Temperature was linked to 3.88% of Puerto Rico's deaths — and cold did more than heat

Most of what is known about temperature and death comes from large cities in wealthy, temperate countries. A study published this month in Environmental Research Letters fills in a gap that its authors describe plainly: robust epidemiological evidence on the health effects of high temperatures — and of low ones — remains scarce in Caribbean small island developing states, particularly outside urban settlements [s1].

The design

The researchers ran a case time-series study of municipality-level mortality and temperature across Puerto Rico from 2015 to 2023 [s1]. Daily death counts were modelled against mean temperature using conditional quasi-Poisson regression combined with a distributed lag non-linear model over a 21-day lag, adjusting for relative humidity, seasonality and day of the week [s1].

The 21-day lag matters. Deaths associated with a cold or hot spell do not all occur on the day of exposure, and a model that only looks at same-day mortality will miss most of a cold effect in particular.

From that model the team estimated the minimum mortality temperature — the point on the temperature curve associated with the lowest death risk — and then measured relative risk at the extremes, defined as the 2.5th and 97.5th percentiles of the daily temperature distribution [s1].

The results

Extreme low temperature was associated with 1.23 times the risk of all-cause mortality (95% CI 1.07 to 1.40) [s1]. Extreme high temperature was associated with 1.16 times the risk (95% CI 1.05 to 1.27) [s1]. Both intervals exclude 1.0, so both effects are statistically distinguishable from no effect in this dataset.

Translated into burden: temperature-related mortality accounted for 3.88% of the island's 280,568 deaths over the study period (95% empirical CI 3.39% to 4.29%) [s1]. Low temperatures contributed 2.02% (95% eCI 1.69% to 2.32%) and high temperatures 1.86% (95% eCI 1.35% to 2.35%) [s1].

Two things about that split are worth stating carefully. First, "extreme cold" in Puerto Rico is not extreme cold anywhere else — it is the coldest 2.5% of days on a tropical island, and the finding says nothing about absolute temperatures. Second, the two confidence intervals overlap substantially, so the data support a claim that both ends of the range matter more than a claim that cold definitively outweighs heat.

The study also reports substantial spatial variability in temperature-related mortality burden across municipalities [s1], and identifies which municipalities are most vulnerable — the practical output the authors point to as a basis for municipality-specific adaptation planning [s1].

Why the location is the point

A commentary published in Nature Climate Change this month makes the structural argument that this Puerto Rico study is an answer to. Over the past decade, the authors write, attribution science has established that climate change is responsible for substantial death, disability and illness — but health impact attribution studies have focused disproportionately on populations in high-income countries, and have mostly quantified the health outcomes of heat and extreme weather [s2].

That is two biases stacked: a geographic one and an outcome one. Places outside high-income settings are underrepresented, and within the studies that exist, the exposures examined skew toward heat and acute weather events rather than the full range of climate-sensitive health outcomes.

The commentary's stated stake is political rather than methodological — a clearer picture of the global burden of climate change, the authors argue, could encourage policymakers to treat the climate crisis as a public health emergency [s2].

Limits

The Puerto Rico analysis covers 2015 to 2023, a nine-year window. It is an association study: it establishes that mortality risk rises at both temperature extremes after adjustment for humidity, season and day of week, not that temperature caused any particular death. Nor does it attribute any portion of the burden to human-caused climate change — that is a separate calculation requiring a counterfactual climate, and this study does not attempt one.

The 3.88% figure is also a fraction of all deaths in the study population, not a projection. It describes what the model estimates happened between 2015 and 2023 under the temperatures that occurred.

What to watch

Whether similar case time-series analyses appear for other Caribbean and Pacific island states is the thing to track. The method used here — municipality-level mortality records paired with local temperature — is portable wherever civil registration is complete enough, and the Nature Climate Change commentary's complaint is precisely that this work has not been done in most of the world [s2].

Sources

• Mortality risk and burden associated with non-optimum temperatures in Puerto Rico — Environmental Research Letters, 2025-09-09
• Health losses attributed to anthropogenic climate change — Nature Climate Change, 2025-09-17