The burden of occupational injury attributable to high temperatures in Australia, 2014–19: a retrospective observational study

To assess the population health impact of high temperatures on workplace health and safety by estimating the burden of heat‐attributable occupational injury in Australia.


Research
to estimate the number of YLDs associated with occupational injuries in Australia: the Safe Work Australia national dataset for compensation-based statistics 12 and the AIHW National Hospital Morbidity Database 13 (further details: Supporting Information). 9The number of occupational injury-related DALYs was calculated by summing the numbers of YLDs and YLLs.
We report the effects of high temperatures nationally and for each of the twelve Köppen-Geiger climate zones in Australia: tropical (Af, Am, Aw), arid (BSh, BSk, BWh, BWk), and temperate or Mediterranean climate zones (Cfa, Cfb, Csa, Csb, Cwa) (Box 1). 10 We calculated the mean, median, and modal daily temperatures for each climate zone from temperature data in the Scientific Information for Land Owners (SILO) dataset 14,15 for the centroid of each Statistical Area Level 2 (SA2) within a climate zone.SA2s include a mean of 10 000 residents. 16e number of employed people in each climate zone and state or territory was derived from 2016 Australian Bureau of Statistics national census SA2-level labour force data (using Census TableBuilder), 17 and the proportions of the national total calculated.Population-adjusted YLDs and YLLs for occupational injury in each climate zone and state or territory were calculated by multiplying the labour force proportion by the national burden of disease, and expressed as number of injuries per 1000 employed people.

Burden of occupational injury attributable to high temperatures
We calculated the heat-attributable burden of occupational injury -the proportion of the total burden attributable to heat exposure -using our previously described methodological framework. 18Heat" was defined as a temperature exceeding that at which the health risk is zero; that is, the theoretical minimum risk exposure distribution (TMRED).The heat-attributable burden is equivalent to the reduction in burden had exposure been limited to temperatures no higher than the TMRED.9 For the current study, we defined the TMRED as being the annual mean temperature for the climate zone or state or territory, 19 as the local distribution of minimum mortality temperatures (ie, temperatures at which mortality risk is lowest) is closely linked with annual mean temperature (based on mean of daily maximum and minimum temperatures in the SILO dataset).20 We calculated the prevalence of exposure to high temperatures during 2014-19 in each climate zone as the proportion of days during the financial year on which the temperature exceeded the TMRED.
We categorised annual mean temperatures by one degree Celsius category and determined the TMRED for each climate zone.To estimate the relative risk (RR) per unit change in temperature, assuming a log-linear relationship between occupational injury and heat, we applied the formula: in which c is the temperature category, z the climate zone, and T the TMRED (ie, annual mean temperature).
The RRs for heat-associated occupational injury by global climate zone (RR z ) were derived from our recent systematic review and meta-analysis. 21This report did not include five of the twelve climate zones relevant to Australia (Af, Am, Aw, BWk, Cwa).For the latter two zones, we substituted RR values for similar climate zones: BSk (semi-arid, cold) for BWk (arid, desert, cold) -the two zones have similar mean annual temperatures (16.5°C and 17.8°C respectively) -and Cfa for Cwa (both humid subtropical zones; mean annual temperatures: 19.4°C and 22.6°C respectively).RRs for the Af, Am, and Aw (tropical) climate zones were based on a study of heat-associated occupational injury among sugarcane harvesters in Guatemala, 22 the only relevant published study for this climate zone type.
The population attributable fraction (PAF) for heat-attributable occupational injury was calculated using the comparative risk Research assessment method, based on the increase in RR for occupational injury associated with exposure to the risk factor and the estimated prevalence of exposure: 9 in which z is the index category (climate zone), RR z is the category-specific relative risk, P z is the prevalence of exposure to the risk factor in the climate zone, and ∑ z is the sum of the proportion of the year during which the temperature exceeded the TMRED for each categorised temperature value in the distribution and the corresponding estimated RR for the climate zone.Population-adjusted YLDs and YLLs were multiplied by the PAF to calculate the heat-attributable burden of disease by climate zone or state or territory (state heat-attributable DALYs were calculated by aggregating the climate zone-specific DALYs within the state).The heat-attributable occupational injury DALYs were then calculated and expressed as a proportion of all DALYs associated with occupational injury.We also report annual rates of attributable burden of disease per 1000 employed workers.
In sensitivity analyses, median or modal temperatures (most frequent mean daily temperature) were used instead of mean annual temperature for the TMRED, alternative exposure periods, data sources, and exposure-response relationships (including linear and non-linear forms) were examined, and alternative temperature exposure references (mean annual maximum and minimum temperatures rather than annual mean temperatures) used.
Analyses were undertaken in Microsoft Excel 2016 and Python 2.0.

Ethics approval
The University of Adelaide Human Research Ethics Committee granted our negligible risk study an exemption from formal ethics review.

Results
Within each climate zone, the annual mean and median daily temperatures were similar, but modal daily temperatures for the BSk, BWk, Cfb, Csa, and Csb climate zones were markedly lower and that of BWh markedly higher than the corresponding mean and median temperatures.The highest mean temperatures were for the two northern climate zones (Am and Aw) (Box 2; Supporting Information, figure 1).

Total burden of occupational injury by climate zone and state or territory
During 2014-19, an estimated 42 884 years of healthy life were lost to occupational injury, comprising 39 485 YLLs (92.1%) and 3399 YLDs (7.9%).By climate zone, the largest proportions of the national occupational injury burden were in the Cfa (18 836 DALYs; 43.9% of the national total) and Cfb zones (11 881 DALYs; 27.7%) (Box 3).By state and territory, the largest proportions of the national occupational injury burden were in New South Wales (13 095 DALYs; 30.5%) and Queensland (10 473 DALYs; 24.4%).The highest occupational injury burden rate was in the Northern Territory (1.76 DALYs per 1000 workers; Australia: 0.80 DALYs per 1000 workers) (Box 4).
By state and territory, the heat-attributable proportions of occupational injury-related DALYs were largest in New South

Sensitivity analyses
Sensitivity analyses using different TMRED indicators yielded slightly different heat-attributable DALY proportions for Australia (1.8-3.0%),but they were generally of similar magnitude to the proportion in our major analysis (2.3%) (Supporting Information, table 1).

Discussion
We found that Australian workers lost 42 884 years of healthy life to occupational injuries during 2014-19, an annual rate of 0.80 DALYs per 1000 workers.Heat-related occupational injuries caused the loss of 967 DALYs (2.3% of all occupational injury-related DALYs).The relative risks per degree increase in temperature exposure for heat-associated occupational injury obtained from the systematic review and meta-analysis were highest in the two tropical climate zones (Am, Aw), 19 and the proportions of DALYs that were heat-attributable were also  7 and Italy (primarily Csb) 23 respectively reported heat-attributable fractions of 0.8% and 1.8%.
Heat-related morbidity and mortality are expected to increase in Australia because of global warming, exacerbating productivity losses for industries involving outdoor labour. 24Construction workers in Australia lost 67 565 hours of work because of heat stress (working in direct sunlight) during 2019, more than twice the 10-year mean of 25 240 hours during 1991-2000, 25 indicating the importance of investigating the impact of climate change on health in burden of disease studies.
Estimating the theoretical minimum level of population exposure beyond which disease or injury risk increases is critical for estimating factor-attributable burdens, and relatively small differences in the TMRED can markedly alter PAF estimates. 26However, determining the temperature associated with the lowest risk of heat exposure-related occupational injury

Limitations
Firstly, PAF calculations were based on RRs derived from a systematic review and meta-analysis of overseas data, 21 and we assumed that these RRs applied to similar Australian climate zones.Most of the RRs we adopted were derived from data for countries with climate and socio-economic characteristics similar to those of Australia, but some were derived from data for countries with similar climate but different socio-economic profiles (eg, Guatemala, 22 used for Australian tropical zones).However, cross-validation of the RRs used with unpublished data from Australian studies indicated that they were consistent with local conditions (data not shown).Further, for each zone the same RR was used to estimate the PAF for calculating both the heat-attributable YLLs and YLDs, as the reviewed studies did not distinguish between fatal and non-fatal occupational injury burden. 21condly, we used the annual mean temperature for the TMRED, as other studies have found that the association of mortality with annual mean temperature is a reasonable indicator of population adaptation. 18,20Cross-validation of TMRED with exposureresponse curves from our earlier study 2 and for other Australian cities (unpublished data) indicated that the temperatures associated with lowest occupational injury risk closely matched the annual mean temperature in each climate zone we examined (data not shown).Further, sensitivity analyses using alternative temperature measures yielded results similar to those of our main analysis.
Thirdly, we did not consider adaptation or acclimatisation in our analysis, nor restricted access to health care in regional and remote areas.Fourth, as the heat-attributable occupational injury burden may differ by specific location within climate zones or jurisdictions, our estimates cannot be applied to specific locations in heterogeneous urban or sparsely populated regional or rural areas.Our findings may also have been influenced by industries with workers at greater risk of heat exposure, such as agriculture or mining.Finally, demographic, health, and socio-economic differences were not considered by our analysis, nor did we stratify our analyses by sex, age, or occupational characteristics.

Conclusion
Despite these caveats, our estimates suggest that the impact of high ambient temperatures on occupational injury in exposed workers is not trivial.Our study, one of the first to estimate the burden of heat-attributable occupational injury, highlights a problem that will increase as temperatures rise with climate change.It is imperative that workplace health and safety be safeguarded during extreme heat, as is protecting workers in industries such as agriculture, transport, and construction, and those in poorly ventilated indoor workplaces.Measures could include restructuring of work hours, providing adequate rest breaks, shaded or cooled rest areas, cool drinking water, personal protective equipment, and wearable cooling devices, and use of health monitoring technologies.
We have quantified the impact of high ambient temperatures on the occupational injury burden in the twelve climate zones of Australia.During 2014-19, 2.3% of the national occupational injury burden was attributable to heat exposure, and the proportion was larger in the tropical climate zones (Am and Aw) and in the Northern Territory.Without adaptive measures and industry-based policies, the heat-attributable occupational injury burden will increase as climate change advances.

July 2014 -30 June 2019, by state and territory Years of life lost Years lived with disability Disability-adjusted life years
Our results are consistent with those of studies that have used other methods to estimate the fraction of occupational injury attributable to heat.For example, we found that 2.0% of workers' compensation claims in Adelaide (climate zone Csa) during 2003-2013 were attributable to high temperatures; 1 our estimate for the Csa climate zone in the current study was 1.7% of occupational injury-related DALYs.Similarly, studies from Spain (primarily Csa)