I’ll admit I opened this paper with the conclusion already half-written in my head: a hotter world, faster-breeding bacteria, more sick toddlers. Then I hit the number at its center. The model pins 1.7 million cases of diarrhea in Ghanaian children under five on “non-optimal” temperature and rainfall, and right beside that number sits a risk curve that climbs to a perfectly ordinary 26.2°C and then bends back down as the months get hotter.
The study, published October 27 in Environmental Health, took monthly under-five diarrhea counts from routine surveillance in 261 districts between 2013 and 2024. The team matched them to temperature and rainfall from the ERA5 climate reanalysis and fed both into a spatiotemporal Bayesian model that lets each exposure act with a delay, two months for heat and three for rain. Etse Yawo Dzakpa of the African Institute for Mathematical Sciences in Kigali is first author, with co-authors at the University of Ghana, Imperial College London, the University of Tokyo and the UK Office for National Statistics.
Over those twelve years, the model credits 916,176 cases to temperature and 927,352 to rainfall, 5.5 percent of cases each, and 1,667,998 to the two jointly. The joint figure is smaller than the first two added together (about 1.84 million), so it isn’t a simple sum. The paper reports it as a single combined count with a credible interval running from about 1 million to 2.3 million. The heat burden clusters in the far north and the rain burden in the wetter south, and both peak from April to July.
So what counts as “optimal” weather for a toddler’s gut? I looked for that definition before anything else, because every one of those cases is a comparison. The paper measures its burden “against a reference selected by a pre-specified rule that resolved to the 95th percentile.” Risk peaked at 26.2°C and at 60 mm of monthly rain, then declined at higher values of both. As I read it, that puts the model’s low-risk benchmark up toward the hot, wet end of the record, past the point where its own curve has already turned down. “Attributable” here means the model’s estimated excess over that benchmark, accumulated across twelve years. It isn’t a count of identifiable children whom the weather made sick.
The excess at any single point is modest, too. At the temperature peak the relative risk was 1.079 (credible interval 1.031 to 1.130), and at the rainfall peak it was 1.082 (1.035 to 1.132). That’s roughly 8 percent above the benchmark at the very worst point of each curve. Spread across twelve years and 261 districts, a bump that size adds up to a seven-figure count.
| Result | Value (RR) | Lower bound | Upper bound |
|---|---|---|---|
| Temperature peak | 1.079 | 1.031 | 1.130 |
| Rainfall peak | 1.082 | 1.035 | 1.132 |
None of this is climate change in the sense most readers hear in that phrase. The paper’s first sentence calls climate change “a major threat to global public health,” and its conclusion offers the method as “a basis for prioritising adaptation planning.” The results it reports, though, describe month-to-month weather inside Ghana’s observed record. They don’t project future warming, and they don’t separate human-caused trends from ordinary variability. Within the range the model studied, the hottest months carried less risk than the 26.2°C peak, so the curve gives no support to the headline that a warmer Ghana means sicker children. (It can’t promise the reverse either, since a curve fit to past months says little about temperatures Ghana hasn’t had yet.)
Then why would risk crest at moderate rain and ease off in wetter months? I love this half of the puzzle, because the leading explanation is so physical. A 2020 meta-analysis in Environmental Health Perspectives by Alicia Kraay and colleagues, drawing on 111 studies, tested what researchers call the concentration-dilution hypothesis. During a dry spell, feces and pathogens build up in the environment, and the first heavy rain flushes that load into the streams, wells and ponds people drink from. Extreme rain after a dry period went with 26 percent more diarrhea (incidence rate ratio 1.26). After a wet period the estimate leaned the other way, at 0.911, though not significantly, which fits the idea that water already flushed now dilutes whatever is left. The pathogens split by season as well: bacterial diarrhea ran 2.70 times higher in rainy seasons, while rotavirus was more common in dry ones.
That evidence concerns extreme rain events rather than monthly totals, so it doesn’t map neatly onto Ghana’s 60 mm peak. It does give the shape a plausible physical explanation, in which whatever sits on the ground when the rain starts ends up in a child’s gut weeks later. Under that mechanism the weather sets the timing, but the contamination has to be there first.
Each exposure on its own accounts for 5.5 percent of cases. That doesn’t mean the other 94.5 percent has some identified cause. An attributable fraction tells you how many cases the model would remove by moving the weather to its benchmark, and nothing about what drives the rest. This model wasn’t built to answer that question, but two of its own authors have already studied it on the ground. A 2024 study in the Journal of Health, Population and Nutrition, with Duah Dwomoh and Mawuli Dzodzomenyo among its authors, surveyed caregivers in five coastal communities and found that 36 percent of children under five had diarrhea in the previous two weeks. The authors called that “high inspite of the reported use of improved water sources and sanitation facilities by majority of households.” National prevalence, according to the 2022 Ghana Demographic and Health Survey they cite, was about 13 percent.
The same paper cites Ghana Health Service figures showing outpatient diarrhea cases recorded in health facilities rising from 287,816 in 2012 to 1,429,990 in 2017. The source doesn’t say what ages those counts cover, so I can’t line them up against the new study’s under-five series. Still, a nearly fivefold jump in what clinics write down raises an obvious question for any model built on routine surveillance: how much of the rise is sicker children, and how much is better counting? The model’s correlation of r = 0.915 shows that it reproduces the recorded counts closely. It doesn’t show that a recorded case meant the same thing in 2013 as in 2024.
Then there’s the question of who built the measuring stick. The study “applies the Standards for Official Statistics on Climate–Health Interactions (SOSCHI) framework,” a 2022 to 2026 project led by the UK Office for National Statistics and funded by Wellcome. It is aimed at producers of official statistics, “especially in low- and middle-income countries,” and waterborne disease heads its list of six priority topics. The UN Statistical Commission endorsed SOSCHI’s proposed headline indicators in March 2026, the final framework came out in June, and the project now continues on a smaller scale, engaging international organizations and other countries. One co-author sits on the ONS Climate and Health Team, and the paper is funded by Wellcome grant 224682/Z/21/Z. Its conclusion calls SOSCHI “a standardised, scalable framework” and argues for “embedding climate–health indicators into official health statistics.”
The authors declare no competing interests, and I take that statement as given. My worry is structural. When an indicator designed in London, funded by Wellcome and labeled “climate” becomes the number a health ministry reports upward, it shapes where attention goes. A district figure for “climate-attributable diarrhea” reads as an argument for adaptation money. Adaptation can include water and sanitation, sure, but the label points at the sky instead of at what’s already sitting on the ground. A project whose deliverable is the climate-health indicator also has an institutional interest in that indicator looking important. That’s an incentive, not evidence that these estimates are biased, and it deserves to be said out loud before other countries adopt the framework.
The map itself is useful. Monthly, district-level resolution across all of Ghana is hard to build, and the north-south split, together with the April-to-July window, tells a district health officer when and where the season will hit. If I ran a clinic in southern Ghana, I’d use it to stock oral rehydration salts and push household water treatment before the first rains. If someone handed me the budget, I wouldn’t spend the first cedi on another climate indicator. I’d spend it finding out why toddlers in households with “improved” water and toilets are still getting sick.