G

Ca

b

a

ARRA

KBMG

1

hhttffhofiic2odota

ftas

h0

Accident Analysis and Prevention 72 (2014) 359–369

Contents lists available at ScienceDirect

Accident Analysis and Prevention

jo u r n al homepage: www.elsev ier .com/ locate /aap

raphical fault tree analysis for fatal falls in the construction industry

hia-Fen Chia,∗, Syuan-Zih Lina, Ratna Sari Dewia,b

Department of Industrial Management, National Taiwan University of Science and Technology, 43 Keelung Road, Section 4, Taipei 106, TaiwanDepartment of Industrial Engineering, Sepuluh Nopember Institute of Technology, Kampus ITS Sukolilo Surabaya 60111, Indonesia

r t i c l e i n f o

rticle history:eceived 7 February 2014eceived in revised form 14 July 2014ccepted 18 July 2014

eywords:oolean algebra

a b s t r a c t

The current study applied a fault tree analysis to represent the causal relationships among events andcauses that contributed to fatal falls in the construction industry. Four hundred and eleven work-relatedfatalities in the Taiwanese construction industry were analyzed in terms of age, gender, experience, fallingsite, falling height, company size, and the causes for each fatality. Given that most fatal accidents involvemultiple events, the current study coded up to a maximum of three causes for each fall fatality. After theBoolean algebra and minimal cut set analyses, accident causes associated with each falling site can be

inimal cut setraphical icon

presented as a fault tree to provide an overview of the basic causes, which could trigger fall fatalities inthe construction industry. Graphical icons were designed for each falling site along with the associatedaccident causes to illustrate the fault tree in a graphical manner. A graphical fault tree can improve inter-disciplinary discussion of risk management and the communication of accident causation to first linesupervisors.

. Introduction

The construction industry has been identified as one of the mostazardous industries in many parts of the world and falls fromeight are a leading cause of fatalities in the construction indus-ry (Sorock et al., 1993). In Taiwan, Chi and Wu (1997) showedhat falls contributed to more than 30% (377) of 1230 work-relatedatalities. In the United States, construction had the highest count ofatal injuries in 2012 while agriculture, forestry, and the fishing andunting sector had the highest fatal work injury rates (U.S. Bureauf Labor Statistics, 2014). Our previous study developed a classi-cation scheme and categorized 621 fatal falls during 1994–1997

n terms of the cause of the fall, falling site, individual factors, andompany size to derive effective prevention strategies (Chi et al.,005). However, our previous analysis of fatal falls only identifiedne primary cause for each fall fatality. Given that most fatal acci-ents involve multiple events, the current study coded a maximumf three accident causes for each fall fatality (Chi et al., 2009). Faultree and Boolean algebra were applied to reduce accident causesnd the results were presented as a fault tree.

As Ale et al. (2006) suggested, causal modeling is a power-ul tool for gaining insight into the interdependencies betweenhe constituent parts of a complex system. It enables policiesnd inspection regimes to be tailor-made to the vulnerabilities inystems and to those activities that pose the most risk (Ale et al.,

∗ Corresponding author. Tel.: +886 227376338.E-mail address: chris@mail.ntust.edu.tw (C.-F. Chi).

ttp://dx.doi.org/10.1016/j.aap.2014.07.019001-4575/© 2014 Elsevier Ltd. All rights reserved.

© 2014 Elsevier Ltd. All rights reserved.

2006). For example, Haslam et al. (2005) developed a hierarchyof causal influences model in construction accidents to describehow accidents arise from a failure in the interaction between thework team, workplace, equipment, and materials giving rise tothe ‘immediate accident circumstances’. Their model examined‘originating influences’ including the permanent works design,project management, construction processes, safety culture, riskmanagement, client requirements, economic climate and educa-tion provisions together with the ‘shaping factors’. The ‘shapingfactors’ included the actions, behavior, capabilities, and communi-cation of the work team affected by or shaped by their attitudes,motivations, knowledge, skills, supervision, health, fatigue, siteconstraints, work scheduling and housekeeping in a workplace(Haslam et al., 2005). The current study applied a causal modelingtechnique, fault tree analysis, to represent the causal relation-ships among events and causes that contributed to a constructionaccident (Hadipriono, 1992). Depending on the available data, afault tree has been applied to the analysis of a catastrophic event,such as the Schoharie Creek bridge collapse (LeBeau and Wadia-Fascetti, 2007) or on a complicated system, such as a power system(Volkanovski et al., 2009).

Construction of a fault tree normally begins with the top eventand proceeds in a top-down manner (Harms-Ringdahl, 2001). TheAND- and OR-gates are used to provide logical connections between

the basic events. The multiple events that cause a fatality can bedescribed graphically using a fault tree diagram. The fault treesymbol OR-gate, which is equivalent to the Boolean symbol “ + ”represents the union of the events attached to the gate. One or

3 is and

mOaeA(

cctartt(lwpwcTt

2

taafeparvaccylorwo

2

sou(pcaadefpthwcvc

A frequency analysis was performed on each coded variable. Theanalysis indicated that the majority of victims were male (387,94.2%), between 25 and 44 years old (220, 53.5%), worked for

Table 1Boolean algebra fundamental laws (Whitesitt, 1995).

Laws Formula

60 C.-F. Chi et al. / Accident Analys

ore of the input events must occur to cause the event above theR-gate to occur. The fault tree symbol AND-gate which is equiv-lent to the Boolean symbol “•” represents the intersection of thevents attached to the gate. All of the input events attached to theND-gate must exist in order for the event above the gate to occur

Vesely et al., 1981).Given that fatal falls can be caused by more than one cause

ombination, the integration of all possible cause combinationsan be regarded as the union of these cause combinations. Sincehese cause combinations have redundancies, when the same eventppeared more than once, minimal cut set (MCS) was applied toeduce the redundancy of basic events. By examining the MCS,he smallest combination of root causes, an analyst can prioritizeheir prevention measures to prevent the top event from occurringDoytchev and Szwillus, 2009). Therefore, the current study ana-yzed 411 work-related fatal falls that occurred during 2001–2005

ith the same coding scheme (Chi et al., 2005) and coded multi-le causes (up to three) for each fall fatality. A fault tree analysisas applied to present the causal relationships among events and

auses that contributed to a construction accident in terms of MCS.hese MCS can be compared in terms of their relative frequencieso prioritize their importance for prevention measures.

. Materials and methodology

In Taiwan, the Labor Safety and Health Act requires the employero report any occupational fatality to the appropriate inspectiongency within 24 h. Trained inspectors must inspect the workplacend file a formal accident report soon after they are notified. Eachormal accident report covers the type of industry, age, gender,xperience level of the victim, the source of the injury, the com-any size measured by the number of workers, the accident type,nd any other factors that are judged to be relevant. The accidenteport can be used as legal evidence against the employer for anyiolation of the Labor Safety and Health Act. Inspection agencies,ll over Taiwan, must report all occupational fatalities to the Coun-il of Labor Affairs on a monthly basis. The Council of Labor Affairsollects and publishes all occupational fatalities on a monthly andearly basis to alert employers to pay special attention to simi-ar accident scenarios and renew of inspection policy. Similar tour previous studies (Chi et al., 2005, 2009), age, gender, expe-ience, falling site, falling height, company size, and the causesere extracted from accident reports of 411 work-related fatal falls

ccurring during 2001–2005 for further analysis.

.1. Classification of accident causes

Instead of a single primary cause for each fatality, the currenttudy identified up to three causes for each fatality. The causef the fall was coded into 3 categories: (1) unsafe behavior, (2)nsafe machines & tools and (3) unsafe environment as in Chi et al.2005, 2009). For unsafe behavior, our study included bodily action,oor work practices, improper use of PPE, overexertion, insuffi-ient capacities, unsafe climbing, being pulled down (pulled by

heavy object, pulled by a hoist, and pulled by a wheelbarrow),nd improper use of PPE. Unsafe machines & tools were furtherivided into unsafe ladders or tools, and mechanical failure. Unsafenvironment included lack of complying scaffold (improper scaf-olds, improper platform or improper guardrails), inappropriaterotection (unfixed floor cover, insecure warning barrier, ineffec-ive safety net, and lack of attachment point for safety harness),armful environment (poor weather, bumpy and restricted walk-

ay, and poor lighting), and other unsafe environments. These

ategories were chosen to link directly to obvious and useful pre-ention measures, as in our previous study on fatal falls in theonstruction industry (Chi et al., 2005).

Prevention 72 (2014) 359–369

Similar to our previous coding of electrocution fatalities (Chiet al., 2009), parsimony of coding was taken into consideration toreduce the number of redundant causes and minimize complexitiesin the statistical analysis. For workers that fall from a scaffold due tounsafe climbing of a scaffold, the root cause could be that employersfailed to provide an access ladder or equivalent safe access. In suchcases, we choose to code only the primary cause of unsafe climbing.Weak supporting structure was coded for the following two typesof fatalities: for workers that fell from scaffold due to the scaffoldbeing unstable to support the worker and the intended load as wellas workers fell through roof surface due to weak roofing materialor lack of secured platforms.

Another concern is the difficulty in differentiating between thefollowing three types of accident causes: bodily action, overexer-tion, and poor work practices. After putting together cases of theclosest nature, we came to the following suggestion. For bodilyaction, workers often took an unsafe posture, such as standingwith their back toward an unguarded opening, while overexertionemphasized that the material handling activity (such as lifting, car-rying, or disassembly) caused the worker’s loss of balance. Lastly,for poor work practice, workers used an incorrect procedure.

2.2. Statistical analysis

Standardized mortality ratios (SMR) for each stratified gender,age, and company size group were calculated, along with 95% con-fidence intervals (Kelsey et al., 1996), using the working populationof the Taiwanese construction industry in 2001–2005 (DirectorateGeneral of Budget and Accounting Statistics, 2001–2005) as thereference group for the calculations. Using the same reference pop-ulation can aid in making a direct comparison of the SMRs amongdifferent demographic segments.

2.3. Fault tree analysis and Boolean algebra representation

A fault tree can always be translated into an entirely equivalentMCS, which can be considered as the root causes for these fall fatali-ties (Vesely et al., 1981). Our fault tree analysis begins by identifyingmultiple cause combinations for each fatality. These multiple causecombinations can be connected by AND-gate (for 2-causes and3-causes combinations) indicating that these 2 or 3 events con-tributed simultaneously to these fatal falls. The integration of allpossible cause combinations for each falling site can be regardedas the union (connected with an OR-gate) of these cause combi-nations. Fundamental laws of Boolean algebra (see Table 1) wereapplied to reduce all possible cause combinations to the ‘smallest”cut set (Vesely et al.,1981) that could cause the top event to occur.Eventually, all cause combinations associated with each falling sitecan be simplified and presented in a fault tree diagram.

3. Results

3.1. Frequency analysis

Commutative Laws X + Y = Y + XAssociative Laws X + (Y + Z) = (X + Y) + ZDistributive Laws XY + XZ = X(Y + Z)Laws of Absorption X + XY = X

C.-F. Chi et al. / Accident Analysis and Prevention 72 (2014) 359–369 361

Table 2Frequency distribution and SMR for stratified gender, age, and company size, together with SMR groupings derived from the statistical analysis.

Factor This study Chi et al. (2005)

Freq. % SMR 95% CI SMR groupings Freq. %

GenderMale 387 94.2 1.64 (1.29–2.04) B 572 92.1Female 18 4.4 0.11 (0.03–0.29) A 49 7.9Unknown 6 1.4 – – – – –

Age

15–24 26 6.3 0.60 (0.20–1.39) A 70 11.325–34 102 24.8 0.86 (0.53–1.33) A 138 22.235–44 118 28.7 0.97 (0.62–1.46) AB 181 29.145–54 101 24.6 1.16 (0.71–1.79) AB 116 18.7≥55 57 13.9 1.64 (0.83–2.91) B 113 18.2Unknown 7 1.7 – – – 3 0.5

Company size

<5 153 37.2 1.87 (1.27–2.66) E 54 8.75–9 117 28.5 1.01 (0.64–1.51) D 60 9.710–29 97 23.6 1.48 (0.89–2.31) C 50 8.130–49 11 2.7 0.44 (0.06–1.49) BC 11 1.850–99 11 2.7 0.48 (0.06–1.66) B 7 1.1

0.25

0.24

–

caTbSigbassm(aa4ttf

TF

100–499 9 2.2

≥500 4 0.9

Unknown 9 2.2

ompanies with less than 30 workers (367, 89.3%) (see Table 2),nd had less than 1 year of working experience (308, 74.9%) (seeable 3). The SMRs of various population segments were taken toe significantly different when the 95% confidence intervals of theirMRs did not overlap (Chi and Chen, 2003). Significant differencesn SMRs were found among different gender, age, and company sizeroups. Different categories, according to the SMRs, are indicatedy alphabetical letters (see Table 2). The SMR analysis of gender,ge, and company size generated similar result as in our previoustudies. Male, workers aged 55 and above, workers employed bymaller companies or workers with less 1 year of experience areore likely to be victims of fatal falls in the construction industry

Chi et al., 2005). Additionally, a rising trend was found between agend SMR (p < 0.05). In the United States, fatal work injury rates, byge group, in 2012 indicated that fatal work injury rates for workers

5 years of age and over were higher than the overall U.S. rate andhe rate for workers 65 years of age and over was around 3 timeshe rate for all workers (U.S. Bureau of Labor Statistics, 2014). Fatalalls of older workers (aged 55 and above) could have been caused

able 3requency distributions of worker experience, falling height and falling sites.

Factors

Experience

Under 1 year

1–15 years

Over 15 years

Unknown

Falling height

0–2.0 m

2.1–3.0 m

3.1–10.0 m

10.1–30.0 m

>30 m

Unknown

Falling sites

Falls from scaffolds

Falls from building girders or steel

Falls through existing floor opening

Stepping through roof surface

Falls from ladders

Falls from roof edge or opening

Falls from dock or floor edge

Falls from scaffolds with wheels

Falls from deck or balcony

Others (falls from construction vehicle,falls down stairs, collapsing material orstructure)

(0.02–0.97) AB 1 0.2(0.00–1.57) A 1 0.2– – 437 70.4

by reduced sensory capability (e.g., decline of vision and hearing),and reduced physical strength and flexibility (Chi and Wu, 1997).

The working height is one of the most hazardous aspects of aconstruction site (Hu et al., 2011). Occupational Safety and HealthAdministration (OSHA) requires applying a fall prevention pro-gram for any construction project with a height above 1.83 m (6 ft)(Occupational Safety and Health Administration, 1996). In Taiwan,fall protection measures are required for performing any task above2 m. However, in our database, there were 24 cases (5.8%) of fallsfrom less than 2 m (see Table 3).

In terms of the falling site, out of the 411 cases analyzed,23.8% (98 cases) of the fatalities were classified as falls from scaf-folds, from a building girder or other steel structure (78 cases,19.0%), through an existing floor opening (56 cases, 13.6%), step-ping through roof surface (49 cases, 11.9%), from a ladder (37 cases,

9.0%), from the roof edge or opening (29 cases, 7.1%), from the dockor floor edge (26 cases, 6.3%), from a scaffold with wheels (12 cases,2.9%), from a deck or balcony (9 cases, 2.2%) and other falling sites(17 cases, 4.1%) (see Table 3).

This study Chi et al. (2005)

Freq. % Freq. %

308 74.9 500 80.579 19.2 106 17.0

8 1.9 9 1.416 4 6 1.0

24 5.8 – –24 5.8 – –

200 48.7 – –119 29.0 – –

30 7.3 – –14 3.4 – –

98 23.8 189 30.478 19.0 70 11.356 13.6 128 20.649 11.9 44 7.137 9.0 26 4.229 7.1 67 10.826 6.3 51 8.212 2.9 - -

9 2.2 - -17 4.1 13 2.1

362 C.-F. Chi et al. / Accident Analysis and

Table 4Frequency distribution of accident causes.

Accident cause This study Chi et al. (2005)

Freq. % Freq. %

Unsafe behavior 315 47.9Bodily action 38 5.8 62 10.0Poor work practices 26 4.0 44 7.1Overexertion 11 1.7 16 2.6Insufficient capacities 2 0.3 11 1.8Unauthorized access 0 0 5 0.8Unsafe climbing 28 4.3 – –Being pulled down 20 3.0

Pulled by heavy object 8 1.2 24 3.9Pulled by hoist 7 1.1 9 1.4Pulled by wheelbarrow 5 0.8 6 1.0

Improper use of PPE 190 28.9 35 5.7

Unsafe machines & tools 36 5.5Unsafe ladders or tools 26 4.0 7 1.1Mechanical failure 10 1.5 3 0.5

Unsafe environment 215 32.7Unguarded opening 60 9.1 104 16.7Weak roofing material 59 9.0 - -Removal of protection measure 16 2.4 15 2.4Unfixed floor cover 1 0.2 15 2.4Improper platform 41 6.2 58 9.3Improper guardrails 18 2.7 113 17.2Improper scaffolds 20 3.0 – –

Harmful environment 92 14.0Poor weather 8 1.2 13 2.1Bumpy and restricted walkway 14 2.1 7 1.1Poor lighting 4 0.6 3 0.5Hit by falling objects 2 0.3 22 3.5Other unsafe environments 0 0 6 1.0Lack of attachment for safety harness 35 5.3 1 0.2

gstacb9fmwco

TC

Ineffective safety net 29 4.4 2 0.3

Unknown 0 0 40 6.4

Table 4 shows the frequency distribution of accident causesiven that multiple causes were identified for each fatality. Table 5hows all cause combination (single, 2-causes, and 3-causes) forhe 411 cases in terms of unsafe behavior, unsafe machines & tools,nd unsafe environment. There were 179 single cause cases, 217ases of two-cause combinations, and 15 cases of three-cause com-inations. Overall, these 411 fatality cases can be described by1 distinctive combinations of accident causes and 90.8% of theseatalities were associated with unsafe behavior and unsafe environ-

ent. About 48% of fatalities were caused by unsafe behavior, forhich improper use of PPE (28.9%), bodily action (5.8%), and unsafe

limbing (4.3%) were the most common causes. An unguardedpening and weak roofing (supporting) structure were the most

able 5ause combination of falling fatality.

Accident cause (N = 411) Freq.

Single cause 179Unsafe environment 83Unsafe behavior 73Unsafe machines & tools 232-causes 217Unsafe behavior•Unsafe environment 120Unsafe behavior•Unsafe behavior 51Unsafe environment•Unsafe environment 32Unsafe machines & tools•Unsafe environment 12Unsafe behavior•Unsafe machines & tools 23-causes 15Unsafe behavior•Unsafe behavior•Unsafe environment 7Unsafe behavior•Unsafe environment•Unsafe environment 4Unsafe environment•Unsafe environment•Unsafe environment 2Unsafe behavior•Unsafe behavior•Unsafe behavior 1Unsafe behavior•Unsafe machines & tools•Unsafe environment 1

Prevention 72 (2014) 359–369

frequent unsafe environmental situations, each contributed to 9.1%and 9% of the falls, respectively. On the other hand, an unsafe ladder(4%) was the most common unsafe tools.

3.2. Falling height and posture

As stated by Hammer (1989), falls do not have to be of agreat height to be fatal. A head injury is the most serious andfrequent trauma from a fall. However, most occupational safetyand health regulations protected workers at heights greater than2 m (in Taiwan and in Australia). The United States has a simi-lar Safety and Health Standards for the construction industry. TheOccupational Safety and Health Administration (1996) stated that“each employee on a walking/working surface with an unprotectedside or edge which is 6 feet (1.8 m) or more above a lower level shallbe protected from falling by the use of guardrail systems, safetynet systems, or personal fall arrest systems”. According to the U.S.Bureau of Labor Statistics (2014) falls to a lower level accountedfor 570 fatal work injuries in 2012. Ten percent of falls to a lowerlevel involved falls of 6 feet or less. Our statistics indicated that24 victims (about 5.8% of all 411 fatalities) (see Table 3) fell fromheights of less than 2 meters, where workers may ignore the risk offalling. In these cases, most of these victims did not wear a helmetand fell backwards hitting the back of their head. More than half(14 of 24 cases) of these cases were falls from ladders. The otherfalls from less than 2 meters include falls from scaffolds (3 cases),from a building girder or steel (2 cases), through a floor opening (3cases), and from scaffold with wheels (2 cases).

3.3. Boolean algebra simplification

Take the 35 distinctive cause combinations associated with the98 falls from scaffold cases as an example (see Table 6). Out of these98 cases, 24 cases were caused by 8 single causes, with improperuse of PPE (8 cases) as the most frequent single cause. Seventycases were caused by 23 two-cause combinations with improperscaffolds and improper use of PPE (16 cases) as the most frequentcombination. Lastly, four falling from scaffold cases were causedby a three-cause combination. According to the fault tree analysis,two-cause and three-cause combinations would be connected bythe AND-gate, as shown at the bottom layer of Fig. 1. The unionof the 35 distinctive cause combinations can be connected with anOR-gate. In terms of Boolean algebra, accident causes associatedwith falling from scaffold can be presented as follows. Abbrevia-tions of these Boolean representations are summarized in Table 7.

8PPE + 4Practice + 3Attach + 3Climb + 2Scaffold + 2Tool

+ Act + Remove + 16(PPE • Scaffold) + 11(PPE • Guard)

+ 10(PPE • Platform) + 5(PPE • Climb) + 4(PPE • Practice)

+ 4(Platform • Net) + 2(Scaffold • Net) + 2(PPE • Structure)

+ 2(Attach • Scaffold) + PPE • Act + PPE • Object

+ PPE • Mechanical + PPE • Remove + PPE • Tool

+ Platform • Bumpy + Platform • Climb + Scaffold • Act

+ Attach • Net + Guard • Climb + Tool • Pull + Tool • Climb

+ Exert • Guard + Structure • Net + Pull • Platform • PPE

+ Act • Scaffold • PPE + Guard • Bumpy • PPE

+ Climb • Scaffold • PPE (1)

Subsequently, the absorption law in the MCS analysis wasapplied to reduce the cause combinations by absorbing a secondary

C.-F. Chi et al. / Accident Analysis and Prevention 72 (2014) 359–369 363

Fig. 1. The fault tree illustration for falling from scaffold before the minimum cut set analysis.

364 C.-F. Chi et al. / Accident Analysis and

Table 6Cause combinations of falling from scaffold.

Cause combinations Freq. Sub total

Improper Use of PPE 8

24

Poor Work Practices 4Lack of Attachment Point for Safety Harness 3Unsafe Climbing 3Improper Scaffolds 2Unsafe Tools 2Bodily Action 1Removal of Protection Measure 1Improper Use of PPE•Improper Scaffolds 16

70

Improper Use of PPE•Improper Guardrails 11Improper Use of PPE•Improper Platform 10Improper Use of PPE•Unsafe Climbing 5Improper Use of PPE•Poor Work Practices 4Improper Platform•Ineffective Safety Net 4Improper Scaffolds•Ineffective Safety Net 2Improper Use of PPE•Weak Supporting

Structure2

Lack of Attachment Point for SafetyHarness•Improper Scaffolds

2

Improper Use of PPE•Bodily Action 1Improper Use of PPE•Hit by Falling Objects 1Improper Use of PPE•Mechanical Failure 1Improper Use of PPE•Removal of Protection

Measure1

Improper Use of PPE•Unsafe Tools 1Improper Platform•Bumpy and Restricted

Walkway1

Improper Platform•Unsafe Climbing 1Improper Scaffolds•Bodily Action 1Lack of Attachment Point for Safety

Harness•Ineffective Safety Net1

Improper Guardrails•Unsafe Climbing 1Unsafe Tools•Being Pulled Down 1Unsafe Tools•Unsafe Climbing 1Overexertion•Improper Guardrails 1Weak supporting Structure•Ineffective Safety

Net1

Being Pulled Down•ImproperPlatform•Improper Use of PPE

1

4Bodily Action•Improper Scaffolds•ImproperUse of PPE

1

Improper Guardrails•Bumpy & RestrictedWalkway•Improper Use of PPE

1

Unsafe Climbing•Improper Scaffolds•Improper 1

cfaccr

TA

use of PPE

Total 98 98

ause with a primary cause. For example, 11 falling from scaffoldatalities were caused by the combination of improper use of PPEnd improper guardrails. Because improper use of PPE itself can

ause falling from a scaffold fatalities, while improper guardrailsannot, improper use of PPE and improper guardrails would beegarded as primary and secondary, respectively. Improper use of

able 7bbreviations of Boolean representation for accident cause.

Act: bodily action Exert: overexertionAttach: lack of attachment for safety

harnessPlatform: improper platform

Capacity: insufficient capacities PPE: improper use of PPEClimb: unsafe climbing Remove: removal of protection

measureCover: unfixed floor cover Practice: poor work practicesGuard: improper guardrails Pull: being pulled downLight: poor lighting Structure: weak roofing material or

supporting structureMechanical: mechanical failure Scaffold: improper scaffoldsNet: ineffective safety net Tool or Ladder: unsafe ladders or toolsObjects: hit by falling objects Bumpy: bumpy and restricted

walkwayOpen: unguarded opening Weather: poor weather

Prevention 72 (2014) 359–369

PPE would absorb improper guardrails and become the only acci-dent cause presented in the fault tree after the MCS analysis. Forother two-cause or three-cause combinations where each acci-dent cause alone can contribute to a fatality, such causes wouldbe regarded as primary. Each such primary cause would absorb thetotal number of cases divided by the number of primary causesof the same combination. For example, five falling from scaffoldfatalities were caused by the combination of improper use of PPEand unsafe climbing. Given that improper use of PPE and unsafeclimbing each can cause falling from scaffold fatalities, the currentstudy assumed that improper use of PPE and unsafe climbing eachabsorbed 2.5 cases out of these 5 two-cause fatalities. For caseswhere none of the causes in a multiple cause combination hadcaused the fatality by itself, the absorption law would not be appliedand the multiple cause combination will remain in its original for-mat. One last possibility for the three-cause combination is that twocauses are primary and one is secondary. In this case, the numberof cases associated with the secondary cause would be absorbedequally by the other two primary causes. For example, falling froma building girder was caused by the combination of bodily action(primary), mechanical failure (primary), and improper use of PPE(secondary); the improper use of PPE would be absorbed equallyby the other two primary causes and give 0.5 bodily action + 0.5mechanical failure. The above consideration allows us to keep trackof the fatality frequency after applying the MCS analysis.

Eventually, all cause combinations associated with fall fromscaffold can be simplified as follows:

49.67PPE + 14.17Scaffold + 8.33Climb + 6Practice + 5Attach

+ 4Tool + 2.33Act + 1.5Remove + 4Platform • Net

+ Platform • Bumpy + Exert • Guard + Structure • Net (2)

The original fault tree of 35 distinctive accident causes associ-ated with falling from a scaffold (see Fig. 1) can be reduced to alogically equivalent fault tree consisting of 12 MCSs connected byan OR-gate beneath the top event (see Fig. 2). Each MCS can be abasic event (denoted by a circular shape) or an intermediate event(denoted by a rectangular shape) consisting of multiple basic eventsconnected by an AND-gate that causes the top event to occur. Sim-ilarly, the same MCS analysis can be applied to other falling sitesto derive logically equivalent fault trees for further development ofprevention measures (see Table 8). All of the MCSs are necessaryand sufficient to cause the fall from a scaffold fatalities.

4. Discussion

As described earlier, fault tree models can be analyzedqualitatively or quantitatively. Qualitative analysis leads to thedetermination of the MCSs contributing to a fatality. For quan-titative analysis, Khakzad et al. (2011) suggested to present theprobability of the top event in terms of the occurrence probabilityof MCS events. However, one problem in this respect is the denom-inator; the population or exposure data, such as the total number ofevents in which a certain fraction led to a fault, are rarely available(Ale et al., 2006). For example, our previous study (Chi and Chen,2003) on occupational fatalities focused only on the controllabledemographic factors (e.g., age, gender and industry). Other factorssuch as working experience, company size, and source of injurycannot be analyzed because the working population with a similarbackground was not available.

Although the probability of a top event in terms of probabilityof MCS events cannot be derived without the exposure data as thedenominator, the frequency of the MCSs can still be prioritized toindicate the importance of the cause(s) and be used to customize

C.-F. Chi et al. / Accident Analysis and Prevention 72 (2014) 359–369 365

from

tiybtf

sfapti

TO

Fig. 2. The fault tree illustration for falling

he construction site inspection to improve the utility of humannspectors (Chi and Chen, 2003). Through fault tree and MCS anal-ses on 411 fatal falls, MCSs associated with each falling site cane presented as a fault tree as shown in Fig. 3 to guide the inspec-ion agency or the construction company to conduct safety auditsocusing on these basic causes.

Based on focus group discussion to investigate why accidentstill happen in the construction industry, Haslam, et al. (2005)ound that construction supervisors frequently have little safety

wareness and a poor understanding of accident causation andrevention. Similarly, Hung et al. (2013) explored fall-protectionraining needs, through 29 semi-structured interviews among res-dential roofing subcontractors, with respect to recommendations

able 8riginal cause combinations and simplified result by applying Boolean algebra.

Falling site (Freq) Accident causes combinations

Scaffolds (98) 8PPE + 4Practice + 3Attach + 3Climb + 2Scaffold + 2To16(PPE•Scaffold) + 11(PPE•Guard) + 10(PPE•Platform4(PPE•Practice) + 4(Platform•Net) + 2(Scaffold•Net) +2(Attach•Scaffold) + PPE•Act + PPE•Object + PPE•MecPPE•Tool + Platform•Bumpy + Platform•Climb + ScaffGuard•Climb + Tool•Pull + Tool•Climb + Exert•GuardPull•Platform•PPE + Act•Scaffold•PPE + Guard•Bump

Building girders (78) 10PPE + 4Attach + 3Climb + 2Scaffold + Act + Open +

5(PPE•Net) + 4(Structure•PPE) + 4(Scaffold•PPE) + 3(2(Open•Attach) + 2(PPE•Weather) + 2(Remove•PPE)

Act•Open + Attach•Net + Bumpy•PPE + Climb•WeathMechanical•PPE + Practice•PPE + Object•PPE + Practic+ Pull•Net + Pull•Guard + Scaffold•Attach + StructureAct•Platform•Weather + Climb•Attach•Net + Climb•P

Floor opening (56) 11Open + 6Practice + 5Remove + 2Attach + PPE + Stru4(Pull•PPE) + 3(Act•Open) + 2(Act•PPE) + 2(MechanicExert•Attach + Exert•Guard + Open•Exert + Open•LigRemove•PPE + Structure•Attach + Exert•Pull•Attach

Roof surface (49) 40Structure + Open + 5Structure•PPE + Act•StructureLadders (37) 16Ladder + 5PPE + 3Act + Bumpy + Capacity + Climb +

2(Ladder•PPE) + Act•Ladder + Open•Attach + Open•P

Roof edge or opening (29) 5PPE + 3Attach + Open + Pull + 3(Open•PPE) + 2(Act•Exert•Open + Exert•PPE + Open•Attach + Open•BumpPPE•Net + Pull•PPE + Scaffold•PPE + Weather•AttachPractice•Pull•PPE

Floor edge (26) 4PPE + 2Open + Bumpy + Scaffold + 3(Open•PPE) + 2(+ Act•Pull + Exert•PPE + Open•Attach + Open•WeathPull•PPE + Remove•PPE + Remove•Weather + Scaffol

Scaffolds with wheels (12) 3Mechanical + 2Act + Act•Light + Act•PPE + Bumpy•GClimb•Net + Practice•Pull + Act•Remove•Light

Deck or balcony (9) 3Practice + Open + Act•PPE + Practice•Attach + PractiRemove•PPE

Others (17) 2Act + 2Bumpy + 2Open + Mechanical + PPE + 2ScaffoClimb•Structure + Mechanical•PPE + Open•Weather

Remove•Bumpy

scaffold in terms of the minimum cut sets.

for the design of fall-protection training. In general, it was foundthat pictures work better than words especially in communicat-ing with foreign (Hispanic) workers. Including as many pictures aspossible on safety materials can increase their effectiveness andworkers’ engagement (Hung et al., 2013).

Fault trees can be very effective in creating an overview of acomplex occurrence and in communicating our research findingsof potential accident causes and scenarios to construction workers(Svedung and Rasmussen, 2002). To improve effectiveness in com-

munication of accident causation, pictures depicting features of thereal-world object or concept they represent (Dewar, 2004) (graph-ical icons) were incorporated into our fault tree analysis. Graphicalicons were designed, for each falling site scenario and associated

Minimal cut set

ol + Act + Remove +) + 5(PPE•Climb) +

2(PPE•Structure) +hanical + PPE•Remove +old•Act + Attach•Net +

+ Structure•Net +y•PPE + Climb•Scaffold•PPE

49.67PPE + 14.17Scaffod + 8.33Climb +6Practice + 5Attach + 4Tool + 2.33Act +1.5 Remove + 4Platform•Net +Platform•Bumpy + Exert•Guard +Structure•Net

Pull + Structure + 6(Climb•PPE) +Act•PPE) + 3(Open•PPE) ++ 2(PPE•Scaffold) + Act•Bumpy +er + Cover•PPE + Guardrail•PPE +e•Attach + Pull•PPE + Pull•Attach•Net + Act•Mechanical•PPE +PE•Net + Structure•Attach•Net

37.5PPE + 9Attach + 8Climb + 5.5Act +5.5Scaffold + 4.5Structure + 4Open +4Pull

cture + Tool + 7(Open•PPE) +al•PPE) + Act•Guard +ht + PPE•Net + Remove•Exert ++ Pull•Open•Attach

20Open + 14PPE + 6.5Remove +6Practice + 5Attach + 1.5Structure +Tool + Guard•Act + Guard•Exert

+ Open•Net + Open•PPE 46Structure + 3Open Exert + Scaffold +2(Act•PPE) +

PE + Practice•Ladder18.5Ladder + 8PPE + 4.5Act + Bumpy +Capacity + Climb + Exert + Scaffold +Open•Attach

PPE) + Attach•Net + Bumpy•PPE +y + Open•Net + Open•Weather +

+ Open•Bumpy•Attach +

13.5PPE + 7.5Open + 6Attach + 2Pull

Open•Net) + Act•Open + Act•PPEer + PPE•Net + Practice•PPE +d•PPE + Exert•Guard•PPE

13PPE + 8.5Open + 1.5Scaffold +Bumpy + Act•Pull + Remove•Weather

uard + Climb•Bumpy + 5Act + 3Mechanical + Bumpy•Guard +Practice•Pull + Climb•Bumpy +Climb•Net

ce•PPE + Practice•Scaffold + 6Practice + Open + PPE•Act +PPE•Remove

ld•PPE + Capacity•PPE ++ PPE•Net + Pull•PPE +

6.5PPE + 3Bumpy + 3Open + 2Act +1.5Mechanical + Climb•Structure

366 C.-F. Chi et al. / Accident Analysis and Prevention 72 (2014) 359–369

Fig. 3. A complete fault tree of fatal falls for all falling sites.

C.-F. Chi et al. / Accident Analysis and Prevention 72 (2014) 359–369 367

atwwa

ifbefi

TE

Fig. 4. Graphical icon designed for accident cause “being pulled down”.

ccident cause, to illustrate the fault tree in a graphical manner sohat accident causation can be easily understood by constructionorkers who may not read English (or Chinese). Similar figuresere seen in Drury and Brill (1983) to describe commonly seen

ccident patterns related to consumer product usage.As an example, Fig. 4 shows the icon designed for represent-

ng the accident cause “being pulled down”. Based on the concept

or image-related icons (Chi and Dewi, 2014), the icon was createdased on the accident scenario found in the database that work-rs could be pulled down while he was dumping a wheelbarrowlled with heavy materials. Other considerations include that all

Fig. 5. Graphical fault tree for fatal fall from roof surface.able 9xamples of similar fatal fall scenarios in the United States during 2010–2013.

Date and Falling site Event description

Date: Jul 24, 2012Falling site: scaffoldSource: OSHA

Victim was working from a job-built wooden scaffold constructed on the fourth floor exterior balconywalkway. A wooden guardrail system was in place along the length of the 4th floor balcony and along theexterior side of the elevated scaffold platform. The victim unsecured the end of the top guardrail, nearest thework being performed, and let it drop down onto the scaffold while the opposite end of the guardrail was stillattached. This created an unprotected open area along the exterior side of the scaffold. He was on one knee,along the unprotected edge of the scaffold, and proceeded to strike the block with his hammer. On the finalstrike, the block came free. The victim’s momentum and force while swinging the hammer caused him topitch forward through the open unprotected space. He fell about 36–40 feet to the asphalt below.

Date: Mar 27, 2011Falling site: scaffoldSource: NIOSH

A male municipal custodian was fatally injured while painting an interior section of a school lobby. Althoughthe incident was un-witnessed, it appears that the victim was on a mobile scaffolding when he fell from thescaffolding landing on the lobby floor.

Date: Jan 4, 2013Falling site: tower (building girders)Source: OSHA

A communications worker was on a cellular communications tower, at a position more than 25 meters abovethe ground, when he fell. Although he was using a PPE, the anchorage he was using failed.

Date: May 21, 2012Falling site: walk board (floor opening)Source: OSHA

An employee was engaged in residential framing work from a walk board eight feet above the ground whilerelocating the end truss that had been placed in the wrong location. The employee fell from the walk board tothe concrete surface below, suffering a concussion.

Date: Sep 25, 2012Falling site: roofSource: OSHA

The owner of a roofing and repair company was working on the roof of a three-story hotel. The owner, whileinstalling a tarp, stepped close to the roof’s eave. The eave collapsed, and he fell 26 feet to the parking lot.

Date: Oct 10, 2012Falling site: ladderSource: NIOSH

The three-man crew of a construction business was removing shingles to re-roof a private recreation facility.The victim carried a tarp, up an extension ladder, to the roof so the owner could cover the roof in anticipationof impending rain. He fell an estimated 10 to 12 feet to the foundation retaining wall. A stabilizer bar attachedto the ladder had come apart from the ladder and a nylon tie-down strap used to secure the stabilizer bar tothe roof had broken when the ladder slid sideways.

Date: Oct 11, 2012Falling site: decking (roof edge & opening)Source: OSHA

The victim and a coworker were covering an open-air atrium with a 24 ft by 45 ft rolled poly plastic sheeting.They were told to install the rolled poly plastic over the existing plastic from the 2nd floor and roof deck. Theypulled the plastic over the open atrium, when the victim fell. The victim was wearing a full body harness withretractable lanyard device, his fall was not arrested, due to the failure of the anchors, and fell 20 ft from theroof atrium opening onto the floor.

Date: Sep, 2010Falling site: highway bridge (Floor edge)Source: NIOSH

A worker fell from a highway bridge in the process of leveling jacks for concrete forms. He had a fall protectionharness, with an attached lanyard, then stepped outside the lifeline onto a 10′′ × 2′′ wooden board to verify thejacks were level. As he stepped over the lifeline, he did not attach the lanyard and fell approximately 28 feetonto the railroad tracks below.

Date: Sep 22, 2012Falling site: Scissor Lift (Scaffold with wheels)Source: OSHA

The victim was using an upright scissor lift on the forks of a forklift to access light fixtures. The scissor lift wasextended to a height of 22.33 ft. As he went to change a bulb, the scissor lift tilted and fell off the forks. Thevictim fell to the asphalt parking lot.

Date: Oct 18, 2012Falling site: unguarded opening (deck/balcony)Source: OSHA

A male carpenter was working on the overhead archway on a balcony of a wood framed residential house,when he fell more than 14 ft. to the ground from an unguarded second story opening. The victim was notwearing or using any type of fall protection at the time of the fall.

368 C.-F. Chi et al. / Accident Analysis and Prevention 72 (2014) 359–369

fatal fa

gcb2rrhifdas

rfnsiyeocctHfhi

Fig. 6. Graphical fault tree for

raphical icons should have borders to make icons appear orderly,onsistent, and uniform (Campbell et al., 2004). Also, icons shoulde legible and comprehensible by the viewers (meaningful) (Dewar,004). Following the standard format of fault tree analysis, iconsepresenting different falling site scenarios are designed to haveectangular borders, while icons indicating accident causes wouldave circular borders. As examples, graphical fault trees, as shown

n Figs. 5 and 6, were developed for falling through roof surface andalling from a roof edge or opening, to support inter-disciplinaryiscussion of risk management (Svedung and Rasmussen, 2002)nd facilitate the communication of accident causation to first lineupervisors (Hung et al., 2013).

This research applied a fault tree analysis to represent the causalelationships among events and causes that contributed to fatalalls in the construction industry. The causes for the falls and theature of the accident events were classified and coded with theame coding scheme as in Chi et al. (2005). A limitation of the studys that the fatalities occurred during 2001–2005, which is severalears ago. However, the reuse of the coding scheme proved thatven though construction methods and practices may have devel-ped during the intervening period, the essence of the accidentausation remains the same. As shown in Table 9, based on accidentases published by the Occupational Safety & Health Administra-ion (OSHA) and the National Institute for Occupational Safety and

ealth (NIOSH), the scenario for fatal falls associated with each

alling site in the U.S. construction industry during 2010 to 2013ave a similar nature. The major contribution of the current study

s to develop and demonstrate the proposed graphical fault tree

lls from roof edge or opening.

approach with the fall fatality data. This approach consists of a com-bination deriving from mutually exclusive accident causes for eachfatality, reducing complexity by Boolean algebra, and presentingthe fault tree analysis in a graphical manner using meaningful iconsto facilitate the inter-disciplinary communication of accident cau-sation. The classification of accident causes into major and mutuallyexclusive categories, as in our previous classification of assistivedevice (Chi et al., 2012), allows us to classify the majority of thefatal fall accidents in Taiwan and presented them as a comprehen-sive and systematic graphical fault tree. Such graphical fault treescan serve as a taxonomy framework for the analysis of fatal falls at amacro-level where each accident cause category can be revised andexpanded based on the particular needs of a construction company(Chi et al., 2012).

5. Conclusion

The current study analyzed 411 work-related fatal falls fromthe years 2001 to 2005 and coded up to three accident causes foreach of them. Boolean algebra was applied for these 411 fall fatal-ities to derive MCSs associated with each falling site. Eventually,all 411 accident causes can be presented in a fault tree to providean overview of all accident causes for fall fatalities included in thisstudy. Also, the relative frequency of the MCSs can be prioritized

to customize the construction site inspection to improve the util-ity of human inspectors. Furthermore, icons representing differentfalling site scenarios and accident causes were designed to illus-trate the fault tree analysis to support inter-disciplinary discussion

is and

oo

R

A

C

C

C

C

C

C

C

D

D

D

H

C.-F. Chi et al. / Accident Analys

f risk management and improve effectiveness in communicationf accident causation to first line supervisors.

eferences

le, B.J.M., Bellamy, L.J., Cooke, R.M., Goossens, L.H.J., Hale, A.R., Roelen, A.L.C., Smith,E., 2006. Towards a causal model for air transport safety – an ongoing researchproject. Safety Sci. 44, 657–673.

ampbell, J.L., Richman, J.B., Carney, C., Lee, J.D., 2004. In-Vehicle Display Icons andOther Information Elements. Volume I: Guidelines (No. FHWA-RD-03-065).

hi, C.F., Wu, M.L., 1997. Fatal occupational injuries in Taiwan – relationship betweenfatality rate and age. Safety Sci. 27, 1–17.

hi, C.F., Chen, C.L., 2003. Reanalyzing occupational fatality injuries in Taiwan witha model free approach. Safety Sci. 41, 681–700.

hi, C.F., Chang, T.C., Ting, H.I., 2005. Accident patterns and prevention meas-ures for fatal occupational falls in the construction industry. Appl. Ergon. 36,391–400.

hi, C.F., Yang, C.-C., Chen, Z.-L., 2009. In-depth accident analysis of electrical fatali-ties in the construction industry. Int. J. Ind. Ergon. (39), 635–644.

hi, C.F., Tseng, L.K., Jang, Y., 2012. Prune a decision tree of selecting computer-related assistive devices for the disable user. IEEE Trans. Neural SystemRehabilitation Eng. 20 (4), 564–573.

hi, C.F., Dewi, R.S., 2014. Matching performance of vehicle icons in graphical andtextual formats. Appl. Ergon. 45, 904–916.

ewar, R., 2004. In: Zwaga, H., Boersema, T., Hoonhout, H. (Eds.), Design and eval-uation of public information symbols in Visual information for everyday use:design and research perspectives. CRC Press.

oytchev, D.E., Szwillus, G., 2009. Combining task analysis and fault tree analysisfor accident and incident analysis: a case study from Bulgaria. Accident Anal.

Prevent. 41 (6), 1172–1179.

rury, C.G., Brill, M., 1983. Human factors in consumer product accident investiga-tion. Hum. Factors 25 (3), 329–342.

adipriono, F.C., 1992. Expert system for construction safety. I: Fault-tree models.J. Perform. Constructed Facilities 6 (4), 246–260.

Prevention 72 (2014) 359–369 369

Hammer, W., 1989. Occupational Safety Management and Engineering, 4th edition.Prentice Hall, NJ, pp. 563–564.

Harms-Ringdahl, L., 2001. Safety Analysis: Principles and practice in occupationalSafety. Taylor & Francis, pp. 121–148.

Haslam, R.A., Hidea, S.A., Gibbb, A.G.F., Gyia, D.E., Pavitt, T., Atkinson, S., Duff, A.R.,2005. Contributing factors in construction accidents. Appl. Ergon. 36, 401–415.

Hu, K., Rahmandad, H., Smith-Jackson, T., Winchester, W., 2011. Factors influencingthe risk of falls in the construction industry: a review of the evidence. Construc-tion Manage. Econ. 29, 397–416.

Hung, Y.H., Winchester, W.W., Smith-Jackson, T.L., Kleiner, B.M., Babski-Reeves, K.L.,Mills, T.H., 2013. Identifying fall-protection training needs for residential roofingsubcontractors. Appl. Ergon. 44, 372–380.

Kelsey, J.L., Whittemore, A.S., Evans, A.S., Thompson, W.D., 1996. Methods in Obser-vational Epidemiology. Oxford University, New York.

Khakzad, N., Khan, F., Amyotte, P., 2011. Safety analysis in process facilities: compar-ison of fault tree and Bayesian network approaches. Reliability Eng. Syst. Safety96 (8), 925–932.

LeBeau, K.H., Wadia-Fascetti, S.J., 2007. Fault tree analysis of Schoharie Creek Bridgecollapse. J. Perform. Constructed Facilities 21 (4), 320–326.

Occupational Safety, Health Administration, 1996. Safety and Health Standards forthe Construction Industry. OHSA 3129.

Sorock, G.S., Smith, E.O., Goldoft, M., 1993. Fatal occupational injuries in the NewJersey construction industry, 1983–1989. J. Occupational Med. 35, 916–921.

Svedung, I., Rasmussen, J., 2002. Graphic representation of accident scenarios: map-ping system structure and the causation of accidents. Safety Sci. 40, 397–417.

U.S. Bureau Labor Statistics (2014). Census of Fatal Occupational Injuries Charts,1992–2012 (revised data). Retrieved July 12, 2014 from http://www.bls.gov/iif/oshcfoi1.htm#charts

Vesely, W.E., Goldberg, F.F., Roberts, N.H., Haasl, D.F., 1981. Fault Tree Handbook.U.S. Government Printing Office, Washington.

Volkanovski, A., Cepin, M., Mavko, B., 2009. Application of the fault tree analysisfor assessment of power system reliability. Reliability Eng. Syst. Safety 94 (6),1116–1127.

Whitesitt, J.E., 1995. Boolean Algebra and its Applications. Courier Dover Publica-tions.