Preventing Workplace Incidents in Construction

Data Mining and Analytics Applications

Imriyas Kamardeen

• Publisher: Routledge
• Year: 2019
• ISBN:
  9781351616133
• Edition: 1
• Topic: Architektur
• Subtopic: Wohnarchitektur

The construction industry is vital to any national economy; it is also one of the industries most susceptible to workplace incidents. The unacceptably high rates of incidents in construction have huge socio-economic consequences for the victims, their families and friends, co-workers, employers and society at large. Construction safety researchers have introduced numerous strategies, models and tools through scientific inquiries involving primary data collection and analyses. While these efforts are commendable, there is a huge potential to create new knowledge and predictive models to improve construction safety by utilising already existing data about workplace incidents. In this new book, Imriyas Kamardeen argues that more sophisticated approaches need to be deployed to enable improved analyses of incident data sets and the extraction of more valuable insights, patterns and knowledge to prevent work injuries and illnesses.

The book aims to apply data mining and analytic techniques to past workplace incident data to discover patterns that facilitate the development of innovative models and strategies, thereby improving work health, safety and well-being in construction, and curtailing the high rate of incidents. It is essential reading for researchers and professionals in construction, health and safety and anyone interested in data analytics.

1 Introduction

Dark side of construction

The construction industry is a vital sector in the national economy of any country; it creates the physical infrastructure essential for the functioning of the nation, provides jobs for a large number of people and contributes significantly to GDP. At the same time, it is one of the most vulnerable sectors for workplace incidents globally (Cigularov et al. 2010). In the Australian context, for example, Safe Work Australia (2015a) reported that 417 construction workers were killed due to work injuries over the 12-year period from 2002–2003 to 2013–2014 (2.24 fatalities per 100,000 workers). In 2013–2014, the construction industry accounted for 12 per cent of all work-related fatalities when it employed only 9 per cent of the national workforce. Moreover, the construction industry accepted an average of 12,600 workers’ compensation claims per year for injuries and diseases involving one or more weeks off work over the five years from 2008–2009 to 2012–2013, equating to an average of 35 serious claims each day. Likewise, the UK construction industry recorded 35 fatalities in 2014–2015 (1.62 per 100,000 workers), the second highest among the main industry sectors (HSE 2015). The number of fatalities in the US construction industry in 2014 was 874 (9.5 per 100,000 workers), which is the highest among all industry sectors (gBLS 2015). Construction fatality statistics in developing countries are much worse than these. For instance, Kheni et al. (2010) claimed that Ghana, a Sub-Saharan African country, recorded a fatality rate of 77.6 per 100,000 workers. In Turkey, official statistics for 2011 revealed that the construction sector accounted for 6.3 per cent of the labour force but 33.5 per cent (570 of 1700) of total fatalities for all industries (Gürcanli and Müngen 2013). Yoon et al. (2013) summarised that the Korean construction industry accounts for a fatality rate range of 18.9–39.7 per 100,000 workers and for the Taiwanese construction industry it was 13 fatalities (Cheng et al. 2012). The undesirable situation around the globe was summarised by the International Labour Organization that the construction industry employs 7 per cent of the global workforce but accounts for an excessively disproportionate 30–40 per cent of work fatalities; 2,100,000 workers are killed on construction sites every year (one worker every five minutes) (Murie 2007).

Table 1.1 Socio-economic consequences of construction incidents

Source: adapted from European Union (2011, p. 8).

The unacceptably high rates of incidents in construction have huge socio-economic consequences for the victims, their families and friends, co-workers, employers and society in general. Table 1.1 illustrates the costs borne by these different groups. Several studies have been conducted to quantify these socio-economic costs of construction incidents in monetary terms. For instance, Safe Work Australia (2015b) estimated that the total cost of construction incidents in Australia during 2012–2013 was AU$5.84 billion, which is approximately 0.4 per cent of the GDP of the nation for that period, with an average unit cost of AU$117,180 per incident. It further reported generally for all industries for the period, that in terms of distribution, 77 per cent of the total cost was borne by workers, 18 per cent by the community and 5 per cent by employers. Likewise, in the UK, the total cost of construction injury and illness in 2013–2014 was estimated to be £0.9 billion. Waehrer et al. (2007) reported that construction incidents in the US in 2002 cost US$11.5 billion; the average cost of a construction fatality was US$4 million whilst a non-fatal injury was US$42,000. These statistics prove that the construction industry creates a distressing socio-economic burden globally. It is evident from the disturbing account above that preventing workplace incidents in construction is an urgent need.

Learning from past incidents

Improving safety and preventing incidents in the construction industry has been a top priority for the International Labour Organization and many Work Health and Safety (WHS) authorities around the world. Aligning with the priority, construction safety researchers have so far introduced numerous strategies, models and tools through scientific inquiries involving primary data collection and analyses. While these efforts are commendable, there is a huge potential to create new knowledge and models to improve construction safety by utilising already existing data about workplace incidents. Chua and Goh (2004) advocated that in order for the construction industry to improve its safety performance, it should learn from its mistakes and put the lessons learnt to good use. WHS authorities such as the Health and Safety Executive (HSE) in the UK, Occupational Safety and Health Administration (OSHA) in the US, and Safe Work Australia spent a significant amount of resources to collect data related to construction incidents. An enormous number of incident records exists with such authorities, in anecdotal form, with ample predictive and analytics potentials, which can be leveraged to develop incident prevention strategies (Panthi and Ahmed 2015).

The WHS authorities regularly analyse the data they collect and generate summaries, trends, charts, industry-specific statistics and comparisons to promote policy development initiatives. For example, Safe Work Australia publishes the following reports regularly, drawing from the incident and workers’ compensation claims data it gathers:

• Lost time injury frequency rates for different industry sectors.

• Industry-based statistics that show an overview of the main causes of injuries and fatalities.

• Work-related fatality reports that summarise work-related deaths occurring in Australia.

• Work-related disease reports that identify concerning work-related diseases and their originating sources.

• Work-related injury reports that categorically summarise compensable injuries according to the nature of injury, incident mechanism, agency of injury and characteristics of workers.

• Costing of work-related injury and illness that provides an update of the ‘human cost’ of work-related injury and illness to the Australian economy.

Similar analyses are performed by HSE and OSHA. Construction researchers and commissioned research centres also have utilised these data or samples thereof for developing theoretical models. Notably, Dumrak et al. (2013), as a result of analysing workers’ compensation data obtained from Safe Work South Australia, suggested a conceptual model to explain why some injuries are more severe than others. Haslem et al. (2005) studied 100 incident reports obtained from HSE and developed a model showcasing a hierarchy of causal influences in construction incidents. Huang and Hinze (2003), using data obtained from OSHA, revealed the characteristics and causes of fall incidents in construction. Chua and Goh (2010), utilising 140 construction incident cases obtained from the Occupational Safety Department of Singapore, developed a case-based reasoning system to aid construction hazard identification and safety planning.

Nonetheless, most of the analyses performed are either descriptive, which generate historical summaries, percentages and indexes (Ural and Demirkol 2008), or linear models that evaluate associations between incidents and possible causes (Karra 2005). The descriptive summaries produced by the authorities are helpful to some extent, but the data can be utilised to extract more insights. Likewise, the linear models generated by researchers do not represent the true nature of complex relationships between incidents and causes, and thereby restrict the possibility for explaining and interpreting workplace incidents in terms of the entire range of variables and effectively evaluating hypotheses aimed at curtailing those (Rivas et al. 2011). Therefore, more sophisticated approaches need to be deployed to enable improved analyses of these incident data and the extraction of more insights, patterns and knowledge to prevent workplace incidents.

Data mining and analytics for incident prevention

Data mining and analytics focuses on discovering new and interesting patterns, insights and knowledge in large datasets, differently from traditional statistical methods, which can be utilised for finding yet unrecognised and unsuspected facts (Äyrämö et al. 2009). Because of their powerful predictive capabilities, data mining and analytics approaches are used quite extensively in fields such as medicine, engineering, finance and business (Rivas et al. 2011). In recent years, the automobile sector utilised the approaches to mine traffic incident data to improve road safety (Beshah and Hill 2010; Äyrämö et al. 2009). It has been proven that data mining and analytics techniques can offer advanced predictive and explanatory capabilities for workplace risk and safety management (Martin et al. 2014). The mining and extractive industries have already started benefiting from them (Jacinto and Soares 2008; Silva and Jacinto 2012). However, it is an underexplored area for analysing workplace incidents in the construction industry (Hsueh et al. 2013). There are few studies that demonstrate descriptive analytics only; for instance, Cheng et al. (2012).

Researchers in other industries have indicated many benefits of applying data mining and analytics methods for investigating workplace incident data, namely:

• Data mining and analytics techniques such as decision rules, classification trees and Bayesian networks are reliable tools compared with classical statistical techniques in predicting and identifying factors underlying workplace incidents (Rivas et al. 2011).

• Data mining and analytics allows multivariate analysis of nominal variables with three, four or more categories, which is a challenge for statistical techniques, and provides a much more detailed and complete characterisation of incident patterns (Silva and Jacinto 2012).

• Data mining and analytics techniques enable studying the complex structure of interactions between all variables associated with incidents (Marques et al. 2014).

• Machine learning techniques, a family of data mining and analytics, have the capacity to investigate and automatically detect useful aspects of workplace incidents, which are hidden within the massive amount of information (Ciarapica and Giacchetta 2009). It is not necessary for the researcher to know the solid underlying relationships between input and output variables for model building; any relationship whether linear or nonlinear can be leant and approximated accurately through machine learning (Thipparat 2012).

• Incident related data are multidimensional, heterogeneous and may contain incomplete and erroneous values, which makes their exploration very challenging; yet, data mining and analytics methods are able to produce understandable patterns and useful results (Äyrämö et al. 2009).

Neverth...