You are on page 1of 9

Safety Science 140 (2021) 105303

Contents lists available at ScienceDirect

Safety Science
journal homepage: www.elsevier.com/locate/safety

Improving accident analysis in construction – Development of a


contributing factor classification framework and evaluation of its validity
and reliability
Anastacio Pinto Goncalves Filho a, *, Patrick Waterson b, Gyuchan Thomas Jun b
a
Ministry of Labour, Federal University of Bahia, Rua Aristides Novis, 2, Federação, Bahia, Brazil
b
Human Factors and Complex Systems Group, School of Design and Creative Arts, Loughborough University, Loughborough LE11 3TU, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: Construction industry is still one of the most dangerous industries and its fatal work injuries is almost three times
Construction higher than the average across all sectors. Previous researchers have attempted to apply more systemic models
Accimap and methods to improve accident analysis in construction, but few studies have fully encompassed upstream
Classification system
factors such as decisions and actions at the level of the government and regulator in accident analysis. More
Accident
Investigation
importantly, no previous study has evaluated the validity and reliability of systemic accident analysis methods in
construction. The present study, therefore, has two main aims: to develop a contributing factor classification
framework to support systemic accident investigation in construction and to carry out an assessment of its
validity and reliability. The classification framework was developed and assessed in two phases. The phase one
involved generating a list of contributing factors from the review of 26 articles and the analysis of 532 con­
struction accident reports. Five federal inspectors with expertise in accident investigation were involved in
refining the list into 61 contributing factors and categorizing them into six levels of the Accimap framework. The
phase two involved in assessing the validity and reliability of the framework with five practitioners in con­
struction sector using three real construction accidents. This study contributes to the development of a
contributing factors classification system framework for construction with acceptable validity and reliability.

1. Introduction vehicle) and geo-technical failure (Winge and Albrechtsen, 2018; Siraj
and Fayek, 2019; Woolley et al., 2018). In addition, the diversity of
There is international acknowledgment that the construction in­ construction sites, along with differences in occupations, technology,
dustry is still one of the most dangerous industries in which to work tools and materials used, make safety and health management very
(Melchior and Zanini, 2019; Umeokafor et al., 2019; Tsang et al., 2017). challenging (Haslam et al., 2005; Mitropoulos et al., 2005; Hallowell
According to the International Labour Organisation (ILO, 2017) and Gambatese, 2009). Construction also involves a high degree of
approximately one in six fatal accidents at work takes place in the subcontracting and outsourcing with multiple locations involving the
construction sector, accounting for 60,000 fatal accidents per year simultaneous interaction of remote and adverse geographical and
worldwide. Fatality rates are almost three times higher than the average environment conditions (Gibb et al., 2006; Hale et al., 2012; Harvey
across all sectors. For example, United Kingdom has a fatal injury rate of et al. 2018). In common with other sectors, construction is also influ­
1.31 per 100.000 construction workers compared with 0.45 across all enced by the external environment, such as political, economic, cultural
sectors (HSE, 2019), United States has a rate of 9.5 compared with 3.5 and legal factors which lead to accidents (Umeokafor et al., 2019).
(US Bureau of Labour Statistics, 2018) and Brazil has 13.3 compared
with 6.6 (AEAT, 2018). The construction sector, therefore, is consis­
1.1. Systemic accident analysis and construction
tently identified as priority for accident prevention actions and regula­
tion (Xia et al., 2018; Zhang et al, 2019). Construction accidents range
There has been growing interest in the development and application
from falls from heights to being struck by a moving object (e.g., a
of ‘systemic accident analysis’ (SAA) in the field of accident

* Corresponding author.
E-mail address: anastacifilho@ufba.br (A.P. Goncalves Filho).

https://doi.org/10.1016/j.ssci.2021.105303
Received 19 July 2020; Received in revised form 30 March 2021; Accepted 11 April 2021
Available online 23 April 2021
0925-7535/© 2021 Elsevier Ltd. All rights reserved.
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

investigation and analysis (Rasmussen, 1997; Underwood and Water­ models have investigated a more extensive range of system levels,
son, 2014; Leveson, 2004). SAA methods commonly illustrate the di­ methods with limited coverage remain predominant in construction
versity of causal factors across different levels of systems and, their (Harvey et al., 2018). This is despite the fact that it is widely acknowl­
interactions including the role played by external influences such as edged that there is a need for an ‘holistic’ system thinking approach
political, cultural, financial, and technological circumstances (Branford, within construction both in breadth and depth (Love et al., 2016).
2011). SAA methods are often seen as better suited to forming an un­ More importantly, another limitation of previous studies is that they
derstanding of accidents and to inform the design of effective risk have not been subject to a detailed evaluation of their validity and
management in complex sociotechnical systems. Traditional approaches reliability. The assessment of validity and reliability is widely viewed as
to accident investigation (e.g., Root Cause Analysis), on the other hand, a key attribute of accident models and methods (Salmon, 2016; Un­
tend to offer a simplistic explanation for accident causation by focusing derwood and Waterson, 2013). Stanton (2016) for example, argues that
on a single cause or physical event and human error alone (Canham, “all ergonomics methods need to prove that they can work in the intended
et al., 2018; Dekker et al., 2011). A variety of SAA methods have been domain of application” (p.347).
developed since the 1990s include, but are not limited to, Accimap
(Rasmussen, 1997), STAMP (Leveson, 2004) and FRAM (Hollnagel,
2004). These methods have been used across a wide variety of industries 1.3. Validity and reliability
including: aviation and aerospace industries (Johnson and de Almeida,
2008); led outdoor activities (Salmon et al., 2014); transport (Under­ The validity of a method can be assessed based on the results of the
wood and Waterson, 2014); patient safety (Waterson, 2009; Canham method application (Branford, 2007). This is referred to as empirical
et al., 2018); the process industries (Tabibzadeh and Meshkhati, 2015); validity or external validity, that is, the degree to which the results
and, marine transportation (Kee et al., 2016; Kim et al., 2016; Lee at al., obtained using the method are, in practice, those that it is intended to
2016). produce. The validity of an accident analysis method can be assessed in
Despite the widespread benefit of SAA models in research and multiple ways. Firstly, the validity can be assessed by evaluating results
practice in other safety critical domains, the use this approach in the against objective external criteria. For example, the validity of the re­
construction sector has thus far been very limited. For example, few sults can be assessed by checking their agreement or disagreement with
studies involving the of Accimap (Rasmussen, 1997), STAMP (Leveson, a so-called ’gold standard’, i.e., the results of a previously validated
2004) and FRAM (Hollnagel, 2004) have been conducted for incident method (Böiger and Wright, 1992). This validity evaluation, however, is
and accident investigation in the construction domain (Zhang et al., difficult to use because a ’gold standard’ is rarely available in accident
2019). Rosa et al. (2015) applied the FRAM to assess risk in the con­ investigation. Second, the validity can be assessed by the degree of
struction industry processes, whereas França et al. (2020) also applied similarity between results obtained from different methods (Carael,
the FRAM to understand the levels of complexity and show the relevant 2001). In this study, this approach is not applicable since there are not
human factors that are critical for the safe operations of these work­ multiple results drawn from one accident with similar format and
places. Accimaps have been used by Zhou et al. (2018) to qualitatively wording. Finally, validity can be assessed by the results against those of
build a generic model for a specific type of equipment safety, the tower the expert (Böiger and Wright, 1992; O’Connor, 2008). This study
crane, which is common on construction sites. Little is known about the adopted this approach and the aforementioned accident analysis results
systemic properties of construction as a whole, particularly the interplay by the five federal inspectors (experts) were compared with the results
between accident contributory factors, actors, controls and feedback by engineers without the experience of applying a systemic accident
loops. Specifically, the understanding of how decisions and actions at analysis method (see Section 2.2).
the level of the government and regulator play a role in accident The reliability of an accident analysis method can be assessed in a
causation is limited and the opportunity to respond to and reform number of ways (Carmines and Zeller, 1979; Cohen et al., 2015; Stanton
ineffective government policies and regulations is unlikely and critical and Young, 2003). Firstly, inter-rater reliability can be used when the
risk factors may be overlooked (Woolley et al., 2019). consistency of the results after applying the method by different analysts
can be examined. Secondly, intra-rater reliability refers to a comparison
1.2. Accident investigation methods and models applied in construction between the results drawn by the same analyst about the same data on
different occasions (Cohen et al., 2015; Ross et al., 2004). In general, the
Several attempts have been made to develop and apply accident more stable the results, the more confident one can be that the results
analysis models and methods for the construction sector as shown in are reproducible and trustworthy (Ergay et al., 2016). In this study, both
Table 1, however, there remains some limitations. One of these relates to inter-rater and intra-rater reliability were assessed.
their limited coverage - most of the models and methods, for example,
cover contributing factors only as far as the organisational levels of
analysis. The influence of government and regulatory body level factors 1.4. Study aims and organisation of the paper
on accident causation has rarely been covered by most accident analysis
models and methods in construction. The models developed by Suraji The present study has two main aims:
et al. (2001), Haslam et al. (2005), Mitropoulos et al. (2005) and Harvey
et al. (2018), for example, do not cover regulatory body level factors. Ye To develop a comprehensive contributing factor classification
et al. (2018) covered a variety of system levels including the government framework to support systemic accident investigation in the con­
and regulatory level, but the level of detail provided by the modified struction sector;
Human Factors Analysis and Classification System (HFACS) method is To carry out an assessment of the validity and reliability of the
limited and in need of further refinement and development. Woolley contributing factor classification framework in order to assess its
et al., (2018) also highlighted that SAA methods that identify contrib­ suitability for accident analysis in construction.
uting factors beyond what they term the ‘organisation ceiling’ have not
as yet been developed within construction. One of the most well-known This paper first describes the procedure we used to develop the
models of accident causation in construction is the Construction Causal contributing factor classification framework followed by its validity and
Accident (ConCa) model (Haslam et al., 2005; Harvey et al., 2018). reliability test (Section 2). The findings are the outlined in Section 3. A
ConCa draws on the Swiss Cheese Model developed by Reason (1990) final section (Section 4) discusses our findings and situates the contri­
and is also limited to the investigation of human error as far as organ­ bution of the paper within the wider context of construction accidents
isational level (Woolley et al., 2018). Whilst other sectors systems and accident investigation.

2
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

Table 1
Accident causation models and methods and their application to the construction sector.
Authors Method/model Government Regulatory Organisation Technical Worker Equipment & Coverage
body management activities surrounding

Abdelhamid and Accident Root ✓ ✓ The model is limited to identify


Everett (2000) Causes Tracing unsafe act and unsafe condition.
Model (ARCTM)
Suraji et al. Constraint- ✓ ✓ ✓ The causal model of construction
(2001) Response Model. accident causation considers both
proximal and distal contributing
factors such as site environment,
systems of work, project
management and organisational
issues.
Mitropoulos et al. Construction ✓ ✓ ✓ ✓ It focuses on how the characteristics
(2005) accident causation of the production system generate
model hazardous situations and shape the
work behaviours and analyses the
conditions that trigger the release of
the hazards.
Gibb et al. ConCa ✓ ✓ ✓ ✓ This accident causation model
(2006); Haslam examines contributing factors
et al. (2005); related to organisation influences,
HSE (2003) workers and job site.
Chua and Goh Modified Loss ✓ ✓ ✓ ✓ The model analyses contributing
(2004) Causation Model factors at levels organisational,
(MLCM) safety management, unsafe
conditions, unsafe acts and
personal.
Garrett and Human Error ✓ ✓ ✓ ✓ The method is a modified HFACS
Teizer (2009) Awareness Training framework and identifies
(HEAT) Contributing factors associated
with organisation, supervisory,
precondition for unsafe acts and
act/events.
Wu et al. (2010) Investigative ✓ ✓ The model was developed to
method of identified accident precursors and
Precursors and near miss on construction site from
Immediate Factors historical accident records. It
(PaIFs) examines contributing factors
related to work team, workplace
and equipment.
Hale et al. (2012) Combined ✓ ✓ ✓ ✓ ✓ The framework analyses
influence contributing factors at
framework. organisational level and external
environment (e.g., political,
regulatory, market and societal).
Mroszczyk Construction ✓ ✓ ✓ The model examines contributing
(2015) incident causation factor in job site (hazard,
model. management, equipment,
workplace, worker and unique to
construction).
Rodrigues et al. Causal factors ✓ ✓ ✓ The model identifies Contributing
(2015) model for factors associated with the
construction company, management and
accident frontline work levels.
Wang et al. Accident causal ✓ ✓ The model focus on worker’s safety
(2016) model. risk tolerance and those
contributing factor that lead to it,
such as personal subjective
perception; work knowledge and
experiences; work characteristics;
and safety management.
Harvey et al. ConCa+ ✓ ✓ ✓ ✓ ✓ This an updated ConCa model.
(2018) However, it is still focused on
organisational, workers and site
contributing factors.
Ye et al. (2018) I-HFCAS ✓ ✓ ✓ ✓ ✓ ✓ The framework is an improved
HFACS that was include External
Factors (Regulatory Factors and
Economic, political, social and legal
environment) at level 5. IHFAC
consisted of 18 factors leading to
error.
Xia et al. (2018) Bayesian-network ✓ ✓ ✓ ✓ The method is based on HFACS
(BN)-HFACS hybrid framework that focus on
model contributing factors at
(continued on next page)

3
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

Table 1 (continued )
Authors Method/model Government Regulatory Organisation Technical Worker Equipment & Coverage
body management activities surrounding

organisational level and lower


levels such as management, unsafe
supervision, preconditions for
unsafe acts and unsafe acts (human
error).
Zhang et al. Construction ✓ ✓ ✓ ✓ The model examines contributing
(2019) Accident Causation factors at organisational and
System (CACS) management levels (e.g., contract
management, safety training,
technical management, resource
support, emergency management)

2. Methods of study redundant factors. For example, restricted work area and limited work area
were combined into a single factor labelled limited work area. Then, the
This study involved two main phases of activities: (1) carrying out a resulting list was checked over by other authors. They discussed and any
literature review in order to develop the classification framework; and controversies were solved by consensus. Following this, an initial list of
(2) an assessment of the validity and reliability of the framework. 445 contributing factors was identified.

2.1. Phase 1 – Developing a framework for classifying contributing factors 2.1.2. Categorisation of contributing factors
Accimaps (Svedung and Rasmussen, 2002) were selected as a cate­
Fig. 1 shows the development process of the final contribution factor gorisation framework for three reasons. Firstly, it is a system thinking-
classification framework, which are described in the next section. based approach that enables the inclusion of causal factors beyond the
organisation level and up to legislation, regulations, certification,
2.1.1. Literature review and accident report analysis auditing, and governmental budgeting. Secondly, based upon our
The first phase in developing the classification framework involved experience the Accimap is considered easy to be used and less time
generating a list of contributing factors from literature review and ac­ consuming than other SAA methods. Finally, Accimap has been widely
cident reports. In order to achieve this, a literature review was con­ applied across a range of safety–critical domains (Hulme et al., 2019;
ducted to identify contributing factors to accidents in construction. The Waterson et al., 2016).
following databases were searched: Science Direct, Web of Science, Five federal inspectors with more than ten-year experience with
Scopus, PubMed, PsycINFO, Construction and Building Abstracts (CBA), accident investigations in the construction domain were recruited to
International Civil Engineering Abstracts, Wiley Online Library; Psy­ refine the initial list of 445 contributing factors. The inspectors were
cARTICLES and Google Scholar. The combinations of the search terms asked to consolidate the list on an individual basis and met together to
used were ((incident OR accident) AND causation models AND con­ address any disagreements by consensus. This method of the refinement
struction); (causal factors AND construction); ((incident OR accident) was based on the Delphi method adapted for use in face-to-face meetings
AND investigation AND construction); ((incident OR accident) AND (see more details in Okoli and Pawlowski, 2004). Three meetings were
analysis methods AND construction); ((incident OR accident) AND necessary for consensus to be reached.
analysis model AND construction); (contributory contributing factors The inspectors were subsequently asked to categorise the refined
AND construction). The references which were cited in articles were also contributing factors into one of the six levels of the Accimap framework:
used as an additional source for our searches. Articles initially identified 1. government; 2. regulatory body (external); 3. organisations; 4. tech­
were subsequently filtered by checking whether the article identify nical & operational management; 5. physical processes & actor activ­
contributing factors or present causation models. Finally, our search was ities; and, 6. equipment & surroundings and physical environment.
limited to publications written in the English language and published in Again, the inspectors met together and addressed any disagreements by
peer-reviewed academic journal. In total, 26 articles were included in eventually achieving consensus. The first author facilitated the
final review (see Table 2). consensus building process but had no influence on the process of
Construction-related accident reports which were published between developing the resultant framework.
2009 and 2014 in Brazil were also analysed, in order to identify
contributing factors beyond those that are in literature. In total, 532 2.2. Phase 2 – Assessing the validity and reliability of the classification
reports were obtained from Sistema Federal de Inspeção database, framework
which is under the authority of the Ministry of Labour of Brazil. Each
accident report described the contributing factors that led to accident. 2.2.1. Pilot test of the contributing factor classification framework
The first author, as a subject matter expert and federal inspector on The contributing factor classification framework was pilot-tested and
accident investigation in construction, reviewed both the articles and covered the following characteristics of the use methods for accident
the accident reports and extracted contributing factors from them. The investigation and analysis (Underwood and Waterson, 2013): useful­
list was then simplified by eliminating duplications and combining ness, reliability, suitability for real-world environment, and

Literature review
(26 articles) Initial list of 445 Refined and Contributing Factors
contributing categorised Classification
factors by five experts Framework
Analysis of 532
accident reports

Fig. 1. Development process of contributing factor classification framework.

4
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

Table 2 day lecture-based training session on SAA methods and the classification
Literature Review – Sample. framework and another half-day discussion-based training session on
Author Title detailed descriptions of each level of the framework and case studies.
The participants were provided with the final version of the contributing
1. Sawacha et al. (1999) Factors affecting safety performance on construction
sites factor classification framework and definitions of all the terms used in
2. Abdelhamid and Identifying root causes of construction accidents the framework alongside worked examples that were discussed during
Everett (2000) the training.
3. Suraji et al. (2001) Development of causal model of construction accident
causation
4. HSE (2003) Causal factors in construction accidents
2.2.3. Method for validity and reliability assessment and criterion for
5. Teo et al. (2005) Framework for project managers to manage construction acceptance
safety The Index of Concordance was used to assess validity and reliability.
6. Haslam et al. (2005) Causal factors in construction accidents This has been used to calculate agreement by dividing the total number
7. Gibb et al. (2006) What causes accidents?
of agreements between pairs with the total number of agreements
8. Aksorn and Critical success factors influencing safety program
Hadikusumo (2008) performance in Thai construction projects possible between all pairs. The formula for the index of concordance is
9. Choudhry and Fang Why operatives engage in unsafe work behavior: IOC = A/(A + D), where A is the total number of agreements and D is the
(2008) Investigating factors on construction sites total number of disagreements (see more details in Ross et al., 2004;
10. Garrett and Teizer Human factors analysis classification system relating to Wallace et al., 2002). There are no generally agreed criteria for what
(2009) human error awareness taxonomy in construction safety
11. Hale et al. (2012) Developing the understanding of underlying causes of
level of Index of Concordance is acceptable (Olsen, 2013; Wallace et al.,
construction fatal accidents 2002), but this study has decided to use a threshold proposed by Wallace
12. Oswald et al. (2013) Exploring factors affecting unsafe behaviours in and Ross (2006), that is, 70% as an acceptable validity and reliability
construction (Goode et al., 2017; Olsen, 2011; Olsen and Shorrock, 2010).
13. Khosravi et al. Factors influencing unsafe behaviors and accidents on
(2014) construction sites: A review
14. Mroszczyk (2015) Improving construction safety A team effort 2.2.4. Applying the contributing factors classification framework
15. Rodrigues et al. Correlation of causal factors that influence the Three recent construction accident scenarios with fatal victims ob­
(2015) constructionsafety performance: A model tained from Accident Workplace Communication (AWC) database in
16. Wang et al. (2016) Critical factors and paths influencing construction Brazil were used in this study. They ranged from a lift crash, an exca­
workers’ safetyrisk tolerances
vation collapse to an excavation flood. Fatal accidents were chosen
17. Carrillo-Castrillo Construction accidents: identification of the main
et al. (2017) associations between causes, mechanisms and stages of because they tend to have a greater number of contributing factors than
the construction process minor accidents (e.g., slips, trips or falls). We expected that fatal acci­
18. Tsang et al. (2017) Development of an accident modelling in the Hong Kong dents would provide an opportunity for the participants to use greater
construction industry
range of possible contributing factors present in the framework.
19. Harvey et al. (2018) Beyond ConCA: Rethinking causality and construction
accidents The participants were given a brief description of the three accidents,
20. Ye et al. (2018) Improved HFACS on human factors of construction as shown in Table 3, and asked to analyse them using the contributing
accidents: A China perspective factor classification framework. This brief description was not enough to
21. Woolley et al. Moving beyond the organizational ceiling: Do analyse the three accidents; therefore, the participants should investi­
(2018) construction accident investigations align with systems
gate the accident where they occurred and collect all data to apply the
thinking?
22. Xia et al. (2018) A Hybrid BN-HFACS model for predicting safety contributing factor classification framework. All the five participants
performance in construction projects stated that they have not prior knowledge of these three accidents.
23. Siraj and Fayek Risk identification and common risks in construction: The participants were asked to work independently and not to
(2019) literature review and content analysis
discuss the process or analysis with the other participants. It was
24. Umeokafor et al. Causal inferences of external–contextual domains on
(2019) complex construction, safety, health and environment
explained that this restriction was necessary to ensure that the results of
regulation the reliability assessment were not compromised. There was no time
25. Woolley et al. Have we reached the organisational ceiling? A review of limit for completing each analysis as long as all three accident analyses
(2019) applied accident causation models, methods and were completed (Time 1) and they should return later (Time 2) to
contributing factors in construction
analyse the same three accident again. The time gap between Time 1 and
26. Zhang et al. (2019) Identification of critical causes of construction accidents
in China using a model based on system thinking and 2 ranged from one to two months. Furthermore, there was no attempt to
case analysis influence the level of detail provided or the extent to which factors at
different levels in the Accimap framework were represented in the ac­
cidents analysed. The ‘experts’ were also asked to carry out the same
comprehensibility in terms of investigation data collection and analysis. three accident analysis. Fig. 2 shows schematically how the validity and
Each of the five inspectors was asked to apply the classification frame­ reliability analyses were carried out.
work to the analysis of three real past construction accidents. After the
individual application, the five inspectors met together, compared notes,
Table 3
and modified the framework. The third meeting was arranged in order to
Accident brief description.
discuss and agree upon the final contributing factor classification
framework. The final results of the three accident analyses by the five Accident Year Consequences Brief description

federal inspectors (experts) were used for the validity assessment. Lift crash 2017 Ten deaths A cabin of a passenger elevator
collapsed and went down from 100-
meters height. The ten passengers who
2.2.2. Training of participants and materials
were in the cabin died.
Five new participants were recruited who were engineers with 5–10 Excavation 2017 One death Two workers were carrying out an
years’ experience in the construction industry. None of the participants collapse excavation in a construction site when
had previous experience in the use of SAA methods. The first author, the excavation collapsed. In
who had been an active SAA method user over the last five years, pro­ consequence, one worker dead.
Excavation 2018 One death An excavation was being carried out by
vided two half-day training sessions covering SAA methods to the par­ flood workers in a construction site flooded
ticipants and introduced the classification framework (eight hours of quickly as pipe with high pressure
training in total). The two half day training sessions consisted of a half- water broke. A worker died.

5
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

Fig. 2. Validity and reliability analysis.

3. Findings additional factors such as “poor infrastructure for inspection” and


“inadequate recruitment policy of inspectors” at government level and
3.1. Phase 1 – Contributing factor classification framework “inappropriate or lack inspection at workplace” and “lack or insufficient
number of inspectors” at the regulatory body level.
Fig. 3 shows 61 contributing factors distributed across the six levels. The “Other (type of contributing factors here)” option was added at
The experts refined the initial comprehensive list, but they also added the top four levels (government, regulatory body, organisational and

Level 1 – Government
Poor or lack Poor infrastructure to inspection Economics policy Inadequate inspection policy
communication of Inadequate or Lack
legislations of legislation
Reduced budgets Inadequate recruitment policy of inspectors Other (type of contributing factors here)

Level 2 – Regulator body


Poor or lack Inappropriate or lack Inadequate Lack or insufficient Priority definition Other (type of
communication of inspection at enforcement of number of inspectors inadequate contributing factors
legislations workplace legislation here)

Level 3 – Organisational
Inappropriate or lack Financial constraints Resource allocation Judgement and Incompatible goals Acceptance or encouragement of
of flow of problems decision-making error between safety and short cuts, non-compliance rules
communication production and legislations production

Inadequate Poor or lack of risk


Inadequate manuals construction process assessment and Poor or lack of Insufficient number of supervisors, Lack or inadequate
and procedures management recruitment policy managers and coordinators safety policy

Poor accident Production priority Inadequate or Lack Inadequate or lack of inspection Inadequate or lack of Other (type of
learning systems over safety of training policy and maintenance policy outsourcing policy contributing factors here)

Level 4 – Technical and


operational management Failure to follow Failure to Lack of planning Provide poor instructions Failure to provide Inappropriate or lack
policy adequately plan task task appropriate of flow of
equipment or tool communication
Lack experience, qualifications Failure to provide Personal Lack or inadequate training to
and competence Protective Equipment front-line workers

Acquiescence with procedures Overtime


Other (type of excess
Poor implementation of safety system Inadequate work schedule violations and unsafe acts
contributing factors
here)

Level 5 – Physical processes


and worker activities Mental and physical Lack experience, qualifications and Violations of procedure Human errors, lapses or Worker Inadequate or
condition adverse competence and rules slips attitude lack of procedure

Level 6 – Equipment, surroundings


and meteorological conditions Inadequate or lack Inadequate, insufficient Inadequate, insufficient Inadequate Fault equipment Inadequate Lack of
of Personal or missing proactive or missing reactive equipment or tools or tools materials materials
Protective system defenses system defenses
Equipment
Lack or
Poor site Poor site Housekeeping and Poor working inappropriate ways
Adverse weather conditions layout/space site organisation environment permanent access
conditions
Equipment

Fig. 3. Contributing Factors Classification System Framework (CFCSF).

6
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

technical and operational management) so that the end users of this Table 5
framework can include other contributing factors beyond predefined Inter-rater reliability: Index of Concordance (%) among participants at time 1
factors if needed. and 2.
Accident Mean overall
3.2. Phase 2 – Validity and reliability assessment Time 1 Standard Deviation Time 2 Standard Deviation

1 71.3 2.3 70.4 2.7


3.2.1. Validity 2 74.3 4.1 71.5 2.1
Descriptive statistics of the validity assessment are shown in Table 4. 3 73.4 2.3 73.0 3.6
The results show that the mean overall of the Index of Concordance was Mean 73.0 71.6
consistently above the acceptable threshold for time 1 (first accident
analysis) and time 2 (second accident analysis one month later) for the
three accidents. There is a reasonably strong agreement on contributing Table 6
factors at Equipment, surroundings and meteorological condition level Intra-rater reliability: Index Concordance (%).
(e.g., “adverse weather conditions”, “poor working environment” and Participant Accident 1 Accident 2 Accident 3
“inadequate materials”). However, the results indicate that also exist
1 100 93.0 84.5
some variation in agreement amongst participants with the expert. For 2 78.2 89.3 89.3
example, the participants tend not to identify the contributing factors at 3 76.7 76.7 74.5
the government level (e.g. poor infrastructure to inspection and inadequate 4 79.7 75.3 78.7
inspection policy), and at the regulatory body level (e.g., inadequate 5 65.7 47.8 100
Mean overall 80.1 76.4 85.4
enforcement of legislation). Although variations in agreement amongst
participants with the expert existed, the standard deviations indicate
that the Index of Concordance values did not vary substantially between participants strongly agreed that the framework is easy to understand,
the levels of the Accimap framework. easy to apply and clearly presented.

3.2.2. Inter- and intra-rater reliability 4. Discussion


Table 5 show that the mean overall Index of Concordance for inter-
reliability was consistently above the acceptable threshold for time 1 4.1. Summary of findings
and 2 in all three accidents. The results indicate that there were some
disagreement amongst the participants at the regulatory body level (e.g., The aim of this study was to develop a comprehensive contributing
lack or insufficient number of inspectors and inadequate definition of pri­ factor classification framework to support systemic accident analysis in
ority) and the technical and operational management level (e.g., acqui­ construction and provide the evidence of its validity and reliability. The
escence with procedures violations and unsafe acts and lack or inadequate framework was developed from the review of 26 academic articles and
training to front-line workers). However, the Index of Concordance values 532 construction accident reports and refined by five experts. It consists
did not vary substantially between the levels of the Accimap framework, of a total of 61 contributing factors populated throughout the six levels:
as demonstrated by a comparison using standard deviations. government (7 contributing factors); regulatory bodies (5 contributing
A summary of the intra-rater reliability for each participant is shown factors); organisational (17 contributing factors); technical and opera­
in Table 6. The results indicate that the mean overall Index of Concor­ tional management (13 contributing factors); physical processes and
dance between the two investigations was significantly above the worker activities (6 contributing factors); and equipment, surroundings
acceptable threshold for the three accidents. All participants, aside from and meteorological conditions (13 contributing factors). The participant
participant 5, have the same agreement pattern and Index of Concor­ 5 have Index of Concordance values in Accident 1 and 2 below the
dance values above the acceptable threshold in all three accidents. The acceptable threshold in intra-rater reliability. This was an outlier found
participant 5 have Index of Concordance values in Accident 1 and 2 in the data because he reached Index of Concordance 100% in Accident
below the acceptable threshold, 65.7% and 47.8%, respectively. How­ 3. Therefore, this did not impact the reliability and practical application
ever; he reached Index of Concordance 100% in Accident 3. In addition, of the overall framework, because the results of the validity and reli­
there was no obvious relationship amongst the days between time 1 and ability evaluation demonstrated that the classification framework meets
2 and the index of concordance . the acceptable criteria (over 70% index of concordance) for validity and
inter and intra reliability.
3.3. Participant’s responses on utility and usability To sum up, this study developed a classification framework for ac­
cident investigation in construction and evaluated its comprehensive­
The participants were slightly more positive about the utility of the ness, reliability, and validity (O’Connor, 2008; Olsen and Shorrock,
framework than the usability. All of the participants strongly agreed 2010). Not surprisingly, the results demonstrate that contributing fac­
with the following statements: the framework is useful in learning from tors identified at the low levels tend to generate the most agreement
the accident; the framework can help in identifying failure in con­ between accident investigation experts and the construction engineers,
struction safety; The framework can help to make recommendations to but experts tend to identify more factors at the high levels (e.g., gov­
strengthen the construction safety. By contrast, only 60% of the ernment and regulatory body levels) than the construction engineers.
This most likely due to the fact that construction engineers with a
Table 4 technical or operational background are more familiar with factors at
Validity test: Index of Concordance (%) among participants and expert at time 1 the low levels and less familiar with higher level factors e.g. This in­
and 2. dicates that the investigators’ skills and experiences influence the way
Accident Mean overall the contributory factors are identified (Goncalves Filho et al., 2019). i.e.
they are inclined to focus on areas of their own expertise/experience. In
Time 1 Standard Deviation Time 2 Standard Deviation
order to avoid this tendency, it is important to provide additional
1 74.2 3.2 73.4 3.5 training to make up for the areas the investigators are not familiar with.
2 71.3 2.3 70.6 2.1
Alternatively, involving multiple investigators with complementary
3 74.0 4.6 73.4 4.3
Mean 73.2 72.5 expertise/experiences involved in accident investigation might be

7
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

effective. To a certain degree, the application of the contributing factor inadequate inspection policy and inadequate recruitment policy of inspectors
framework may weaken the effects of subjective bias or pre-existing as contributing factors at government level, which might not be strongly
world view or the previous experience of investigators. It may also applicable in some other countries. It may be necessary to revise the
reduce the tendency for accident investigator’s to apply what Lundberg classification framework to include broader contributing factors or
et al. (2009) call the principle of ’what you look for is what you find’, that refine it when applied in different contexts. Secondly, the number of
is over relying on information which is easier to find rather than infor­ participants and accident scenarios used for the validity and reliability
mation which may take more time or effort. assessment were still very limited. The literature does not suggest the
minimum number of accident reports to be used for validity and reli­
4.2. Contributions to literature and practice ability evaluation, but it might be important to further develop a
rigorous and practical methodology for the validity and reliability
The findings from this study make two main contributions to the evaluation.
current literature and accident investigation practice. First, the
contributing factor framework developed in this study is an advance on 5. Conclusion
the existing models in the literature considering its holistic coverage in
breadth and depth. The framework evaluation also demonstrates its This study set out to develop a comprehensive, valid and reliable
validity and reliability to a certain degree. Second, this study explores contributing factor classification framework to support systemic acci­
ways to systematically evaluate the validity and reliability of systemic dent analysis in construction. The framework provided an opportunity
accident models. The evaluation in this study involved practitioners who to for in-depth analysis of contributing factors at each of six levels of the
work within the construction domain (see Section 2.2 – Phase 2) in line socio-technical system. The study explored a rigorous and practical way
with what Wallace and Ross (2006) and Olsen (2011) have argued; the to evaluate the validity and reliability of a systemic accident model and
participants used in the validity and reliability assessment should be demonstrated that the validity and reliability of the framework devel­
those who will use the classification system in the “real world”. oped in this study reached the acceptable threshold of the overall Index
Furthermore, the evaluation was made using real accident investigation of Concordance to a certain degree. The findings from our study fill a gap
rather than coding of pre-existing incident reports. This was an attempt existing in the literature for the construction sector and provide a
to minimise some potential limitations (e.g., lack of information on type practical framework to be applied for the construction accident analysis
of training to participants) that have impacted previous studies on val­ in Brazil, but the framework developed in this study should be consid­
idity and reliability (Olsen, 2011, 2013). ered as a starting point for further application, evaluation and refine­
In practical terms, as mentioned early, systemic accident analysis in ment. We would also argue that the classification framework needs to be
depth and breadth has not been developed for accident investigation in further evaluated under more natural conditions by different
construction sector. The validity and reliability of the accident analysis practitioners.
method have not been properly evaluated in the past. The contributing
factor classification framework developed in this study overcome these Declaration of Competing Interest
limitations by considering the influences of multiple detailed factors at
each of six levels. The framework is underpinned by the principles of The authors declare that they have no known competing financial
contemporary accident causation (Love et al., 2016; Salmon, 2016; interests or personal relationships that could have appeared to influence
Underwood and Waterson, 2013). The framework can help to identify the work reported in this paper.
contributory factors and actors as across the whole system and conse­
quently enables a better understanding of the entire construction sys­ References
tem. In addition, construction practitioners frequently find it difficult to
use complex techniques for accident investigation (Taroun, 2014; Un­ Abdelhamid, T.S., Everett, J.G., 2000. Identifying root causes of construction accidents.
derwood and Waterson, 2013). In the case of the classification frame­ J. Construct. Eng. Manage. 126 (1), 52–60.
Aksorn, T., Hadikusumo, B.H.W., 2008. Critical success factors influencing safety
work described in this paper we would argue that it is relatively simple
program performance in Thai construction projects. Saf. Sci. 46, 709–727.
and straightforward to learn and use. Because the classification frame­ AEAT, 2018. Anuário Estatístico de Acidentes do Trabalho. Published by Ministry of
work was designed for use in real world environments; therefore, Social Securityed by (Accessed 30 September 2019).
reasonable reliability may be achieved despite the limitations of those Böiger, F., Wright, G., 1992. Reliability and validity in expert judgment. In: Wright, G.,
Bolger, F. (Eds.), Expertise and Decision Support. Plenum Press, New York.
environments (i.e. limitations in training time and resources or position Branford, K., 2007. An Investigation into the Validity and Reliability of the Accimap
and experience of personnel conducting coding). Finally, the framework Approach. PhD Thesis. Australian National University, Canberra.
could also be used to categorise information, identify and monitor trends Branford, K., 2011. Seeing the big picture of mishaps: applying the Accimap approach to
analyse system accidents. Aviat. Psychol. Appl. Human Factors 1 (1), 31–37.
over time, and act as what Star and Griesemer (1989) describe as a Bureau of Labor Statistics, 2018. Illnesses, injuries and fatalities, https://www.bls.
‘boundary object’, that is a means of providing a common reference gov/iif/ (accessed date: 16 September 2019).
point for accident investigators, as well as supporting the process of Canham, A., Jun, G.T., Waterson, P., Khalid, S., 2018. Integrating systemic accident
analysis into patient safety incident investigation practices. Appl. Ergon. 72, 1–9.
longer-term sensemaking involved in investigation (Weick, 1995; Carael, M., 2001. Assessing the validity and reliability of self-reported behavioral data.
Macrae, 2019). In: Magnani, R. (Ed.), Evaluating Programs for HIV/AIDS Prevention and Care in
Developing Countries: A Handbook for Program Managers and Decision. Makers.
Arlington Virginia, Family Health International.
4.3. Limitations and future work Carmines, E.G., Zeller, R.A., 1979. Reliability and Validity Assessment. Sage
Publications, London, England.
The classification framework inevitably has some limitations, which Carrillo-Castrillo, J.A., Trillo-Cabello, A.F., Rubio-Romero, J.C., 2017. Construction
accidents: identification of the main associations between causes, mechanisms and
can suggest some directions for future studies. Firstly, the classification stages of the construction process. Int. J. Occupat. Safety Ergon. 23 (2), 240–250.
framework was developed based on the comprehensive literature review Cohen, T.N., Wiegmann, D.A., Shappell, S.A., 2015. Evaluating the reliability of the
and the analysis of accident reports which were selected by experts, human factors analysis and classification system. Aerosp. Med. Hum. Perform. 86
(8), 728–735.
Brazilian federal inspectors. Consequently, some contributing factors
Choudhry, R.M., Fang, D., 2008. Why operatives engage in unsafe work behavior:
included in the framework developed in this study may be specific to the investigating factors on construction sites. Saf. Sci. 46, 566–584.
Brazilian context. The types of accidents and their frequency in different Dekker, S., Cilliers, P., Hofmeyr, J.H., 2011. The complexity of failure: Implications of
countries are likely to be influenced by their socio-economic factors and complexity theory for safety investigations. Saf. Sci. 49 (6), 939–945.
Ergai, A., Cohen, T., Sharp, J., Wiegmann, D., Gramopadhye, A., Shappell, S., 2016.
national characteristics. For example, the experts in this study included Assessment of the human factors analysis and classification system (HFACS): intra-
poor or lack communication of legislations, poor infrastructure to inspection, rater and inter-rater reliability. Saf. Sci. 82, 393–398.

8
A.P. Goncalves Filho et al. Safety Science 140 (2021) 105303

França, J.E.M., Hollnagel, E., Santos, I.J.A.L., Haddad, A.N., 2020. FRAM AHP approach Reason, J., 1990. Human Error. Cambridge University Press, Cambridge.
to analyse offshore oil well drilling and construction focused on human factors. Rodrigues, F., Coutinho, A., Cardoso, C., 2015. Correlation of causal factors that
Cogn. Technol. Work 22, 653–665. influence construction safety performance: a model. Work 51 (4), 721–730.
Garrett, J.W., Teizer, J., 2009. Human factors analysis classification system relating to Rosa, L.F., Haddad, A.N., Carvalho, P.V.R., 2015. Assessing risk in sustainable
human error awareness taxonomy in construction safety. J. Construct. Eng. Manage. construction using the Functional Resonance Analysis Method (FRAM). Cogn.
135 (8), 754–763. Technol. Work 17, 559–573.
Gibb, A.G., Haslam, R., Gyi, D.E., Hide, S., Duff, R., 2006. What causes accidents? Proc. Ross, A.J., Wallace, B., Davies, J.B., 2004. Technical note: measurement issues in
ICE – Civil Eng. 159 (6), 46–50. taxonomic reliability. Saf. Sci. 42 (8), 771–778.
Goncalves Filho, A.P., Jun, G.T., Waterson, P., 2019. Four studies, two methods, one Salmon, P.M., 2016. Bridging the gap between research and practice in ergonomics
accident – An examination of the reliability and validity of Accimap and STAMP for methods: Part II. Methodological issues in ergonomics science. Theor. Issues Ergon.
accident analysis. Safety Sci. 113, 310–317. Sci. 17 (5–6), 459–467.
Goode, N., Salmon, P.M., Taylor, N.Z., Lenné, N.G., Finch, C.F., 2017. Developing a Salmon, P.M., Goode, N., Lenn, M.G., Finch, C.F., Cassell, E., 2014. Injury causation in
contributing factor classification scheme for Rasmussen’s Accimap: Reliability and the great outdoors: a systems analysis of led outdoor activity injury incidents. Accid.
validity evaluation. Appl. Ergon. 64, 14–26. Anal. Prevent. 63, 111–120.
Hale, A., Walker, D., Walters, N., Bolt, H., 2012. Developing the understanding of Sawacha, E., Naoum, S., Fong, D., 1999. Factors affecting safety performance on
underlying causes of construction fatal accidents. Saf. Sci. 50 (10), 2020–2027. construction sites. Int. J. Project Manage. 17, 309–315.
Hallowell, M.R., Gambatese, J.A., 2009. Construction safety risk mitigation. J. Construct. Siraj, N.B., Fayek, A.R., 2019. Risk identification and common risks in construction:
Eng. Manage. 135 (12), 1316–1323. literature review and content analysis. J. Construct. Eng. Manage. 145 (9), 1–13.
Harvey, E.J., Waterson, P., Dainty, A.R.J., 2018. Beyond ConCA: rethinking causality and Star, S.L., Griesemer, J.R., 1989. Institutional Ecology, `Translations’ and Boundary
construction accidents. Appl. Ergon. 73, 118–121. Objects: Amateurs and Professionals in Berkeley’s Museum of Vertebrate Zoology,
Haslam, R.A., Hide, S.A., Gibb, A.G.F., Gyi, D.E., Pavitt, T., Atkinson, S., Duff, A.R., 2005. 1907-39. Soc. Stud. Sci. 19 (3), 387–420.
Contributing factors in construction accidents. Appl. Ergon. 36 (4), 401–415. Suraji, A., Duff, A.R., Peckitt, S.J., 2001. Development of causal model of construction
Hollnagel, E., 2004. Barriers and Accident Prevention. Ashgate, Aldershot. accident causation. J. Construct. Eng. Manage. 127 (4), 337–344.
HSE, 2003. Research Report 156. Health and Safety Executive, UK. Stanton, N.A., 2016. On the reliability and validity of, and training in, ergonomics
HSE, 2019. Health and Safety in Construction in Great Britain. methods: a challenge revisited. Theor. Issues Ergon. Sci. 17 (4), 345–353.
Hulme, A., Stanton, N.A., Walker, G.H., Waterson, P., Salmon, P.M., 2019. What do Stanton, N.A., Young, M.S., 2003. Giving ergonomics away? The application of
applications of systems thinking accident analysis methods tell us about accident ergonomics methods by novices. Appl. Ergon. 34 (5), 479–490.
causation? A systematic review of applications between 1990 and 2018. Saf. Sci. Svedung, I., Rasmussen, J., 2002. Graphic representation of accident scenarios: mapping
117, 164–183. system structure and the causation of accidents. Saf. Sci. 40, 397–417.
ILO, 2017. Inspections on Construction. A Guide for Labour Inspector. International Tabibzadeh, M., Meshkati, N., 2015. Applying the AcciMap Methodology to Investigate a
Labour Organisation, Geneva, Switzerland. Major Accident in Offshore Drilling: A Systematic Risk Management Framework for
Johnson, C.W., de Almeida, I.M., 2008. An investigation into the loss of the Brazilian Oil and Gas Industry. In: SPE Western Regional Meeting. Society of Petroleum
space programme’s launch vehicle VLS-1 V03. Saf. Sci. 46 (1), 38–53. Engineers.
Kee, D., Jun, G.T., Waterson, P.E., Haslam, R., 2016. A systemic analysis of South Korea Taroun, A., 2014. Towards a better modelling and assessment of construction risk:
Sewol Ferry accident - striking a balance between learning and accountability. Appl. insights from a literature review. Int. J. Project Manage. 32, 101–115.
Ergon. 1–13. Teo, E.A.L., Ling, F.Y.Y., Chong, A.F.W., 2005. Framework for project managers to
Kim, T., Nazir, S., Øvergård, K.I., 2016. A STAMP-based causal analysis of the Korean manage construction safety. Int. J. Project Manage. 23 (4), 329–341.
Sewol ferry accident. Saf. Sci. 83, 93–101. Tsang, Y.T., Fung, I.W.H., Tam, V.W.Y., Sing, C.P., Lu, C.T., 2017. Development of an
Khosravi, Y., Asilian-Mahabadi, H., Hajizadeh, E., Hassanzadeh-Rangi, N., Bastani, H., accident modelling in the Hong Kong construction industry. Int. J. Construct.
Behzadan, A.H., 2014. Factors influencing unsafe behaviors and accidents on Manage. 17 (2), 124–131.
construction sites: a review. Int. J. Occupat. Safety Ergon. 20 (1), 111–125. Umeokafor, N., Windapo, A., Evangelinos, K., 2019. Causal inferences of
Lee, S., Moh, Y.B., Tabibzadeh, M., Meshkati, N., 2016. Applying the Accimap external–contextual domains on complex construction, safety, health and
methodology to investigate the tragic Sewol Ferry accident in South Korea. Appl. environment regulation. Saf. Sci. 118, 379–388.
Ergon. 1–9. Underwood, P., Waterson, P.E., 2013. Systemic accident analysis: examining the gap
Leveson, N., 2004. A new accident model for engineering safer systems. Saf. Sci. 42, between research and practice. Accid. Anal. Prev. 55, 154–164.
237–270. Underwood, P., Waterson, P., 2014. Systems thinking, the Swiss Cheese Model and
Love, P.E.D., Ding, L., Luo, H., 2016. Systems thinking in workplace safety and health in accident analysis: a comparative systemic analysis of the Grayrigg train derailment
construction: bridging the gap between theory and practice. Accid. Anal. Prev. 93, using the ATSB, AcciMap and STAMP models. Accid. Anal. Prev. 68, 75–94.
227–229. Wallace, B., Ross, A.J., 2006. Beyond Human Error. CRC Press, U.S.A.
Lundberg, J., Rollenhagen, C., Hollnagel, E., 2009. What-You-Look-For-Is-What-You- Wallace, B., Ross, A., Davies, J., Wright, L., White, M., 2002. The creation of a new minor
Find – the consequences of underlying accident models in eight accident event coding system. Cogn. Technol. Work 4, 1–8.
investigation manuals. Saf. Sci. 47 (10), 1297–1311. Wang, J., Zou, P.X.W., Li, P.P., 2016. Critical factors and paths influencing construction
Macrae, C., 2019. Investigating for improvement? Five strategies to ensure national workers’ safety risk tolerances. Accid. Anal. Prev. 93, 267–279.
patient safety investigations improve patient safety. J. R. Soc. Med. 112 (9), Waterson, P.E., 2009. A systems ergonomics analysis of the Maidstone and Tunbridge
365–369. Wells infection outbreaks. Ergonomics 52, 1196–1205.
Melchior, C., Zanini, R.R., 2019. Mortality per work accident: A literature mapping. Waterson, P.E., Jenkins, D.P., Salmon, P.M., Underwood, P., 2016. ‘Remixing
Safety Science 114, 72–78. Rasmussen’: the evolution of Accimaps within systemic analysis. Appl. Ergon. 59
Mitropoulos, P., Abdelhamid, T.S., Howell, G.A., 2005. Systems model of construction (PartB), 483–503.
accident causation. J. Construct. Eng. Manage. 131 (7), 816–825. Weick, K., 1995. Sensemaking in Organisations. Sage, London.
Mroszczyk, J.W., 2015. Improving construction safety: a team effort. Profess. Safety 60 Winge, S., Albrechtsen, E., 2018. Accident types and barrier failures in the construction
(6), 55. industry. Saf. Sci. 105, 158–166.
O’Connor, P., 2008. HFACS with an additional layer of granularity: validity and utility in Woolley, M.J.I., Goode, N., Read, G.J.M., Salmon, P., 2018. Moving beyond the
accident analysis. Aviat. Space Environ. Med. 79 (6), 599–606. organizational ceiling: do construction accident investigations align with systems
Okoli, N.S., Pawlowski, S.D., 2004. The Delphi method as a research tool: an example, thinking? Human Factors Manage. 28, 397–1308.
design considerations and applications. Inform. Manage. 42, 15–29. Woolley, M.J.I., Goode, N., Read, G.J.M., Salmon, P., 2019. Have we reached the
Olsen, N.S., 2011. Coding ATC incident data using HFACS: inter-coder consensus. Safety organisational ceiling? A review of applied accident causation models, methods and
Sci. 49 (10), 1365–1370. contributing factors in construction. Theor. Issues Ergon. Sci. 20 (5), 533–555.
Olsen, N.S., 2013. Reliability studies of incident coding systems in high hazard Ye, G., Tan, Q., Gong, X., Xiang, Q., Wang, Y., Liu, Q., 2018. Improved HFACS on human
industries: A narrative review of study methodology. Appl. Ergon. 44, 157–184. factors of construction accidents: a China perspective. Adv. Civil Eng. 1–15.
Olsen, N., Shorrock, S., 2010. Evaluation of the HFACS-ADF safety classification system: Xia, N., Zou, P.X.W., Liu, X., Wang, X., Zhu, R., 2018. A hybrid BN-HFACS model for
inter-coder consensus and intra-coder consistency. Accid. Anal. Prev. 42, 437–444. predicting safety performance in construction projects. Saf. Sci. 101, 332–343.
Oswald, D., Sherratt, F., Smith, S., 2013. Exploring factors affecting unsafe behaviours in Zhang, W., Zhu, S., Zhang, X., Zhao, T., 2019. Identification of critical causes of
construction. In: Proceedings 29th Annual Association of Researchers in construction accidents in China using a model based on system thinking and case
Construction Management Conference, ARCOM 2013. Association of Researchers in analysis. Saf. Sci. https://doi.org/10.1016/j.ssci.2019.04.038.
Construction Management, pp. 335–344. Zhou, W., Zao, T., Liu, W., Tang, J., 2018. Identification of critical causes of construction
Rasmussen, J., 1997. Risk management in a dynamic society: a modeling problem. Saf. accidents in China using a model based on system thinking and case analysis. Saf.
Sci. 27 (2), 183–213. Sci. https://doi.org/10.1016/j.ssci.2018.05.001.

You might also like