Professional Documents
Culture Documents
Rozenfeld2010 PDF
Rozenfeld2010 PDF
Safety Science
journal homepage: www.elsevier.com/locate/ssci
a r t i c l e i n f o a b s t r a c t
Article history: Job Safety Analysis (JSA), which is also known as Job Hazard Analysis, is an efficient proactive measure for
Received 29 January 2008 safety risk assessment used in industrial manufacturing settings. However, unlike the manufacturing set-
Received in revised form 17 September tings for which JSA was developed, at construction sites the physical environment is constantly changing,
2009
workers move through the site in the course of their work, and they are often endangered by activities
Accepted 21 December 2009
performed by other teams. To address this difficulty, a structured method for hazard analysis and assess-
ment for construction activities, called ‘‘Construction Job Safety Analysis” (CJSA), was developed. The
method was developed within the framework of research toward a lean approach to safety management
Keywords:
Construction planning
in construction, which required the ability to predict fluctuating safety risk levels in order to support
Health and safety safety conscious planning and pulling of safety management efforts to the places and times where they
Lean construction are most effective. The method involves identification of potential loss-of-control events for detailed
Risk identification stages of the activities commonly performed in construction, and assessment of the probability of occur-
rence for each event identified. It was applied to explore 14 primary construction activities in an exten-
sive trial implementation that included expert workshops and a series of 101 interviews with site
engineers and superintendents. Detailed quantitative results were obtained for a total of 699 possible
loss-of-control events; the most frequent events are those related to exterior work at height.
Ó 2009 Elsevier Ltd. All rights reserved.
0925-7535/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ssci.2009.12.017
492 O. Rozenfeld et al. / Safety Science 48 (2010) 491–498
2004; Gyi et al., 1999; Hinze, 2008; Kartam and Bouz, 1998; Shep- The fundamental change is that accidents are replaced by loss-
herd et al., 2000). of-control events and the potential for any victim to be exposed
Applying lean thinking (Womack and Jones, 2003) to construc- to them. To implement this in practice requires knowledge of con-
tion in this context leads to the hypothesis that, like production struction activity types, including the nature and probability of
control itself, activities to enhance safety should be pulled by cur- loss-of-control events, the impact of environmental intensifying
rent system needs rather than pushed uniformly onto workers and factors, the use of protective gear, and the potential severity of
activities. The CHASTE (Construction Hazard Assessment with Spa- accident scenarios. Each of these must be compiled in a knowl-
tial and Temporal Exposure) approach has been developed in a re- edge base in a form that can be used by software that implements
search project for the Preventive Action Unit of the Industrial Labor the CHASTE approach to compute risk levels for specific construc-
Inspector’s office of the Israel Ministry of Labor (Rozenfeld et al., tion projects. The CJSA method was devised to collect this
2009). The basic idea of the CHASTE is that although construction knowledge.
projects as a whole are unique and dynamic, individual construc-
tion tasks and methods are fairly well-defined and expected. For
1.2. Job Safety Analysis
example – pouring concrete using a crane on site is a common well
understood trade activity, but the level of risk associated with it
The process of JSA includes three main stages (Chao and Hen-
can differ depending on its context. At one site, it may be per-
shaw, 2002):
formed at the end of the day when no other tasks are being per-
formed, while at another, it might be performed at the middle of
(1) Identification – choosing a specific job or activity and break-
the day when many other workers are located either on or below
ing it down into a sequence of stages, and then, identifying
the element being cast.
all possible loss-of-control incident that may occur during
By separating the potential for loss-of-control from the poten-
the work.
tial for presence of victims, it becomes possible to compute a
(2) Assessment – evaluating the relative level of risk for all the
time-dependent risk level forecast, using a database of probabil-
identified incidents.
ities of loss-of-control for standard work methods, coupled with
(3) Action – controlling the risk by taking sufficient measures to
site-specific computation of workers’ exposure to possible loss-
reduce or eliminate it.
of-control events. The result is a more accurate assessment of
actual risks than is available using current methods, such as Pre-
For determining a priority order of treatment, the level of each
liminary Hazard Analysis (PHA) (Elzarka et al., 1995; Hansen,
incident risk is evaluated by assessing the incident’s probability of
1993; Saurin et al., 2004) which disregards the exposure factor.
occurrence and its expected outcome (the level of injury). Those
The predicted risk levels can be computed for various planning
two measures place the risk in a standard scale from most negligi-
windows, and used either to pull safety interventions or to
ble to the most severe.
change production plans, both of which enhance safety. Thus
In essence, the JSA method has proven to be effective for plan-
management efforts to enhance safety can be less wasteful and
ning the safest way to perform a task (Holt, 2001). However, in its
more effective.
current form, it is impractical for the construction industry. Unlike
A statistical approach based on historic accident data is
other industries, construction projects are highly dynamic; the
unsuitable for computing the risk levels needed in the trade
production environment changes in time and place, and work
method risk database for two main reasons. Firstly, the CHASTE
crews change frequently. Moreover, construction products are un-
approach considers location, exposure to other teams, work
ique, and are almost always prototypical; standardized procedures
method, and personal factors to assess risk levels, producing
that may be considered safe in one project may be hazardous in the
very large combinatorial number of possible accident scenarios.
environment of a different project. Another drawback of the tradi-
The number of documented accidents is many orders of magni-
tional JSA is that in construction, workers commonly endanger
tude smaller than the number of potential accident scenarios, so
other workers, who may be performing a different activity at a dif-
that for most scenarios, the sample size of accidents recorded
ferent location. The standard JSA method is not designed to reveal
would be zero or too small to be considered statistically signifi-
these dangers since it focuses on production activities in isolation,
cant. Secondly, we are concerned with the probability of loss-of-
at predetermined workstations. For these reasons, a different
control while performing a task rather than with the probability
method is needed for construction in general, and to support the
of an accident occurring. For every serious construction accident,
CHASTE approach in particular. This research proposes an im-
there are multiple actual dangerous events (near misses) that
proved technique, called Construction Job Safety Analysis (CJSA),
end with no injury (Shapira and Lyachin, 2008); these should
in which the job analysis is performed independently of any spe-
be taken into account when assessing loss-of-control risk levels,
cific consideration of time and place. This is achieved by separating
but the vast majority are not recorded and do not appear in sta-
the loss-of-control that precedes any accident from the potential
tistical records. Hence, an approach based on aggregated acci-
presence of a victim in the path of harm. Loss-of-control events
dent statistics is not suitable and cannot be used to build a
are assessed in the CJSA, which is generic across any local construc-
database useful for assessing the likelihood of loss-of-control
tion industry, while exposure of potential victims in time and
events at any particular place during any particular time frame
space is assessed for specific construction projects.
on a construction site.
To overcome this problem, a different conceptual approach
was adopted in the development of CHASTE. Instead of assessing 2. CJSA process
risk as a function of likelihood of an accident and its potential
severity (two parameters), the risk level was divided into three The Construction Job Safety Analysis (CJSA) method generates a
parameters: large knowledge-base describing all possible loss-of-control events
in construction. The knowledge is structured in a form that can be
1. The probability of a loss-of-control event occurring. used by software implementing the CHASTE approach to compute
2. The exposure of potential victims in time and in space. the predicted levels of risk for the activities of specific projects, by
3. The likely severity level of an accident (which is also dependent using a three-dimensional building model and a construction
on the use of personal safety gear). schedule.
O. Rozenfeld et al. / Safety Science 48 (2010) 491–498 493
The CJSA process comprises three major steps: The last step of the workshop is to identify all the possible loss-
of-control events that may occur during each working stage of the
Step 1: identify hazards – identify the set of direct and support- activity, regardless of their likelihood.
ing construction activities needed for a domain, define their pro- Finally, the researchers compile a set of accident scenario types
cedures, and analyze all possible loss-of-control events that may (Table 1), and match each loss-of-control event recorded with one
occur during their execution. or more types. Association with accident types is needed so that
Step 2: assess probability – evaluate the likelihood of occurrence the potential victims of each loss-of-control event can be identified
of each loss-of-control event, the levels of possible intensifying when a project’s risk levels are calculated. For any accident sce-
factors, and the likelihood of use of personal safety gear. nario type, workers who are adjacent to the loss-of-control event,
Step 3: assess severity – associate the possible loss-of-control below it, and/or above it may be exposed to the hazard. For some
events with possible accident scenarios, and assess the expected types, only the workers performing the activity in question are ex-
degree of severity for each type of accident scenario. posed – these are classified as ‘self impacting only’. The third col-
umn of Table 1 provides these logical relationships.
2.1. CJSA step 1 – Identification In order to classify the accident scenario types, it must be pos-
sible to calculate the level of exposure for each type as a function of
The first step of the CJSA process is performed in a set of work- the geometric relationships between the locations and any equip-
shops in which the researchers interview experts in the execution ment involved. This requires a unique algorithm for each class of
of construction activities, usually senior construction superinten- accident types. The necessary exposure algorithms have been
dents. The activities relevant to the domain being explored (e.g. developed and their application has been tested. Details can be
‘multistory residential construction’) are identified, and each ex- found in Sacks et al. (2009).
pert is asked to analyze one or more activities with which they
are familiar. The procedure is detailed in Fig. 1, and an example
of the activity breakdown structure is shown, together with an
example, in Fig. 2. 2.2. CJSA step 2 – Assessment
The experts begin by dividing each activity into sub-activities.
They determine the start and finish times of each sub-activity in The second step of the CJSA procedure seeks to determine the
relation to the overall activity duration as it would be defined in following information about the activities that were detailed in
a construction plan. Values are set as percentages of the planned the first step:
duration (activity start = 0%, activity end = 100%). In some cases,
sub-activities may begin before 0% of the activity (e.g. preparation 1. The expected rate of occurrence for each possible loss-of-con-
of scaffolding), and some may end after 100% of the activity (e.g. trol event.
curing concrete, clearing formwork). 2. The degree of influence of the different managerial and environ-
Next, each sub-activity is divided into work stages. The likely mental factors that affect the expected rates of occurrence.
composition of the work teams and their expected locations 3. The expected degree of use of personal safety gear.
while performing the task are detailed. Each stage may involve
work at a primary location (usually defined in a construction Fig. 3 provides a flow chart for this step. The information is col-
plan) and at secondary locations, such as storage or loading lected by means of an extensive survey that is conducted through
areas. It is unreasonable to set fixed relative periods for the face to face interviews with construction superintendents. The sur-
stages at this level of detail, and they may be repetitive, thus vey instrument is a set of structured questionnaires, one for each
only the proportions of their durations within the sub-activities activity type. The questionnaires can be produced automatically
need be recorded. from the database of activities, sub-activities, stages, typical loss-
Fig. 2. Information items defined in the knowledge base, relationships, and examples.
Table 1
Accident scenario type.
on the rate of the accident occurrence (e.g. Hide et al. (2003) who 4. Death.
examined the influence of causal factors on 100 documented acci-
dents, and Hinze and Raboud (1988) who found a relationship be- For each of these four possible outcomes, the expert assesses its
tween project’s attributes and safety performance). A summary likelihood relative to the other possible outcomes, as a percentage
report of all claims for compensation from construction work re- part, so that the cumulative likelihood of all the possible outcomes
lated accidents (Bar et al., 2005) revealed significant tendencies for a single event is 100%. Two distributions are assessed for each
of factors affecting accident rates. For example, it showed that con- type; one assuming use of personal safety gear and the other
struction workers were most likely to be injured on the first work- assuming absence of the appropriate gear. The assessment is based
ing day of the week, and that the number of days of absence tended on the assessor’s professional experience.
to decrease as company size increased. Some factors, such as per- The distributions are used in the CHASTE approach to calculate
sonal risk aversion, personal discipline, schedule delays and others, a single severity level, which is multiplied by the likelihood of
are difficult to monitor, and so are not of practical use in predicting loss-of-control and the exposure levels for each event in order to
risk levels. calculate an overall risk level estimate. To transform the severity
The CJSA method acknowledges the importance of these factors distributions into single weighted values, a set of weighting factors
and their integration in any application of the CHASTE approach. In must be applied to the severity levels, which express the relative
the trial implementation described below, four specific factors importance a risk assessor attributes to death, say, in relation to
(schedule delays; a work group’s first day on site; crowding of other injuries. This is a value judgment, which must be made by
workers in the work area; and short notice before work begins) the end user. An example of four possible weight values can be
were tested for because they were of particular interest for re- seen in the second column, ‘‘Severity weight”, of Table 2. In this
search of the application of Lean Construction on the site in which example, the user determined a scale for severity from 1 (for the
the CHASTE method was implemented. Future users of the CJSA lowest level) to 100 (for the highest), and set intermediate values
method should select factors relevant to the context of their indus- of 5 and 25 for levels 2 and 3 respectively.
try in order to increase the reliability of the model. The example provided in Table 2 illustrates the overall proce-
Although these factors are collected, a word of caution is dure used for setting the expected severity level for the case of
needed: little empirical evidence is available to calibrate their im- an accident scenario ‘‘falling from over 5 m height” while ‘‘casting
pact, and the professional literature does not provide indications of concrete for exterior walls using industrialized forms”. The values
the potential correlations between these and other factors. that derive from step 2 of the CJSA survey appear in the ‘‘Expected
The final aspect tested for in step 2 of CJSA is the expected de- occurrence” column. Two distributions are provided, reflecting the
gree of use of personal safety gear during each particular activity, likelihood of occurrence of each outcome dependent on the use or
including helmets, appropriate working shoes, gloves, safety har- non-use of personal safety gear (a safety harness in this case). The
nesses, and safety goggles. This information is needed for the likelihood of use is 33%, and of non-use is 67%. The resulting
CHASTE model, because the model assumes that the degree of weighted severity level is 52.6 (out of maximum possible value
severity of injury resulting from any possible accident will be dis- of 100).
tributed differently depending on the potential victim’s use of per-
sonal safety gear. The expected severity outcome for any event is
3. Trial implementation
the weighted average of the two severity distributions that result
when safety gear is used and when it is not. In the questionnaire,
The CJSA method was developed and first applied in practice
every respondent is asked to assess the expected level of use of rel-
within the framework of the CHASTE research project. The scope
evant personal safety gear by the workers in each production stage.
for this implementation covered 14 common construction activi-
This information reflects the sample population; during implemen-
ties from all phases of a typical multi-story building project.
tation for any particular project, a safety manager can adjust the
levels of expected use of safety gear to local conditions.
3.1. Step 1 – Identification
2.3. CJSA step 3 – Assess Severity
In step 1, the knowledge was elicited in a series of workshops
The final step of the CJSA method determines the relative prob- with safety experts and senior site managers, who are legally
abilities of severity for each accident scenario type. The distribu- responsible for site-safety. Each expert was asked to analyze a sin-
tions are obtained by asking safety expert interviewees to gle construction activity according to his or her experience.
distribute the likelihood of the severity of the outcome for each Table 3 lists the range of activities covered with details of the
type among four distinct possible outcomes: number of work stages and loss-of-control events that were iden-
tified for each activity. After sorting, filtering and dismissing over-
1. Minor injury (up to one day of absence) – scratch, wound. lapping data, 699 different possible loss-of-control events were
2. Medium injury (long absence) – burn, fracture. defined for 14 construction activities, out of the 875 loss-of-control
3. Severe injury – permanent disability. events enumerated in Table 3.
Table 2
Calculating the expected severity of a fall from above 5 m height while casting concrete for exterior walls using industrialized forms.
Table 3
Activity analysis summary.
3.2. Step 2 – Assessment per year of work per person, i.e. the expected number of times a
single event might occur, if a single worker performs a single task
The population for the survey in step 2 consisted of 91 senior for a time period of one year. Table 4 lists five sample events out of
superintendents from 45 construction companies. The majority the total 699 events that could arise from the 14 construction
were interviewed in depth about a single construction activity type. activities covered in the survey. The table includes the most fre-
A small number of them were interviewed twice, because they were quent and the most infrequent events.
familiar with more than one activity type; a total of 101 interviews The knowledge elicited from the expert workshops (in step 1)
were conducted. The questionnaires were filtered by comparing the included data describing the proportional duration of every work
separate descriptive and numeric responses of each interviewee to stage for each sub-activity. In any given construction project, the
the same questions, and by examining whether the responses were planned start and finish times for the sub-activities can be ob-
logical and complete. Of the 101 interviews, 14 were rejected due to tained from the schedule, and so the expected duration of each
inconsistencies that indicated misunderstanding of the question- work stage can be determined. Multiplying the duration of a stage
naire, leaving 87 valid complete questionnaires. by the average event frequency and the number of workers in the
The respondents’ average period of construction experience was work group gives the number of event occurrences expected for
21.8 years. In terms of company size, 44% were employed in small the stage during any period of activity.
firms (up to 50 employees), 26% in medium-sized firms (51–200 To compare the results of the survey between different activi-
employees), and 30% worked for large firms (over 201 employees). ties and types of events, the likelihood of occurrence was calcu-
7% of the respondents were working on small projects (up to lated for all loss-of-control events as if each one of the 14
1500 m2), 61% on medium-sized projects (1500–7500 m2), 21% activities was being performed continuously for one year, and
on medium-large projects (7500–15,000 m2), and 11% worked on the work stage duration was set respectively. The expected num-
large projects (over 15,000 m2). ber of events for each activity is shown in Fig. 4. The most hazard-
ous activity in terms of expected number of event occurrences is
3.2.1. Likelihood of loss-of-control event occurrences the application of exterior stucco, with 704 expected event occur-
Average values for likelihood of occurrence for all loss-of-con- rences for a single worker over a year (this is equivalent to a single
trol events were summarized in measures of number of events plasterer causing an average of more than three loss-of-control
Table 4
Eight examples of event likelihood, including the most frequent and infrequent loss-of-control events.
800 control events would be increased during each work stage of the
entire activity, in the presence of each of the following intensifying
704
factors: schedule delays; a work group’s first day on site; crowding
of workers in the work area; and short notice before work begins.
As previously mentioned, the intensifying factors should be de-
600 fined by the user in relation to the nature and context of the work
analyzed; these particular factors were chosen in this research due
Event likelihood
ls
ng
g
o
ng
ls
ry
ab
ab
single work stage (out of more than 300 work stages) was mea-
n
io
fin
ilin
al
cc
lin
al
ar
di
di
on
at
zi
w
sl
sl
oo
w
rt
tu
Pi
ad
ad
bo
la
ul
as
d
e
st
rs
oo
pr
st
G
ns
cl
cl
an
et
m
ca
ca
er
rio
Fl
e
cr
fi
su
k
ns
e-
at
on
on
e-
oo
on
te
ic
yp
Pr
W
m
Pr
Br
St
st
Ex
R
C
ith
co
w
e
ls
any of the factors: examples are curing of stucco layers and tying
et
al
cr
w
on
Activity found for similar work stages belonging to other activities. This re-
flects high reliability in spite of the small sample for each activity,
Fig. 4. Number of event occurrences during a year of work according to activity since different respondents answered for different activities.
type.
4. Conclusions
events a day, assuming 220 working days per year). As can be seen,
activities with high levels of loss-of-control event occurrence are The CHASTE approach represents a progressive way to evaluate
those performed outside and at height, whilst activities performed risks in construction. It confronts the difficulties and unique haz-
indoors have relatively low levels of occurrence. The least frequent ards of the construction industry by considering likelihood of
event type is ‘floor collapse’, which according to the interviewees is loss-of-control events and exposure of potential victims to their
expected to occur once in ten years of continuous work. Although it consequences separately. The CJSA method provides a mechanism
is very rare, the expected outcome of this event is disastrous and for collecting the extensive knowledge of the likelihood of loss-of-
therefore, it is quite a significant risk. control events in construction that is needed for implementation of
the CHASTE approach. The CJSA method is loosely based on the
3.2.2. Intensifying factors standard JSA approach to safety planning in manufacturing; it cov-
Implementation of the CJSA assessment step included examina- ers the first two stages of traditional JSA (identification and assess-
tion of factors affecting the expected likelihood of occurrence of ment), but does not extend to the ‘action’ stage, taken in order to
loss-of-control events. The respondents in step 2 were asked to as- reduce or eliminate the risk level, as defined by Chao and Henshaw
sess, based on their past experience, how the likelihood of loss-of- (2002).
5.00
4.00
3.00
2.00
1.00
0.00
C ma n
y
bs
st er g
la n
g
ls
ds
ng
ns ng
g
g
cr onr
tio
ith bs din
ec ctio
in
cc
io
fin
al
rio in
a
ar
zi
i
dd
l
at
w
yp r stu
i
ec
sl
te Pil
oo
t
la
ls sla lad
s
bo
on e
ul
R oor
d
e
G
er
pr
an
et
Fl
ls
er
e
fi
k
su
ns
al
in e-c ton
ic
oo
at
on
tw
Br
Ex
m
w
S
G
lu
al st
as
co
gw a
w
c
e-
e
et
Pr
rm Pr
cr
on
C
Fo
Activity
The CJSA method described was implemented for the con- Bobick, T.G., 2004. Falls through roof and floor openings and surfaces, including
skylights: 1992–2000. Journal of Construction Engineering and Management
struction activities and methods typical of the Israeli building
ASCE 130 (6), 895–907.
construction industry, and a comprehensive analysis was con- Brown, D.B., 1976. System Analysis & Design for Safety. Prentice-Hall Inc.,
ducted of its results. A number of lessons were learned from its Englewood Cliffs, New Jersey.
implementation: Chao, E.L., Henshaw, J.L., 2002. Job Hazard Analysis. OSHA Publication 3071 2002
(Revised). Occupational Safety and Health Administration, US Department of
Labor, Washington.
The method is tractable, despite the large number of individual Elzarka, H.M., Minkarah, I.A., Pulikal, R., 1995. A knowledge-based approach for
loss-of-control events that must be explored. In step 2, assess- automating construction safety management. Computing in Civil Engineering 2,
997–1002.
ment of likelihood of loss-of-control events, each interviewee Findley, M., Smith, S., Kress, T., Petty, G., Enoch, K., 2004. Safety program elements in
was able to respond to up to 85 events within 2 h. construction. Professional Safety 49, 14–22.
The need to obtain measures of likelihood of loss-of-control Goetsch, D.L., 1996. Occupational Safety and Health in the Age of High Technology.
Prentice-Hall, Englewood Cliffs, New Jersey.
events, rather than of accident occurrence, meant that the inter- Gyi, D.E., Gibb, A.G.F., Haslam, R.A., 1999. The quality of accident and health data in
viewers had to explain the principle to each interviewee thor- the construction industry: interviews with senior managers. Construction
oughly in order to avoid responses based on misconceptions. Management and Economics 17, 197–204.
Hansen, L., 1993. Safety management: a call for (r)evolution. Professional Safety 38
(3), 16–21.
The major contribution of the CJSA method is that relative Hide, S., Atkinson, S., Pavitt, T., Haslam, R., Gibb, A., Gyi, D., Duff, R., Suraji, A., 2003.
quantitative measures for each event are obtained. CJSA does not Causal factors in construction accidents. 156, Manchester Centre for Civil and
Construction Engineering, Manchester.
provide risk reduction measures in and of itself. Rather, it supports
Hinze, J., 2008. Construction Safety. Safety Science 46 (4), 565.
the compilation of essential data that is sufficiently rich to support Hinze, J., Raboud, P., 1988. Safety on large building construction projects. Journal of
the CHASTE approach. Construction Engineering and Management, ASCE 114 (2), 286–293.
Through the trial implementation, loss-of-control likelihood Holt, A.S.J., 2001. Principles of Construction Safety. Blackwell Science, London.
Kartam, N.A., Bouz, R.G., 1998. Fatalities and injuries in the Kuwaiti construction
data were collected for 14 common construction activities from industry. Accident Analysis and Prevention 30 (6), 805–814.
all phases of a typical multi-story building project. Not surpris- Mitropoulos, P., Howell, G.A., and Reiser, P., 2003. Workers at the edge; hazard
ingly, the activities with the highest likelihoods of loss-of-control recognition and action. In: 11th International Group for Lean Construction
Conference, Blacksburg, VA, USA.
events were those performed outdoors and at height. Rozenfeld, O., Sacks, R., Rosenfeld, Y., 2009. CHASTE – construction hazard analysis
with spatial and temporal exposure. Construction Management & Economics 27
(7), 625–638.
Sacks, R., Rozenfeld, O., Rosenfeld, Y., 2009. Spatial and temporal exposure to safety
References hazards in construction. ASCE Journal of Construction Engineering and
Management 135 (8), 726–736.
Ahmed, S.M., Kwan, J.C., Ming, F.Y.W., Ho, D.C.P., 2000. Site safety management in Saurin, T.A., Formoso, C.T., Guimaraes, L.B.M., 2004. Safety and production: an
Hong Kong. Journal of Management in Engineering, November 2000, 34–42. integrated planning and control model. Construction Management and
Ale, B.J.M., Bellamy, L.J., Baksteen, H., Damen, M., Goossens, L.H.J., Hale, A.R., Mud, Economics 22, 159–169.
M., Oh, J., Papazoglou, I.A., Whiston, J.Y., 2008. Accidents in the construction Shapira, A., Lyachin, B., 2008. Identification and analysis of factors affecting safety
industry in the Netherlands: an analysis of accident reports using Storybuilder. on construction sites with tower cranes. Journal of Construction Engineering
Reliability Engineering and System Safety 93 (2008), 1523–1533. and Management 135 (1), 24–33.
Bar, S., Shtrosberg, N., Prior, R., and Neon, D., 2005. National Insurance Shepherd, G.W., Kahler, R.J., Cross, J., 2000. Crane fatalities – a taxonomic analysis.
Compensation Claims. Research Report 89. National Insurance Institute, Safety Science 36, 83–93.
Research and Planning Administration, Jerusalem, Israel. <http:// Womack, J.P., Jones, D.T., 2003. Lean Thinking: Banish Waste and Create Wealth in
www.btl.gov.il/SiteCollectionDocuments/btl/Publications/mechkar_89.pdf>. Your Corporation. Free Press, New York.