Investigation of Causative Factors Associated with Summertime Workplace Fatalities
methodology
Data Collection and Collation, and Review of Workplace Fatality Investigation Reports
The Department of Labour provided the researchers with a database of 365 workplace fatality records from 2000 to 2005 inclusive. It was appreciated that while this data set might appear to be relatively small for the scale of the analysis to be conducted, it was the most reliable data set available at the time of the study and was considered sufficient to obtain some good indications regarding the objectives of work.
The fatality data set was sorted for any possible recording errors and/or record duplications. As a result, a total of 362 fatal incident records were identified to be usable for further detailed analysis.
On the basis of the fatality data, a master datasheet was developed with various information fields against each record. Some of the fields were already available with the basic data set (e.g., date of incident, Workplace Services (OSH[2]) region, age of the victim), and some were added by the project team in order to facilitate the required analyses. The additional incident classification fields added by the project team were used to record the contributing factors associated with each incident. The classification system was based on the Human Factors Analysis Classification System (HFACS) (Wiegmann & Shappell, 1997; Shappell and Wiegmann, 2000) and is summarised in Table 1 below. The classification system was agreed to in advance by the Department of Labour project management team. Each of the fatality reports provided by the Department of Labour (315 in total) were systematically reviewed and the contributing factors were recorded accordingly.
Table 1: Classification System
In addition, complex formulae were developed which allowed for geographical location, time of year and daylight savings to determine whether incidents occurred during daylight hours, dawn, dusk or at night time.
The original incident data were classified into five broad categories, including Agriculture, Construction, Forestry, Mines & Quarries, and Industrial/Commercial. Each incident was recorded against one of these five categories. While it may be relatively simple to record and to analyse the incident data under these broad categories, they were considered too broad to provide a useful classification for targeting preventative strategies. 'Industrial/Commercial', for example, included a great number of very varied industry sub-sectors. Therefore, each incident record was re-coded using the ANZSIC (Australian New Zealand Standard industrial Classification Codes) system. This classification was considered to be the most appropriate in order to accurately determine which industry sectors are most at risk of fatality events. It should be noted that the ANZSIC system classifies the 'Agriculture, Forestry & Fishing' as one industry sector and this report uses this as a standard classification (and sometimes 'agriculture' for short), although the Department of Labour data does not cover fishing. The ANZSIC coding of the fatality data was completed at the Department of Labour by a safety expert and checked by the project technical team. A high level of the ANZSIC system is shown in Appendix A.
Consultation with ISEs (Industry Safety Experts)
Semi-structured interviews were conducted with a number of ISEs with expertise and experience in safety and health, and/or in related subject areas. The ISEs represented different industries, research organisations and regions in New Zealand.
Most consultations were carried out through face-to-face interviews in Wellington and Auckland, and each interview lasted between 1.5 and 2 hours. The interviews were guided by a list of questions and before they attended the interview, the ISEs were sent a 'Briefing Note' that explained the purpose of the interview. The note contained a list of questions/issues to be discussed.
ISEs from a particular type of industry were asked to complete a survey form towards the end of the interview regarding their opinions on the seasonal activities within that industry. They were asked to give a relative rating (in terms of 'high', 'medium' or 'low') on a number of workplace factors for their seasonal changes across the year. Some of these factors include 'hours worked per day'; 'level of staffing'; 'physical or mental work demands'; 'non-work related activities'; 'work involving machineries'; 'variation in work activities'; 'need to work in remote locations'; 'fatigue/tiredness' and so on (16 topics in total). The ISEs were encouraged to add more topic issues if they were not included in the rating list. It was the intention of the exercise to help the research team better understand the industry-specific issues across different seasons of the year. The Briefing Note and the Seasonal Activity Survey Form are provided in Appendix B.
The ISEs consulted represented a wide range of industries and type of organisations. For practical reasons, it was not possible to interview representatives from all industry sub-sectors across the country. However, sufficient information was obtained to achieve the objectives of the present study. The list of organisations is shown below:
- Department of Labour: Team Leader; Health and Safety Advisor; Principal Advisor; Senior Policy Advisor; Business Advisor; Health & Safety Inspectors; Safety Engineer and Data Entry Contractor.
- Employers & Manufacturers Association (EMA) (Northern region): Occupational Health & Safety Managers.
- Federated Farmers of New Zealand (Inc.): President; Policy Advisor.
- Horticulture New Zealand: Chief Executive; Senior Business Managers.
- Site Safe New Zealand: Executive Director; Training Manager.
- Roading New Zealand: Chief Executive.
- EMA (Central region): Health & Safety Consultant.
- New Zealand Safety Council: Executive Director.
- University of Otago: Injury Prevention Research Unit, Dunedin School of Medicine.
- University of Auckland: Department of Management and Employment Relations.
- Massey University: Occupational Health & Safety
- Fletcher Construction Auckland: Operations Manager.
- Works Infrastructure: General Manager.
- OnTrack, New Zealand Railways Corporation: Risk & Safety Manager.
- Amalgamated Workers Union: Auckland Representative.
Literature Review
A comprehensive literature review was conducted. The focus of the literature review was to review research findings of seasonal issues in relation to workplace safety and to identify independent evidence pertaining to the factors which are shown to be significantly associated with the occurrence of workplace fatalities. A wide range of information sources were used for the literature search. Some of the primary sources of the research data are listed below:
- Human Engineering's internal library (both in Australia and UK) and electronic database, which contains a wide range of published literature on workplace safety and general ergonomics/human factors reports.
- Human Factors Research Catalogue CD (UK, 2005-06)
- HSE research reports database online[3].
- SCIRUS[4] (one of the most comprehensive science-specific search engine on the Internet).
- NCBI[5] (National Center for Biotechnology Information).
- Statistics New Zealand[6]
- National Farm Injury Data Centre[7]
Detailed Data Analysis
Trend analysis
The workplace fatality data were analysed for their annual and seasonal trends, firstly based on the overall fatality data, and secondly on the cases related to different industries and regions. Detailed analysis was also performed to investigate the relationship between such factors as employment status and workers age, and the occurrence of workplace fatalities with seasonal changes.
Data normalisation was carried out as necessary and when a reliable denominator was available. This reduces or eliminates the potential bias of incident occurrence due to the variation of some 'risk exposure' factors, such as the change in total population of the workforce over time.
Statistical analyses were conducted, using chi2 tests, on the annual/seasonal trends of fatality occurrence to test if the variation between seasons (and years) was statistically significant. The basic principle of these tests was to identify whether a certain 'peak season' (i.e., with a relatively higher number of fatal incidents) was due to chance. If the data show that for a certain month the number of incidents is greater than the monthly 'norm', the chi2 test will be able to verify whether or not the peak is for certain, or due to chance. The minimal acceptable level of significance was set at p≤0.05, which means that the likelihood of a peak (or a trough) occurrence of the event (in this case, the number of incidents per month) by chance is less than 5%. Where p≤0.05, it is significantly likely that the variation of event occurrence is due to other factors.
In most cases in this report, a trend graph is presented only when the results are found to be statistically significant. This is because the small numbers involved can lead to very visible peaks and troughs in groups that are not significant and entirely due to chance, thus the figures can be misleading. In certain instances, however, some normalised results (although not statistically significant) are also presented, but these are for illustration purpose only, and this will be clearly stated in the relevant section in relation to the illustration.
95% Confidence Intervals (CI) were calculated (shown in the graph as error bars at each data point) with some normalised data when reliable denominators were available. These would normally indicate that 95% of the observed (or recorded) data are expected to fall within the specified range of variation.
Identification of a wide range of associated factors for fatality occurrence
The fatality database has recorded only a small number of relevant factors in relation to each incident (e.g. date, industry, region, employment status, age, etc.), but there are far more factors that may have played an important role in the occurrence of fatal incident across the seasons. These include environmental factors, sociological factors, individual factors, and workplace factors. Statistical tests with correlation analysis (using both parametric and non-parametric tests depending on the type of data) were conducted to identify the level or strength of association between each type of the factors and the workplace fatalities. The significance levels of association were divided into four levels, including p≤0.05, p≤0.01, p≤0.001, and NS (not significant, p>0.05). These will also be described in the relevant section.
