally, analytic hierarchy process (AHP), developed by Saaty
(1990), illustrates how to determine the relative importance of alter-
natives in MCDM problem. The advantage of this method is easy to
M.-c. Hsieh et al.
Fig. 1. The framework of HFACS (Wiegmann and Shappell, 2003).
understand; it can efficiently handle both qualitative and quantitative
data (Kahraman et al., 2003). AHP method is based on three principles,
namely, structure the model, comparative judgment of the alternatives
and the criteria, and synthesis of the priorities (Dağdeviren et al.,
2009).
AHP has been widely used in solving many complicated MCDM
problems. It can also be used along with other processes (i.e. TOPSIS) to
analyze complicated decision-making problems. Dağdeviren et al.
(2009) utilized AHP to determine the weight values of criteria in
TOPSIS, this approach made the weighting process more systematic and
objective. Accordingly, the current study adopts AHP to determine the
weight of each criterion and TOPSIS was applied to identify the im-
portant error factors based on different criteria.
1.4. The objective of this study
Based on the literature reviewed above, the objective of this study is
to analyze the ED adverse events and identify the error factors by
HFACS, and to evaluate the importance of the error factors by fuzzy
TOPSIS and AHP. By integrating HFACS, fuzzy TOPSIS, and AHP, the
results of this study can not only pinpoint the error factors among the
ED adverse events but also identify the important error factors based on
the different criteria in each category of HFACS rather than using one
criterion to determine the importance of the error factors. It is expected
that this research may enhance depth and broadness of the metho-
dology for medical adverse events analysis and improve patient safety
as well as the reliability of healthcare system.
2. Method
2.1. ED adverse events selection
This is a retrospective study that focuses on adverse events in ED
173
International Journal of Industrial Ergonomics 67 (2018) 171–179
and cooperating with hospitals is necessary. Thus, the study involves a
hospital, which is also a medical center, with averagely 300–400 ED
patients per day in Taipei. The electronic medical records (EMRs) re-
lated to ED patients were provided by the hospital for the conduct of
this research. This study was approved by the Institutional Review
Board (approved number: 14MMHIS212) of the hospital.
This study included all the EMRs had occurred in the ED between
1st January, 2012 and 30th June, 2014 as the cases. However, there
were many EMRs but not all EMRs were indeed medical adverse event.
Thus, downsizing the number of EMRs and ensuring that the selected
EMRs were adverse events related to human errors were necessary.
Three steps were performed for the purposes. Firstly, a medical doctor
(MD) who has worked in ED for more than 20 years and an expert from
human factors and ergonomics area were recruited to identify the initial
EMRs. To effectively identify EMRs associated with medical adverse
events, we proposed six scenarios, in which the medical adverse events
occurred frequently, based on the existing literature. These scenarios
were used as indexes for the initial identification of EMRs. The six
scenarios were derived from the previous studies, including being
transferred to the intensive care unit or operation room, cardio-
pulmonary arrest occurred during the observation period in ED, death
during diagnosis period in ED, the diagnoses of hospitalization were not
consistent with discharge, returned to the ED in 72 h, and returned and
hospitalized within 7 days (Wolff and Bourke, 2002; Hendrie et al.,
2007; Calder et al., 2010). Secondly, the diagnostic content of EMRs
related to Taiwan Triage and Acuity Scale (TTAS) level 1 and level 2,
which meant that the patient was in an urgent situation, were selected;
medical errors could result in death of patients. Thirdly, from the pool
of EMRs, error factors that were unrelated to human errors were ex-
cluded, such as false labor, vaginal spontaneous delivery, and out-of-
hospital cardiac arrest. Through the steps, this study could ensure that
the selected EMRs contained at least one serious medical adverse event
associated with human errors in ED.
M.-c. Hsieh et al. International Journal of Industrial Ergonomics 67 (2018) 171–179

Table 1 evaluation procedure is necessary to assure consistency. Both the con-

The exact value for each linguistic variable (Chen and Hwang, 1992). sistency of decision-makers and overall hierarchy can be evaluated by
Linguistic variables exact value Linguistic variables exact value the measurement of consistency (Dağdeviren et al., 2009).

Extremely high 0.954 More or less low 0.410 2.3. The fuzzy TOPSIS method
Very high 0.864 Low 0.333
High-very high 0.701 Low-very low 0.299
High 0.667 Very low 0.136 With regard to fuzzy set theory that introduced in section 1.3, a TFN
More or less high 0.590 None 0.064 can be represented as a triplet (l, m, u). The membership functions of a
Medium 0.500 fuzzy number μ A∼ (x ) are defined as shown (Chiu and Hsieh, 2016):

● l to m is increasing function
The final pool of ED adverse events was generated following the ● m to u is decreasing function
step-by-step selection process. The chosen ED adverse events were ● l≤m≤u
analyzed with the help of three medical doctors, two nurses, and one
human factors and ergonomics expert. The medical doctors and the ⎧0 for x < l; x > u
nurses have more than fiften-year work experience in ED in Taiwan. ⎪ x−l
μ A∼ (x ) = m − l for l ≤ x ≤ m
The human factors and ergonomics experts has more than five-year ⎨ u−x
⎪ u − m for m ≤ x ≤ u
work experience in National Tsing Hua University. The final pool of ED ⎩ (4)
adverse events was analyzed using HFACS. During the period when the
The linguistic variables, represented by a TFN, are words or sen-
analysis was ongoing, experts could request to review and check the
tences of a natural and/or artificial language. Based on TFN, Chen and
detailed diagnosis contents of EMR for each ED adverse event to make
Hwang (1992) developed a set of eight scales, each contains different
their decision about which category that a specific error belonged to.
numbers of variables. To strengthen the power of explanatory, the
The numbers of errors reported by the seven experts mentioned above
current study employed scale eight (Table 1), which has 11 linguistic
were recorded. To further calculate the probabilities of occurrence, the
variables, to assess the levels in each TOPSIS criterion.
numbers of errors in each error type were averaged respectively.
Developed by Hwang and Yoon (1981), TOPSIS is one of the clas-
sical methodologies for dealing with MCDM problems. The main ob-
2.2. The AHP method jective of TOPSIS is to generate a ranking of potential alternatives
(Kannan et al., 2014). It is achieved through, firstly, identifying the
In order to determine the weights of the criteria for TOPSIS, AHP positive ideal solution (PIS) and the negative ideal solution (NIS). Then,
was used in this study. Dağdeviren et al. (2009) indicated that AHP was the distance between each potential alternative and the ideal solution is
based on three main steps: 1) structure of the model, 2) comparative measured. The PIS maximizes the benefit criteria and minimizes the
judgment of the criteria, and 3) synthesis of the priorities. At the first cost criteria, vice versa for the NIS (Dağdeviren et al., 2009). The se-
step, a complex decision problem needs to be structured hierarchically. lected alternative is the one with the smallest distance to the PIS and
The second step is to judge the importance of each criterion by pairwise the largest distance to the NIS. The TOPSIS consists of the following
comparison based on a standardized comparison scale with nine levels steps (Kannan et al., 2014).
(1 represents equally important, and 9 represents extremely more im- Step 1: Constructing the decision matrix.
portant). Let C = {Cj |j = 1, 2, …, n} be the set of criteria. The pairwise A group with k decision-makers (D1, D2,…, Dk), m alternatives (A1,
comparison on n criteria can be summarized in a (n × n) evaluation A2,…, Am) and n criteria (C1, C2,…, Cn) for a MCDM problem which is
matrix, A. Every element aij (ij = 1, 2, …, n) of the matrix A is the clearly expressed in a matrix format as:
quotient of weights of the criteria, which are as follows:
C1 C2 … Cn
a a ⋯ a1n A1
⎡ 11 12 ⎤ A r11 r12 … r1n
⎢ a21 a22 ⋯ a2n ⎥ D = 2⎡ r r22 ⋯
⎤
r2n ⎥
A=
⎢ ⋮ ⋮ ⋱ ⋮ ⎥ ⋮ ⎢ 21
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢ an1 an2 ⋯ anm ⎥ AM ⎢ r rm2 ⋯ rmn ⎥
⎣ ⎦ ⎣ m1 ⎦
where aii = 1, aji = 1 aij , aij ≠ 0 . where rmn be the rating of alternative Am with respect to criterion Cn.
At the final step, normalization and identification of the relative Step 2: Aggregate the evaluation of decision-makers.
weights were performed for individual criterion through the mathe- At this step, the aggregate of decision-makers was evaluated to
matical process. The relative weights are given by the right eigenvector determining the criteria weights. Let Wj be the weight of the criterion Cj
(W) corresponding to the largest eigenvalue (λmax ), which are as fol- that evaluated from AHP. The detailed calculation steps of AHP were
lows: descripted in section 2.2.
Aw = λmax W (1) Step 3: Construct the normalized decision matrix.
Assume that the decision matrix be X = [x ij]m × n . The decision matrix
Importantly, the quality of the AHP results and the consistency of for m alternatives and n criteria can be normalized as:
the pairwise comparison judgements are strictly related with each
S = [sij]m × n
other. The consistency is defined by the relation between the entries of
A: aij × ajk = aik . The consistency index (CI) is as follows: where
CI = (λmax − n)/(n − 1) (2) rij
sij = m
∑i = 1 rij2 (5)
The measurement of the final consistency ratio (CR) is to allow re-
searchers to conclude whether the evaluations are sufficiently con- Step 4: Construct the weighted normalized decision matrix.
sistent, which is calculated as the ratio of the CI and the random index The weighted normalized fuzzy decision matrix was constructed to
(RI). evaluate the level of importance of each criterion. Let the weighted
CR = CI / RI (3) normalized decision matrix be V = (vij )m × n .

If the final CR exceeds the accepted upper limit, 0.1, repetition of the vij = sij × Wj (6)

174
M.-c. Hsieh et al. International Journal of Industrial Ergonomics 67 (2018) 171–179

where i = 1, 2, …, m; j = 1, 2, …, n; Wj represents the weight of the jth Table 2

criterion. The average number and frequency of errors in ED (n = 35).
Step 5: Determine the PIS and NIS. Error factors of HFACS Average number of Frequency of the error
The PIS and the NIS of this study can be defined as: error factor

A+ = {(maxvij |j ∈ J ), (minvij |j ∈ J ′)|i = 1,2…,m} = {v1∗, v2∗, …, vn∗} (7) Unsafe acts
Decision errors 24.4 69.8%
A− = {(minvij |j ∈ J ), (maxvij |j ∈ J ′)|i = 1,2…,m} = {v1−, v2−, …, vn−} (8) Skill-based errors 20.7 59.2%
Perception errors 10 28.6%
Step 6: The distance of each alternative from A+ and A− is calcu- Violations 16.3 46.5%
lated respectively Preconditions for unsafe acts
Physical environment 9 25.7%
n Technological environment 11.1 31.8%
d+ = ∑ (vij − v∗j )2 , i = 1,2…m ; j = 1,2, …, n Adverse mental states 23.3 66.5%
j=1 (9) Adverse psychological states 4.1 11.8%
Physical/Mental limitations 10.4 29.8%
n Crew resource management 26 74.3%
d− = ∑ (vij − v−j )2 , i = 1,2…m ; j = 1,2, …, n Personal readiness 7 19.6%
j=1 (10) Unsafe supervision
Inadequate supervision 24 68.6%
Step 7: Calculate the closeness coefficient. Planned inappropriate 17.6 50.2%
With the formula below, the closeness coefficient of each alternative operations
Failed to correct a problem 18.4 52.7%
was calculated.
Supervisory violations 4.6 13.1%
d− Organizational influence
CCi = Resource management 18.1 54.7%
d+ + d− (11)
Organizational climate 21.1 56.7%
Step 8: Rank the order of alternatives according to the closeness Organizational process 15 43.7%
coefficient.
Based on the closeness coefficient of each alternative, all alter-
natives are ranked from the highest to the lowest closeness coefficient. 2.5. Application process of fuzzy TOPSIS
The importance of human error factors in the MCDM problems are also
represented by the order of the ranking. A five-step process was applied to select the important factors in
human errors in the four categories of HFACS. The process is described
below:
2.4. Identification of criteria and the hierarchy for human error factors
Step 1: Identify the criteria.
selection
With the assistance of the experts and by reviewing existing litera-
ture, this study identified the criteria that could be employed to de-
Criteria to be considered when selecting the important error factors
termine the factors of importance in human error selection. Three cri-
were determined by the expert team. Past experience and their back-
teria were involved in the current study and the definition for each is
ground were applied to the criteria determination, and three criteria
shown in section 2.4.
related to the feature of importance were established. The relevant
Step 2: Evaluate the error factors.
definition of each criterion is shown as below.
The seven experts need to evaluate the error factors based on the
three criteria in each ED adverse events using the linguistic variables in
➢ Influence: The severity of the errors in the ED adverse event.
Table 2.
➢ Reproducibility: The frequency of the errors occurs in ED.
Step 3: Normalize and weight the Fuzzy decision matrixes.
➢ Preventability: The preventability of the errors in ED.
The fuzzy decision matrix was normalized using formula (5). The
weights used in this step were determined by AHP, it was multiplied
Based on the categories of HFACS for important human error factors
with the normalized matrix to form a weighted normalized fuzzy de-
selection problem, four decision hierarchy structures were obtained.
cision matrix.
Each decision hierarchy has three levels. The goal of the decision pro-
Step 4: Determine the PIS and NIS.
cess, i.e. the selection of the important human error factors, is the first level
After generating the matrix, the PIS and the NIS were determined
of each hierarchy. The second level consists of three criteria of this
using formula (7) and formula (8) respectively.
decision problem. This level is identical across the four hierarchies. The
Step 5: Calculation and ranking.
third level is the alternatives, which represent the error factors of
The distance of the criteria from the PIS and the NIS was calculated
HFACS. One of the four hierarchies, the decision hierarchy structure of
using the formula in (9) and (10). In order to select the important active
unsafe acts, is used as an example to illustrate how the hierarchy is
and latent human error factors based on their closeness to the PIS and
presented in this study (Fig. 2).

Fig. 2. The hierarchy structure for unsafe acts.

175
M.-c. Hsieh et al. International Journal of Industrial Ergonomics 67 (2018) 171–179

remoteness to the NIS, the closeness coefficient was calculated using Table 4
formula (11). Based on the results of the closeness coefficient, the ranks The results of the weights from AHP.
of the most important error factors were provided. Criteria Weights λmax , CI, RI CR

3. Results Influence 0.322 λmax = 3.06 0.055 < 0.1

Reproducibility 0.202 CI = 0.03
Preventability 0.476 RI = 0.58
3.1. ED adverse event analysis

This study analyzed the EMRs which occurred in the ED between Table 5
January 1st, 2012 and June 30th, 2014. There were 13,759 EMRs se- The decision matrixes of “unsafe acts”.
lected based on the six scenarios at the first selection step. Out of the
Unsafe acts Influence Reproducibility Preventability
13,759 EMRs, 840 EMRs were identified with triage level 1 and 2 at the
second selection step. From the 840 EMRs, 35 were identified as ED Decision errors 0.549 0.498 0.493
adverse events that serves the purpose of the study. None of these 35 Skill-based errors 0.471 0.431 0.439
events were related to false labor, vaginal spontaneous delivery, or out- Perception errors 0.259 0.229 0.232
of- hospital cardiac arrest. After human error analysis, the number of Violations 0.385 0.355 0.370

errors identified by the seven experts were recorded separately. To

calculate the probabilities of error occurrence, the numbers of errors
Table 6
within the same category were averaged. Table 2 lists the frequency and
The weighted normalized decision matrixes with the PIS and NIS of “unsafe
the average number of the error factors of HFACS. The decision errors, acts”.
crew resource management, inadequate supervision, and organizational
climate were the most frequently occurring error factors in ED. Unsafe acts Influence Reproducibility Preventability

Decision errors 0.206 0.128 0.296

3.2. Results of AHP Skill-based errors 0.176 0.111 0.264
Perception errors 0.097 0.059 0.139
After forming the four decision hierarchies, the weights of the cri- Violations 0.145 0.092 0.223
PIS 0.206 0.128 0.296
teria used in the evaluation process were calculated by AHP method. In
NIS 0.097 0.059 0.139
this phase, the experts in the team were given the task of forming in-
dividual pairwise comparison matrix using the given scale. The values
from the individual pairwise comparison matrix were extracted to Table 7
calculate the group pairwise comparison matrix, shown in Table 3, The distance, closeness coefficient, and ranks of “unsafe acts”.
using geometric means. After standardizing the pairwise comparison
Unsafe acts D+ D- CCi Rank
matrix and calculating the average number of each row vector, the
results of the weights for each criterion were obtained, demonstrated in Decision errors 0.00 0.20 1.00 1
Table 4. Consistency ratio of the pairwise comparison matrix was cal- Skill-based errors 0.05 0.16 0.77 2
culated using formula (3) as 0.055 < 0.1. Thus, the weights were Perception errors 0.20 0.00 0.00 4
Violations 0.10 0.10 0.50 3
consistent and could be applied to the process of TOPSIS.

3.3. Results of fuzzy TOPSIS ➢ The most of importance error factors in “organizational influence”:
resource management
The error category of “unsafe acts” was used as an example to il-
lustrate how the fuzzy TOPSIS worked in this study.
The decision matrixes of “unsafe acts” was shown in Table 5. The 4. Sensitivity analysis
exact values in Table 5 were the averages from the seven experts.
Following the application process, the weighted normalized decision In order to test whether the results of the important human error
matrix, the distance of each criterion to PIS and NIS, closeness coeffi- factors were robust, a sensitivity analysis was conducted to analyze the
cient, and ranks obtained by each criterion were calculated respectively important active and latent human error factors under different criteria
in this study (Tables 6 and 7). weights (Önüt and Soner, 2008). This study presented the error cate-
Based on the results of the closeness coefficients, ranks of the most gory of “unsafe acts” as an example of sensitivity analysis in Table 8 and
important error factors in the four categories of HFACS are summarized Fig. 3. The intention of sensitivity analysis was to assess if the weights
as follows: of the criteria were exchangeable with one another. Thus, three com-
binations of the three criteria were analyzed. For each combination, the
➢ The most of important error factors in “unsafe acts”: decision errors
Table 8
➢ The most of important error factors in “preconditions for unsafe
Sensitivity analysis of “unsafe acts”.
acts”: crew resource management
➢ The most of importance error factors in “unsafe supervision”: in- Weight of criterion CCi of human error factors
adequate supervision
W1 W2 W3 H1 H2 H3 H4

Original 0.322 0.202 0.476 1.000 0.769 0.000 0.496

Table 3 1 0.202 0.322 0.476 1.000 0.773 0.000 0.502
The group pairwise comparison matrix for criteria. 2 0.476 0.202 0.322 1.000 0.749 0.000 0.466
3 0.322 0.476 0.202 1.000 0.751 0.000 0.468
Influence Reproducibility Preventability

Influence 1.00 2.07 0.52

(Note: W1 represents the criterion “influence”, W2 and W3 represent the criteria
Reproducibility 0.48 1.00 0.54 “reproducibility” and “preventability”. H1 represents the error factor “Decision
Preventability 1.91 1.86 1.00 errors”, H2, H3, H4 represent the error factor “skill-based errors”, “perception
errors”, and “violations”.).

176
M.-c. Hsieh et al.
Fig. 3. Sensitivity analysis of “unsafe acts”.
closeness coefficient of the human error factors was calculated. Table 8
summarizes the numerical results of the calculations and Fig. 3 presents
these results based on different categories of human error factors in
HFACS.
According to Table 8 and Fig. 3, “decision errors” had the highest
closeness coefficient value among all combinations. “Perception errors”
had the lowest closeness coefficient value of all combinations. Based on
the results of the sensitivity analysis of fuzzy TOPSIS, this study con-
cludes strongly that the most important human error factor, decision
errors, remains as the first to-be-focused in the category of “unsafe
acts”. It is due to its robust trend of the closeness coefficient value, as
shown in Fig. 3.
In the same manner, “crew resource management” had the highest
closeness coefficient value in all combinations. “Adverse psychological
states” had the lowest closeness coefficient value in all combinations.
Based on the results of the sensitivity analysis of fuzzy TOPSIS, this
study concludes strongly that the most important human error factor,
crew resource management, remains as the first to-be-focused in the
category of “preconditions for unsafe acts” since it has robust trend of
the closeness coefficient value.
Accordingly, “inadequate supervision” had the highest closeness
coefficient value in all combinations. “Supervisory violations” had the
lowest closeness coefficient value in all combinations. Based on the
results of the sensitivity analysis of fuzzy TOPSIS, this study concludes
strongly that the most important human error factor, inadequate su-
pervision, remains the first to-be-focused in the category of “unsafe
supervision” because of its robust trend of the closeness coefficient
value.
“Resource management” had the highest closeness coefficient value
in all combinations. “Organizational process” had the lowest closeness
coefficient value in all combinations. Based on the results of the sen-
sitivity analysis of fuzzy TOPSIS, this study concludes strongly that the
most important human error factor, inadequate supervision, remains
the first to-be-focused in the category of “organizational influence” due
to its the robust trend of the closeness coefficient value.
5. Discussion
After a series of analyzes, both important active and latent human
error factors were identified by this study. Decision error was the most
important active error factor in the category of “unsafe acts”. Crew
resource management, inadequate supervision, and resource
177
International Journal of Industrial Ergonomics 67 (2018) 171–179
management were the most important latent error factors in the cate-
gories of “preconditions for unsafe acts”, “unsafe supervision”, and
“organizational influence” respectively.
The results of the current study could be explained by the recent
states of the ED environment in Taiwan. Because the work environment
of ED is uncontrolled, unpredictable, and punctuated by intermittent
time-critical activities, the diagnosis and medical process might be
particularly susceptible to interrupts (Chisholm et al., 2000). Also,
failing to fulfill a hierarchical medical system, the ED of hospital are
often overcrowded with patients. Consequently, the medical doctors
usually make diagnoses without complete information and might not
have enough time to provide optimal care for each patient due to the
over-crowded ED environment (Trzeciak and Rivers, 2003; Epstein
et al., 2012). Speed-accuracy trade-off leading to misdiagnosis can ea-
sily occur when diagnoses are made in a short time. For examples, in-
appropriate risk evaluations and misinterpreting or missing out im-
portant information in diagnostic procedures can cause misjudgment of
the state of an illness and, therefore, affect patient safety.
In a busy healthcare department, cooperation between medical
staffs are even more important. In the second level of HFACS, crew
resource management is the most important problem that the ED cur-
rently encounters. Taking the hospital that cooperates with this study as
an example, the ED is run twenty-four seven with daily patient flow
about three-to four-hundreds. The medical staffs take turns to cover
three shifts and the number of staffs on each shift is adjusted based on
the patient flow. That is, night shift usually has less medical staffs than
the other shifts. Despite that the allocation of medical staffs is adjusted
for different shifts, patient flow fluctuates within a shift; yet, at the peak
(s) of patient flow(s), the number of staffs does not increase accord-
ingly. The only exception is when a mass-casualty incident occurs.
Consequently, the ED is overcrowded and the doctors need to make
quick diagnoses and schedule relevant check-up(s) or joint-consultation
(s) with doctors from another department. Overcrowding is one of the
factors that hugely increase the pressure of medical staffs. According to
the Inverted-U model, also known as Yerkes-Dodson Law (Yerkes and
Dodson, 1908), the right amount of stress facilitates work performance;
however, when an individual is over-stressed, the performance can
suffer. Therefore, overcrowding has negative effects on healthcare.
Other examples include communication between medical personnel
and communication between medical personnel and patients are also
part of personnel resource management.
The manager of a department plays a critical role in supervision.
M.-c. Hsieh et al.
The results revealed that inadequate supervision, involving the man-
ager of an ED failed to provide timely instructions and supervision, was
an important error factor of ‘unsafe supervision’ in the third level of
HFACS. Taking the hospital involved in the current research as an ex-
ample, the manager of the ED has teaching, research, and managerial
duties in addition to seeing patients. With such a busy schedule, it is not
easy to fully take on the supervisory role. In addition, due to the nature
of emergency service, doctors need to evaluate a patient's condition
within a short period of time, timely supervision from the manager
would be helpful to promote patient safety. For instance, the manager
may offer opinions regarding a patient's conditions or make use of an
incidence occurs on site as a chance to share experience.
Moreover, as for the forth level of HFACS, the majority of hospitals
in Taiwan are for-profit hospitals; therefore, the management ap-
proaches taken by the decision-makers are profit-oriented. The ED is
one of the most profitable department in a hospital. However, the profit
cannot be further increase because of its high operational and personnel
costs, as it is run twenty-four seven, as well as that it involves high risks
and can easily cause medical malpractices. As a result, when medical
resources are being distributed in a hospital, it is difficult for the ED to
request more. Also, the National Health Insurance (NHI) in Taiwan is
one of the factor leading to resource distribution problems within a
hospital. Comparing to many other countries, the system of NHI in
Taiwan is well-established. The NHI covers the majority of medical
expenses. When patients visit hospitals, they only pay part of the
medical expense to use the complete healthcare service. Hospitals, then,
report the treatments received by patients to NHI Administration to
claim the rest of the costs. However, each treatment reported is assessed
by the NHI Administration for its necessity. If a treatment is considered
inappropriate, the NHI Administration rejects the claim and, indirectly,
causes loss for the hospital. The loss of profit affects medical resource
distribution within a hospital.
The results of this study reflect the reality of the ED environment in
Taiwan. To reduce these phenomena, it is necessary to implement a
hierarchical medical system. The current situation in Taiwan is that
people prefer large hospitals, such as regional hospitals, teaching hos-
pitals, and medical centers, to local clinics when it comes to receiving
treatment. Therefore, the concept of a hierarchical medical system still
needs to be further introduced to the public. Recently, the Government
has increased the registration fee of the ED in large hospitals. In addi-
tion to adjusting the registration fee, a fine will be imposed to in-
dividuals who abuse the medical resources of the ED in order to ef-
fectively improve the overcrowding situation. Through implementing
these policies, the patient flow of some major hospitals can be reduced
so that the efficiency and quality of medical service are improved. On
the other hand, the NHI system needs to be re-evaluated to prevent
hospitals from reporting inflated costs as well as to fully reimburse the
medical expenses pre-paid by hospitals in order for the hospitals to run
smoothly.
Many studies have analyzed the error factors causing medical ad-
verse events. Freund et al. (2013) analyzed 34 medical adverse events
occurred in ED and recorded the frequency by error types. The most
frequently occurred error type is proficiency, followed by commu-
nication, procedural, and violation. Freund et al. adopted the analysis
used by Helmreich (2000), which involved five types of human error in
the human error analysis for ED and recorded the numbers of error
occurrences. The research aim of Freund et al. was to understand the
frequencies of different error types; therefore, unlike the current study,
no other criteria were used to evaluate error types. Also, when ex-
ecuting error analysis, managerial and organizational problems were
not analyzed. As a result, the human error types identified by Freund
et al. were not as insightful as those presented by the current study.
Diller et al. (2013) used HFACS to analyze 105 adverse events and re-
corded the numbers and percentages for each error type. Diller et al. not
only used the 19 error types of HFACS to analyze the adverse events but
also recorded deeper error states for the 19 error types such as safety
178
International Journal of Industrial Ergonomics 67 (2018) 171–179
checklist error for skill-based error in the category of “unsafe acts”. The
approaches taken by Diller et al. facilitated the study as well as furth-
ered human error analysis researches. But, the results only demon-
strated the frequency of errors and could not determine whether the
most frequently occurred error was indeed the most important error
factor.
In this study, important human error factors are identified through
comprehensive evaluations instead of applying a single criterion to
determine their importance. However, the results of the current study
are considered as more subjective than those of the previous studies,
although this approach is relatively time-consuming. After the experts
were recruited, a lot of time were spent on explaining the concept of
HFACS in order for the experts to fully understand the connotation of
each error category and type so that to minimize misjudgment. Due to
the fact that HFACS includes a large amount of error types, after re-
viewing a case, the experts spent, averagely, an hour on error analysis
for each case. Also, rather complex algorithms were involved when
applied AHP and TOPSIS to error factor evaluation. Yet, to familiarize
with the concept of HFACS and the procedures of AHP and TOPSIS
algorithm is helpful for the analyst to identify important human error
factors and, further, to enhance patient safety.
6. Conclusion
The purpose of this study is to investigate the most important active
and latent error factors in ED. AHP, fuzzy TOPSIS, and HFACS were
used in this study. The findings show that the most important error
factor in the category of unsafe acts is decision errors; in the category of
preconditions for unsafe acts is crew resource management; in the ca-
tegory of unsafe supervision is inadequate supervision; in the category
of organizational influence is resource management.
Three limitations exist in the current study. Firstly, the most im-
portant error factors found in this study depend on the ED adverse
events so the results might not be applicable to other situation, such as
adverse drug events. Secondly, this study only analyzed 35 adverse
events. To increase the number of ED adverse events is necessary in the
future work. Finally, this research was done in collaboration with only
one ED in a teaching hospital but not all ED in large hospitals in
Taiwan. All the cases involved in this study were provided by the
hospital and, therefore, might not well represent the situations in other
ED. Hence, it is necessary to not only increase the number of cases of
medical adverse event but also actively form collaboration with ED in
other hospitals in the future. It is expected that, using the methods of
analyses involved in this study, important human error factors can be
identified and used as guidance for developing policies to improve
patient safety in ED in Taiwan.
This study recommends that, after human error analyses are per-
formed, MCDM should be applied to further analyze the results based
on the criteria. In the future, this method can also be utilized to enhance
safety in other environments, such as aviation, nuclear power plant,
military, and other industries.
Relevance to industry
Safety is an important issue in hospitals and other fields. The per-
spective of this study can also be utilized to enhance safety in other
industries, such as aviation, nuclear power plant, military, and other
industries.
References
Behzadian, M., Otaghsara, S.K., Yazdani, M., Ignatius, J., 2012. A state-of the-art survey
of TOPSIS applications. Expert Syst. Appl. 39 (17), 13051–13069.
Bai, C., Dhavale, D., Sarkis, J., 2014. Integrating fuzzy C-means and TOPSIS for perfor-
mance evaluation: an application and comparative analysis. Expert Syst. Appl. 41 (9),
4186–4196.
Chen, S.J., Hwang, C.L., 1992. Fuzzy Multiple Attribute Decision-making Methods and
M.-c. Hsieh et al.
Applications. Springer-Verlag, New York.
Chisholm, C.D., Collison, E.K., Nelson, D.R., Cordell, W.H., 2000. Emergency department
workplace interruptions are emergency physicians “Interrupt-driven” and “multi-
tasking”? Acad. Emerg. Med. 7 (11), 1239–1243.
Chang, Y.H., Yeh, C.H., 2002. A survey analysis of service quality for domestic airlines.
Eur. J. Oper. Res. 139, 166–177.
Cosby, K.S., 2003. A framework for classifying factors that contribute to error in the
emergency department. Ann. Emerg. Med. 42 (6), 815–823.
Chang, Y.H., Chung, H.Y., Wang, S.Y., 2007. A survey and optimization-based evaluation
of development strategies for the air cargo industry. Int. J. Prod. Econ. 106, 550–562.
Calder, L.A., Forster, A., Nelson, M., Leclair, J., Perry, J., Vaillancourt, C., et al., 2010.
Adverse events among patients registered in high-acuity areas of the emergency
department: a prospective cohort study. CJEM 12 (05), 421–430.
Chauvin, C., Lardjane, S., Morel, G., Clostermann, J.P., Langard, B., 2013. Human and
organisational factors in maritime accidents: analysis of collisions at sea using the
HFACS. Accid. Anal. Prev. 59, 26–37.
Chiu, M.C., Hsieh, M.C., 2016. Latent human error analysis and efficient improvement
strategies by fuzzy TOPSIS in aviation maintenance tasks. Appl. Ergon. 54, 136–147.
Dağdeviren, M., Yavuz, S., Kılınç, N., 2009. Weapon selection using the AHP and TOPSIS
methods under fuzzy environment. Expert Syst. Appl. 36 (4), 8143–8151.
Diller, T., Helmrich, G., Dunning, S., Cox, S., Buchanan, A., Shappell, S., 2013. The human
factors analysis classification system (HFACS) applied to health care. Am. J. Med.
Qual. 29 (3), 181–190.
Daramola, A.Y., 2014. An investigation of air accidents in Nigeria using the human factors
analysis and classification system (HFACS) framework. J. Air Transp. Manag. 35,
39–54.
Epstein, S.K., Huckins, D.S., Liu, S.W., Pallin, D.J., Sullivan, A.F., Lipton, R.I., Camargo
Jr., C.A., 2012. Emergency department crowding and risk of preventable medical
errors. Intern. Emerg. Med. 7 (2), 173–180.
Fordyce, J., Blank, F.S., Pekow, P., Smithline, H.A., Ritter, G., Gehlbach, S., Benjamin, E.,
Henneman, P.L., 2003. Errors in a busy emergency department. Ann. Emerg. Med. 42
(3), 324–333.
Friedman, S.M., Provan, D., Moore, S., Hanneman, K., 2008. Errors, near misses and
adverse events in the emergency department: what can patients tell us? CJEM 10
(05), 421–427.
Freund, Y., Goulet, H., Bokobza, J., Ghanem, A., Carreira, S., Madec, D., et al., 2013.
Factors associated with adverse events resulting from medical errors in the emer-
gency department: two work better than one. J. Emerg. Med. 45 (2), 157–162.
Hwang, C.L., Yoon, K., 1981. Multiple Attribute Decision-making Methods and
Application. Springer-Verlag, New York.
Helmreich, R.L., 2000. On error management: lessons from aviation. BMJ 320 (7237),
781–785.
Hendrie, J., Sammartino, L., Silvapulle, M.J., Braitberg, G., 2007. Experience in adverse
events detection in an emergency department: nature of events. Emerg. Med.
Australas. 19 (1), 9–15.
James, J.T., 2013. A new, evidence-based estimate of patient harms associated with
hospital care. J. Patient Saf. 9 (3), 122–128.
Kohn, L.T., Corrigan, J.M., Donaldson, M.S., 2000. To Err Is Human: Building a Safer
Health System. National Academies Press, Washington, DC.
Kahraman, C., Cebeci, U., Ulukan, Z., 2003. Multi-criteria supplier selection using fuzzy
AHP. Logist. Inf. Manag. 16 (6), 382–394.
Kuo, R.J., Wu, Y.H., Hsu, T.S., 2012. Integration of fuzzy set theory and TOPSIS into
HFMEA to improve outpatient service for elderly patients in Taiwan. J. Chin. Med.
Assoc. 75 (7), 341–348.
Kannan, D., Jabbour, A.B.L., Jabbour, C.J., 2014. Selecting green suppliers based on
179
International Journal of Industrial Ergonomics 67 (2018) 171–179
GSCM practices: using fuzzy TOPSIS applied to a Brazilian electronics company. Eur.
J. Oper. Res. 233, 432–447.
Lisby, M., Nielsen, L.P., Mainz, J., 2005. Errors in the medication process: frequency,
type, and potential clinical consequences. Int. J. Qual. Health Care 17 (1), 15–22.
Li, W.C., Harris, D., 2006. Pilot error and its relationship with higher organizational le-
vels: HFACS analysis of 523 accidents. Aviat Space Environ. Med. 77 (10),
1056–1061.
Li, W.C., Harris, D., Yu, C.S., 2008. Routes to failure: analysis of 41 civil aviation acci-
dents from the Republic of China using the human factors analysis and classification
system. Accid. Anal. Prev. 40 (2), 426–434.
Liang, G.F., Lin, J.T., Hwang, S.L., Wang, E.M.Y., Patterson, P., 2010. Preventing human
errors in aviation maintenance using an on-line maintenance assistance platform. Int.
J. Ind. Ergon. 40 (3), 356–367.
Madigan, R., Golightly, D., Madders, R., 2016. Application of human factors analysis and
classification system (HFACS) to UK rail safety of the line incidents. Accid. Anal.
Prev. 97, 122–131.
Önüt, S., Soner, S., 2008. Transshipment site selection using the AHP and TOPSIS ap-
proaches under fuzzy environment. Waste Manag. 28 (9), 1552–1559.
Reason, J., 1990. Human Error. Cambridge University Press, Cambridge.
Reason, J., 1995. Understanding adverse events: human factors. Qual. Saf. Health Care 4
(2), 80–89.
Reason, J., 1997. Managing the Human and Organizational Response to Accidents.
Ashgate, Aldershot.
Rothschild, J.M., Churchill, W., Erickson, A., Munz, K., Schuur, J.D., Salzberg, C.A., et al.,
2010. Medication errors recovered by emergency department pharmacists. Ann.
Emerg. Med. 55 (6), 513–521.
Saaty, T.L., 1990. Decision Making for Leaders: the Analytic Hierarchy Process for
Decisions in a Complex World. RWS publications.
Stang, A.S., Wingert, A.S., Hartling, L., Plint, A.C., 2013. Adverse events related to
emergency department care: a systematic review. PLoS One 8 (9), e74214.
Trzeciak, S., Rivers, E.P., 2003. Emergency department overcrowding in the United
States: an emerging threat to patient safety and public health. Emerg. Med. J. 20 (5),
402–405.
Torlak, G., Sevkli, M., Sanal, M., Zaim, S., 2011. Analyzing business competition by using
fuzzy TOPSIS method: an example of Turkish domestic airline industry. Expert Syst.
Appl. 38 (4), 3396–3406.
Taiwan Patient-safety Reporting System, 2016. Annual Report 2015. Joint commission of
Taiwan.
Wiegmann, D.A., Shappell, S.A., 2001. Applying the human factors analysis and classi-
fication system to the analysis of commercial aviation accident data. In: Proceedings
of 11th International Symposium on Aviation Psychology. Ohio State University,
Columbus, OH.
Wolff, A.M., Bourke, J., 2002. Detecting and reducing adverse events in an Australian
rural base hospital emergency department using medical record screening and re-
view. Emerg. Med. J. 19 (1), 35–40.
Wiegmann, D.A., Shappell, S.A., 2003. A Human Error Approach to Aviation Accident
Analysis: the Human Factors Analysis and Classification System. Ashgate,
Aldershot, UK.
Wenner, C.A., Drury, C.G., 2000. Analyzing human error in aircraft ground damage in-
cidents. Int. J. Ind. Ergon. 26 (2), 177–199.
Wang, C.H., Chou, H.L., 2015. Assessment of patient safety management from human
factors perspective: a fuzzy TOPSIS approach. Hum. Factors Ergon. Manuf. Serv. Ind.
25 (5), 614–626.
Yerkes, R.M., Dodson, J.D., 1908. The relation of strength of stimulus to rapidity of habit-
formation. J. Comp. Neurol. 18 (5), 459–482.