Investigation of Causative Factors Associated with Summertime Workplace Fatalities
results of detailed data analysis
Overview of Workplace Fatalities from 2000 to 2005
This section reports the overall trend of fatality occurrence on an annual basis over the past 6 years (2000-05), including the overall annual trend for all industries and all regions; the annual trend for different industries and for different geographical regions. Further detailed analysis on seasonal trend of workplace fatalities is reported in the section entitled 'Seasonal Trend of Workplace Fatalities'.
Overall workplace fatalities by year
The workplace fatality data collected over the past six years (2000-05, 362 cases in total) were first analysed on an annual basis to identify the pattern of changes in the fatality rate over this period. Figure 1 shows that the annual trend in workplace fatalities once peaked between 2001 and 2002 (73-74 cases per year), and since then there was a gradual decline year on year. However, the occurrence of workplace fatalities has increased from 47 cases in 2004 to 65 cases in 2005, or by some 38%. Statistical analysis (chi2 test) has shown that such an increase is statistically significant at p≤0.01 level.
However, there is a possibility that the increase in the number of fatal incidents in recent years is due to a greater number of people working in industry, thus increasing the chance of having more incidents overall during this time. The data were therefore normalised by the total number of employees over the same period, which gives a fatality rate per 100,000 employees per year, as shown in Figure 2.
A similar trend in the occurrence of workplace fatalities was observed with the normalised data as was observed with the original fatality data. The result indicates a 34% increase in workplace fatality rate in 2005 (at the rate of 0.78) relative to 2004 (0.58). It is therefore evident that there has been a recent increase in the number of workplace fatalities since the end of 2004.
It is also useful to see whether or not the trend of workplace fatality rate would be compatible with that of the serious harms data. The serious harm data set is far bigger than the fatalities data set and therefore a potentially more reliable source of information. It is generally accepted that trends in fatality occurrence would follow trends in the occurrence of serious harm incidents. Figure 3 shows the normalised results of workplace fatalities and serious harms over the same period from 2000 to 2005. It is interesting to see that over the same period from 2004 to 2005, while the fatality rate has increased by some 34%, the rate of serious harm injuries has dropped by some 41%. Such a change in the rate of serious harms is statistically significant (p≤0.001).
One of the reasons for such a difference in trends between the two types of data sets is possibly due to the change in incident reporting policies and the method/standard of data recording with respect to the serious harm injuries. Some ISEs commented during interviews that under the government policies and regulations, both employers and employees are encouraged to report serious harm workplace injuries. However, whether or not an injury should be counted as a serious harm injury has always been a very grey area. This has led to the receipt of many reports which should not technically be classified as serious harm, yet all of the reports have been logged in the serious harm data. The issue of over-reporting in the serious harm database was raised by several ISEs. However, there have been some changes in more recent years with respect to the categorisation and recording of serious harms, which may have resulted in reduced number incidents since 2005. Further investigation and analyses of the serious harms database will confirm or discount this type of postulation and provide for a better understanding of the disparity in trends between the two databases. These issues are discussed further in the 'Discussion' section.
Overall workplace fatalities by industries
The workplace fatality data were also analysed for annual trend by different industries. Figure 4 shows the number of workplace fatalities for different industries as classified by ANZSIC system over the period from 2000 to 2006. This indicates a relatively higher fatality rates for the industries including 'Agriculture, Forestry & Fishing', 'Construction', 'Cultural & Recreational Services', 'Transport & Storage', and 'Manufacturing'. Figure 5 shows the normalised results for these industries with relatively higher fatality rates over the past 6 years (2000-05).
The normalised data clearly demonstrate that efforts to reduce or eliminate workplace fatalities should be targeted at the ANZSIC industrial classification 'Agriculture, Forestry & Fishing'. This is discussed further in the section entitled 'Seasonal trend of workplace fatalities by industries'.
One of the useful aspects of these results is that 4 out of 5 major industrial sectors (as shown in Figure 5) have shown an increase in their workplace fatality rates over the past two years. The 'Agriculture' sector in particular started such an upward trend since 2003, followed by the sector of 'Transport & Storage' and 'Construction'. However, it must be noted that the annual variation in fatality rates for all these industries does not reach a significant level (p>0.05). This is mainly due to the fact that the data set is small for each industrial sector. A larger data set may be able to confirm an actual seasonal trend for different industrial sectors, but at this stage, the graphs in Figures 4 & 5 should be used for illustration purpose only.
Overall workplace fatalities by geographical regions
The workplace fatality data were also analysed for their annual trend by geographical regions. There are 17 Workplace Services regional offices throughout New Zealand and the fatal incidents have been recorded under these regional names. However, if the overall fatality data set is further divided by such a number of regions, the resultant data set would be so small that no meaningful analysis could be achieved. In addition, maintaining the granularity of 17 regions is unnecessary since any preventative strategies would not be targeted at such small areas particularly since there is no real geographical difference between many of the regions, the boundaries of which represent administrative divisions only.
Therefore, after the consultation with the ISEs from the Department of Labour, the areas of New Zealand are classified into four larger regions, including Northern Region, Mid North Region, Central Region, and Southern Region. A map to show these regions is in Appendix C.
The 17 Workplace Services regional offices (and the corresponding fatality data) were coded by their geographical locations within the four larger regions. The data were also normalised by the number of employees in the corresponding regions over the same period of time. Figure 6 shows the normalised results.
In general, the normalised regional data show that the 'Mid North Region' had the highest rate of workplace fatalities by the end of 2003, but it was overtaken by the 'Southern Region' from 2004. Overall all regions have indicated an increase in workplace fatalities for 2005. However, due to the small data set, all the annual changes within these regions are not statistically significant (p>0.05), suggesting that these results should be taken as indicative rather than conclusive.
This section reports on the analyses of seasonal trends in workplace fatalities. The analyses explore seasonal trends in workplace fatalities for the data set as a whole; by industry and region; by employment status; and by age group.
Overall seasonal trend of workplace fatalities
The overall workplace fatality data were analysed on monthly and quarterly (seasonal) basis. Figure 7 shows the monthly trend of workplace fatalities from 2000 to 2005. The result is presented with the month 'July' on the far left and 'June' on the far right. This is to accommodate the financial year along which the data were recorded and to show the summertime period (Dec.-Feb.) close to the central section of the graph, making the results easier to read.
Despite the fact that there appears to be a January peak in fatality occurrence over the past six years, statistical analysis with the chi2 test showed that the overall monthly variation of workplace fatalities was insignificant (p>0.05). This is again considered to be due to the relatively small original data set which is further divided into 12 separate data groups for each month of the analysis.
Figure 8 shows the seasonal change in workplace fatalities over the past six years (2000-05). In this analysis, the seasons in New Zealand are defined as the Spring (September-November), Summer (December-February), Autumn (March-May), and Winter (June-August). The results indicate that there is a relative 22% increase in fatality occurrence for the summer (105 cases) as compared with the spring (86 cases). There is also a higher level of workplace fatalities in the autumn. Relatively fewer fatal incidents occurred during the winter as compared to other seasons. The same pattern was confirmed by the normalised data (shown in Figure 9). This seasonal trend has been found to be statistically significant (p≤0.05).
Seasonal trend of workplace fatalities by industries
The seasonal trend of workplace fatality data over the past six years (2000-05) was analysed for all the industries as classified by the ANZSIC system. The results showed that only one industrial sector (i.e., 'Agriculture, Forestry & Fishing') had a significant seasonal variation in their fatality rate (p≤0.01). All the other industries had little or insignificant seasonal variation, as illustrated in Figure 10.
In order to get an insight within the agriculture sector as to which month has the highest number of fatalities and which sub-sectors are at higher risks in relation to fatal incidents, further detailed analyses were carried out. Figure 11 shows that there is a January peak in workplace fatalities within the agriculture sector, followed by two more peaks, one is in April and the other in October. This trend is statistically significant at p≤0.05 level.
Figure 12 shows that within the agriculture sector, 'Forestry and logging' is the sector with the highest summertime fatality rate, followed by 'Horticulture & fruit growing'. However, these apparent seasonal peaks did not reach an acceptable level of significance (p>0.05). The results in Figure 12 therefore need further analysis. An important next step in targeting preventative measures within the agriculture industry sector is to gather employee numbers  for the industrial sub-sectors (e.g. Forestry and logging, Horticulture & fruit growing, Services to agriculture, Dairy cattle farming, etc.) that form part of this industry group. Using these data, spikes in the fatality rates associated with the highest risk sub-sectors and sub-sub-sectors can be identified providing clearer direction for fatality prevention strategies.
Seasonal trend of workplace fatalities by regions
Figure 13 below illustrates the seasonal trends associated with workplace fatality rates by geographical region. Of the four regions, both the 'Mid North region' and the 'Southern region' show a summertime increase in fatality rate. The Southern region also has a higher level of workplace fatalities in the autumn. Due to the limited number of cases recorded over the past six years for these regions, only the data set from the Southern region was strong enough to reach a statistically significant level (p≤0.05).
Seasonal trend of workplace fatalities by employment status
The workplace fatality data over the past six years (2000-05) were analysed by the victim's employment status, including the victims who were either employees (EMP), self employed (SE), or bystanders (BS). Figure 14 shows some indicative results, although all the seasonal trends of the data as analysed by employment status did not reach a statistically significant level (p>0.05).
Figure 14 suggests that it is the employed who are most at risk of having a workplace fatality. However, this assumption cannot be made until denominator data can be obtained. Given that the numbers of self employed are likely to be relatively low, normalised data may show a very different picture.
(Note: EMP-employees; SE-self employed; BS-bystanders)
Seasonal trend of workplace fatalities by employment status for different industries
Further exploration of the employment status by industry, shows that the people working in the agriculture sector, either employed or self-employed, tend to have more fatal incidents in summertime as compared to other seasons. This seasonal trend is very close, but not up to, a significant level (p=0.073, close to 0.05).
(Note: EMP-employees; SE-self employed; BS-bystanders)
Seasonal trend of workplace fatalities by employment status for different regions
The workplace fatality data were analysed for a possible seasonal trend by employment status for geographical regions, including the Northern region, Mid North region, Central region and Southern region. Although some apparent seasonal changes in workplace fatalities (with the EMP data, for example) are shown in Figure 16, these changes are not statistically significant (p>0.05).
The normalised results in Figure 17 (with the EMP data only due to limited data availability) suggest that while in the northern part of New Zealand (Northern and Mid North regions) there tend to be more workplace fatalities in the summertime (highlighted in red bars), in the southern part of the country (Central and Southern regions), more fatal incidents tend to occur in the autumn (orange bars). The Central region also tends to have a higher fatality rate during windy springtime. It needs to be noted here, however, that the evidence obtained so far from the fatality data analysis is technically not strong enough to reach a conclusion, and these results should be interpreted with the support from other independent evidence.
Seasonal trend of workplace fatalities by age groups
Figure 18 below illustrates the seasonal trends of workplace fatalities in relation to the age groups of those involved in the incidents. The data were based on total fatal incidents from 2000 to 2005.
The data indicate that the people with the highest summertime fatality occurrence are between 55 and 64 years of age. This is a significant finding at the at p≤0.05 level. The data also shows that those aged 35-44 are more likely to have fatal incident in the autumn. This was significant at the p≤0.05 level. Other apparent seasonal trends with different age groups were not statistically significant (p>0.05).
(Overall occurrence 2000-05)
Figure 19 shows the normalised results of the seasonal trend by age groups. One of the major findings from the normalised data is that the results confirm the high-risk group of 55-64 for summer season (red columns). The normalised data also confirmed the autumn peak of work-related fatalities with the 35-44 age group.
In addition, the normalised results suggest that there is a gradual increase in fatality rate by those aged 65 and over from the summertime through to the spring, with the summer having the lowest fatality rate and the spring having the highest rate. The overall fatality rate of the 65+ group is much higher than that of any other age groups. The next section investigates which industrial sector these high-risk groups tend to work in, and whether it would follow a similar seasonal pattern.
Seasonal trend of workplace fatalities by age groups for different industries
The workplace fatality data from 2000 to 2005 were analysed by each age group, for each industrial sector as classified by the ANZSIC system. Statistical analysis was carried out to identify if there is a significant seasonal change for each age group within each industry. The results demonstrate that only the agriculture sector had a significant seasonal trend of age-related workplace fatalities, with the 55-64 age group having the highest fatality occurrence in the summer (p≤0.05) and 35-44 age group having the highest fatal incidents in the autumn (p≤0.05) (Figure 20). These results are in line with the previous findings confirming that, as far as workplace fatalities are concerned, the employees aged 55-64 are most vulnerable during the summertime; the employees aged 35-44 are most vulnerable during the autumn, and most of them are working in the agriculture industry sector.
People in other age groups, do not show a significant variation in their seasonal involvement in workplace fatalities for the agriculture industry. No significant seasonal patterns were found with any age groups working in any other industries, except for agriculture.
The normalised results by age groups for the agriculture sector show a summer peak fatality rate with the 55-64 age group (Figure 21). This result is consistent with the previous findings. Figure 21 also demonstrates that those 65+ workers tend to have higher fatality rate during autumn while their involvement in other seasons is relatively stable.
Figure 21: Normalised fatality rate with people in different age groups in the 'Agriculture, Forestry & Fishing' industry (2000-05)
Seasonal trend of workplace fatalities by age groups for different regions
The workplace fatality data over the last 6 years (2000-05) were analysed for each geographical region by age groups. Of the four large regions and a total number of 362 fatalities over the 6-year period, the Northern region had 101 fatalities (or 27.9% of the national total), the Mid North region had 73 (20.2%), the Central region had 101 (27.9%), and the Southern region had 87 (24.0%). However, when dividing the data set by regions and then by different age groups, the actual data size becomes so small that statistical tests indicate only the Southern region had a significant seasonal variation (at p≤0.05 level) in workplace fatalities (between all age groups). In addition, employee population may vary between these regions, thus the data would need to be normalised to make the results meaningful. The normalised results for these regions are shown in Figure 22 below. These results will be discussed later in section 6.
Mid North region
(See footnote 10 for data source)
Time of day
Figure 23 shows the fatal incidents that occurred in the workplace from 2000 to 2005 at different times of day. There are two peaks for the fatal occurrences, one is between 11:00 and 12:00, and the other is around 15:00. These variations are highly significant at p<0.001 level.
Further analysis of the fatality data by industry indicates that such a pattern is in line with the incident occurrence in the agriculture industry (p<0.001), while the daily variations of the incident times for all other industries are insignificant, as shown in Figure 24.
Figure 25 shows the trends of workplace fatalities between time of day and the season. To achieve this analysis, software code was written to work out whether fatalities occurred during daylight, twilight or night time hours. This code looped through each record and calculated the region and then the sunrise/sunset times for that region at that time of year. Five times of day were identified:
- Dawn - Twilight to Sunrise
- Morning - Sunrise to Noon
- Afternoon - Noon to Sunset
- Dusk - Sunset to Twilight end
- Evening=Twilight end to Twilight start
The graphs in Figure 25 suggest that during the summer, fatal workplace incidents are more likely to occur in the morning (from sunrise to noon), and during autumn, more fatal incidents tend to occur in the afternoon (from noon to sunset).
Day of the week
The workplace fatality data over the past 6 years (2000-05) were also analysed by day of the week, and the weekly trend is shown in Figure 26 below. As expected, more fatal incidents are likely to happen during the weekdays (Monday to Friday) as compared to weekends (Saturday to Sunday) and such a difference is statistically significant (p<0.001). However, there is no significant difference between weekdays (Monday-Friday) with regard to the frequency of fatal occurrences (p>0.05). There is also no significant variation in fatality occurrence between Saturday and Sunday (p>0.05), although there are slightly more incidents on Saturdays than on Sundays in the 6-year record. These issues will be discussed in further detail in section 6 with other independent evidence.
 Source: Statistics New Zealand, Household Labour Force Survey (HLFS), 1986-2006.
 Not available within the timeframes for this study
 Data source: Employment by age (1986-2006), Statistics New Zealand, Household Labour Force Survey (HLFS).
 Note: These normalised results were based on the denominators which were available up to 2001. These were all that were available at the time when the analysis was carried out.
 Source: Statistics New Zealand, 2001 Census.