New research combining actuarial modelling with machine learning has found that heat-related workplace injuries in Australia are being understated by a factor of 20, with workers in remote regions and high-risk industries facing the greatest burden.

The findings, presented by Taylor Fry’s Dr Xi Lin at the 2026 UNSW Workshop on Risk and Actuarial Frontiers: Climate Risk and Insurance, draw on a national workers' compensation dataset covering two to three million claims and represent one of the first attempts to quantify the true scale of heat's contribution to workplace injury in Australia.

Official data captures only a fraction of heat-related claims

The discrepancy between recorded and estimated heat-related injuries is stark. Of the millions of claims in the national dataset, just over 1000 were explicitly coded as heat stress, heat stroke, or exposure to environmental heat. That figure represents 0.05 per cent of all claims, or roughly five in every 10,000.

"So if we just rely on official coding and only rely on it, we can only identify maybe 5 out of 10,000 claims," Dr Lin said. "Surely this is not true, given the heat we are feeling."

The research team, which included Associate Professor Fei Huang and Xinyi Yan from UNSW, and Dr Ramona Meyricke from Taylor Fry, used a machine learning model trained on a small pool of explicitly coded heat claims to identify additional injuries that carried the hallmarks of heat involvement without being labelled as such. These included claims coded as falls, wounds, or body stressing that occurred during summer and in locations with elevated temperatures at the time of injury, consistent with international literature pointing to widespread under-recognition of heat's role.

Machine learning uncovers the hidden burden

The model was trained on the explicitly coded heat claims as positive cases, with non-heat-coded summer claims used as the reference group. Bureau of Meteorology grid data was linked to individual claims to establish the temperature conditions at the time and place of each injury. A manual review of 750 high-probability cases was then conducted to validate the model's outputs.

"So we randomly sampled 750 claims that are tagged as potential heat-related claims, because we don't just learn from models; models need to learn from us as well," Dr Lin said.

Learn more: Contract labour mining safety: hidden costs and business risks

The results indicated that the true proportion of heat-affected claims was approximately 1.1 per cent of all workers' compensation claims. Sensitivity analysis using alternative modelling approaches produced a range of 0.3 per cent to 2.3 per cent, but Dr Lin said the core finding held across all scenarios. "Although the exact number might be uncertain, the conclusion is robust, which is that heat-related claims are underestimated."

Remote workers and high-risk industries carry a disproportionate load

The research also produced a heat-related workforce injury risk index, built on three components: hazard (measured using wet bulb globe temperature, which accounts for temperature, humidity, wind speed, and solar radiation), exposure (the size of the at-risk workforce by location and industry), and vulnerability (the relative injury rate as a proxy for heat susceptibility).

Using a threshold of 28 degrees wet bulb globe temperature, the index identified East Pilbara in Western Australia as the highest-risk location in the country, with an index value of 49, compared with a national median of 0.75. In East Pilbara, 205 days out of 365 recorded wet bulb globe temperatures above that threshold, driven by a large concentration of mining workers operating in one of Australia's most thermally demanding environments.

^{Source: When the heat is on: Where Australia’s workers face greatest risk, Taylor Fry}

"So in East Pilbara near Perth, 200 days out of 365, it has a WBGT over 28 degrees," Dr Lin said. "This is driving the heat risk in the workers."

Manufacturing workers were identified as the most vulnerable cohort, partly because large warehouse and factory environments are rarely air-conditioned. Brisbane Airport is also registered as a high-risk area, where workers operate on expansive paved and reflective surfaces. The analysis found that heat-related injuries were more concentrated in regional and remote areas, with far west New South Wales recording a rate around 200 per 10,000 claims compared with 140 in metropolitan Sydney, pointing to a dimension of geographic inequality in heat exposure.

Prevention, reserving, and climate projections form the next phase

Dr Lin outlined three practical applications for the research. The first is financial reserving: insurers and workers' compensation schemes need accurate data on the true volume and distribution of heat-affected claims to set aside adequate reserves. The second is targeted prevention, directing investment toward workers and locations with the highest heat risk. Smart helmets fitted with temperature sensors, already in use in Japan, were cited as one example of technology that could be deployed more strategically if high-risk cohorts are better identified.

Subscribe to BusinessThink for the latest research, analysis and insights from UNSW Business School

"By knowing who is actually more susceptible or vulnerable to heat risk, you can invest in specific measures," Dr Lin said. "If you know who is most vulnerable, maybe you can just invest in some smart gadgets like that."

The third phase of the research involves combining the index with climate projections to model how the risk landscape will shift over the next two decades.

A data problem with systemic consequences

The research highlights a fundamental limitation in Australia's current approach to tracking occupational heat illness. When claims are coded by mechanism of injury rather than contributing environmental factor, heat's role in strains, falls, and other incidents becomes invisible in the data. The result is that prevention investment, scheme reserving, and policy responses are all calibrated against a figure that reflects only a small fraction of the actual problem.

The findings add to a body of international evidence suggesting that the global burden of heat-related workplace injury is substantially larger than official statistics indicate, with some studies estimating undercount factors of 3-7.5 times. Australia's figure of 20 times suggests the gap here may be particularly pronounced, though Dr Lin noted that further research using unit-record data and individual-level vulnerability measures would refine the estimate.

"By knowing who is actually more susceptible or vulnerable to heat risk, you can invest in specific measures"

XI LIN

The collaboration between Taylor Fry and UNSW is continuing, with climate scenario modelling the next planned stage of the work.

6 recommendations for business leaders: Managing and mitigating heat risk

For risk managers and insurers: Review workers' compensation reserving assumptions given the possibility that heat-related claims are understated by a factor of up to 20. Current coding practices may materially underestimate financial exposure to occupational heat illness in scheme models.

For OHS and safety professionals: Audit injury reporting systems to capture environmental contributing factors, including heat, alongside the mechanism of injury. Workers in manufacturing, mining, construction, and logistics require targeted heat management protocols, particularly in remote and regional locations.

For CFOs and finance executives: Heat-related workplace injuries entail both direct costs through workers' compensation claims and indirect costs through productivity losses and workforce absences. Research suggests the financial burden is substantially higher than current data indicate, warranting inclusion in climate-related financial risk assessments.

For HR and workforce planning leaders: Regional and remote workforces face materially higher heat exposure than metropolitan workers. Workforce planning, scheduling, and duty-of-care obligations should account for seasonal and geographic heat risk, particularly as climate conditions intensify.

For actuaries and data analysts: The research methodology, combining machine learning classification with manual case review and Bureau of Meteorology grid data linkage, offers a reproducible framework for identifying latent risk factors in large insurance claims datasets beyond heat injury.

For boards and risk committees: Occupational heat illness is an emerging climate-related physical risk with direct implications for workers' compensation liability, regulatory compliance, and workforce resilience. The 20-fold gap between recorded and estimated claims warrants board-level attention within climate risk governance frameworks.

This research uses data from the National Data Set for Compensation-based Statistics (NDS), compiled and maintained by Safe Work Australia from workers' compensation claims data supplied by Australian workers' compensation authorities. Safe Work Australia is Australia's national policy body for work health and safety and workers' compensation.