The study’s design is a textbook example of modern environmental epidemiology. The team drew on the Ministry of Health Mortality Collection within Stats NZ’s Integrated Data Infrastructure, capturing every registered death in the country from 1 January 1999 to 30 April 2025. Each death was matched to daily heat exposure in the person’s Territorial Authority of residence, reconstructed from weather-station observations in the National Climate Database maintained by the National Institute of Water and Atmospheric Research. To estimate the acute effect of heat, the researchers used a time-stratified case-crossover design, a self-matched method in which each person’s exposure on the day of death is compared with exposure on nearby referent days falling on the same weekday within the same calendar month. Because every individual serves as their own control, time-invariant personal characteristics, seasonality, long-term trends and day-of-week effects are absorbed by the design itself, leaving a clean estimate of the short-term heat-mortality association.

What distinguishes this study from most of its predecessors is the choice of exposure metric. Alongside conventional air-temperature summaries, the team computed the wet-bulb globe temperature, or WBGT, a composite heat-stress index that integrates air temperature with humidity, radiation and wind speed. Developed in the 1950s to prevent heat illness among US military recruits and later codified in occupational health standards such as ISO 7243, WBGT captures the combined thermal load on the human body far better than a thermometer reading alone. The researchers applied the Liljegren physical model to translate standard meteorological measurements into daily WBGT indices based on mean, maximum and minimum temperatures, creating three parallel exposure measures for every district and day. Missing radiation and wind inputs in a small fraction of district-days were filled using multiple imputation by chained equations, a statistically rigorous approach that draws on contemporaneous and recent meteorological information.

The headline results are striking in their consistency. In the all-cause, full-year sample, the odds of death rose by 0.44 percent for every 1 degree Celsius increase in WBGT based on mean daily temperature, and by 0.39 percent per degree of mean air temperature. Restricting the analysis to the warm season, from October to March, sharpened the associations to 0.57 percent and 0.51 percent respectively. Mean-based measures outperformed maximum- and minimum-based measures within both metric families, suggesting that sustained daily heat load, rather than brief afternoon peaks, is what matters most for short-term mortality. When the team compared models statistically, WBGT-only models fit better than temperature-only models, with lower Akaike information criterion values and significant likelihood-ratio improvements, indicating that humidity, radiation and wind carry genuine additional information about lethal heat beyond what a thermometer can see.

To translate these odds ratios into human terms, the researchers performed a simple scaling exercise. A uniform 1 degree Celsius increase in mean WBGT across the country would correspond to roughly 140 additional all-cause deaths per year under full-year assumptions, and about 110 additional non-injury deaths. Restricted to the warm season alone, the same shift implies approximately 85 excess all-cause deaths annually. These are illustrative figures that abstract from spatially differentiated warming and behavioural adaptation, but they give policymakers a tangible benchmark: in a nation of about five million people, even a single degree of additional heat stress claims a measurable annual toll.

The heterogeneity analyses reveal where the burden falls. When Territorial Authorities were grouped by their mean warm-season WBGT into cool, moderate and hot climate tertiles, the per-degree odds increase climbed monotonically from 0.17 percent in the coolest districts to 0.43 percent in moderate ones and 0.67 percent in the hottest. This gradient is consistent with an emerging international pattern in which populations in historically cooler regions may be less adapted, yet the strongest associations here appeared in the warmest districts, hinting that sustained exposure compounds risk. Women showed consistently higher point estimates than men, with a 0.53 percent odds increase per degree of WBGT-Tmean compared with 0.34 percent for men, and the clearest age-specific signal emerged among adults aged 60 and over, where women again faced higher estimates than men. These findings align with physiological expectations that reduced thermoregulatory reserve in older adults heightens vulnerability, and they carry direct implications for aged-care planning and community health services.

Beyond the same-day estimates, the team deployed distributed lag non-linear models to trace how heat effects unfold over the following week. The cumulative exposure-response curves were broadly U-shaped, with fitted minima around 11 to 12 degrees Celsius in the full-year models and rising cumulative odds on the warm side, most clearly supported across the upper-middle of the exposure range. The warm-season curves shifted rightward and became right-skewed, but the qualitative pattern held across metrics, lag windows of 0 to 3, 0 to 7 and 0 to 14 days, and alternative bin thresholds. Robustness checks, including a symmetric 28-day referent window, shifted bin cut points, calendar-period splits and the exclusion of districts with imputed meteorological inputs, all left the main conclusions intact, a testament to the care invested in the empirical design.

The study arrives amid unmistakable evidence that New Zealand’s climate is changing. The country’s seven-station temperature series records a 0.91 degree Celsius rise over the twentieth century, 2024 was the nation’s tenth-warmest year on record, and government projections anticipate that hot days above 25 degrees Celsius could become two to four times more frequent by 2090. The unprecedented 2017-18 austral summer heatwave, with land temperature anomalies of about 2.2 degrees Celsius and sea-surface anomalies reaching 3.7 degrees Celsius in the eastern Tasman Sea, offered a preview of this future. Previous New Zealand evidence had been fragmentary: a Christchurch time-series study from 2000, a Canterbury heatwave pilot, and an expert assessment estimating roughly 14 heat-related deaths per year among older adults in Auckland and Christchurch. The new national analysis replaces these fragments with a coherent, powered, individual-level picture.

The limitations are candidly acknowledged. Exposure was assigned at the district-day level, smoothing over within-district variation that matters in topographically diverse and coastal areas, and the WBGT indices were built from daily summaries rather than hourly data, potentially blurring the timing of peak heat stress. Non-differential exposure error of this kind most likely attenuates rather than inflates the estimated associations, meaning the true burden could be larger. Cause-of-death coding was available only through 2020 for the non-injury analyses, and the case-crossover design, centred on deaths, cannot yet model non-fatal outcomes. The authors flag an individual-day population panel as the next step. Even so, the core message stands firm: measurable heat-mortality associations exist in a temperate maritime climate, they are stronger in the warm season, sharpest among older adults and women, and steepest in the hottest districts. As New Zealand’s climate continues to warm, heat-health warning systems, seasonal preparedness and locally calibrated thresholds will no longer be optional extras but essential infrastructure.

Subject of Research: The association between short-term heat exposure, measured by wet-bulb globe temperature and air temperature, and mortality in Aotearoa New Zealand.

Article Title: Heat exposure and mortality in Aotearoa New Zealand: a time-stratified case-crossover study

Article References: Lu, S., Noy, I., & Stone, D. (2026). Heat exposure and mortality in Aotearoa New Zealand: a time-stratified case-crossover study. Regional Environmental Change, 26(3), Article 184. https://doi.org/10.1007/s10113-026-02667-y

Image Credits: AI Generated

DOI: 10.1007/s10113-026-02667-y

Keywords: heat exposure, mortality, wet-bulb globe temperature, New Zealand, case-crossover study, climate change, heat stress, public health, distributed lag non-linear models, warm season, older adults, temperature