Association between meteorological variables and diabetic ketoacidosis admissions: a single-center observational study

Abstract

Abstract Background Evidence on hot weather and acute diabetes complications comes mainly from temperate, high-income countries; hot semi-arid regions and children are underrepresented. We examined emergency department admissions for diabetic ketoacidosis (DKA) in relation to meteorological variables in a city with a hot semi-arid climate. Methods Pediatric and adult DKA admissions to Mardin Training and Research Hospital in 2025 were studied in a single-center retrospective cohort. Cases met age-appropriate diagnostic criteria (ISPAD 2022 under 18 years; 2024 consensus report for adults). The outcome was the daily number of DKA admissions over 365 days; the primary objective was its variation by season and the secondary objective its association with 14 meteorological and air quality variables. Daily counts were modeled with Poisson regression; adequacy was checked with negative binomial, zero-inflated and quasi-Poisson models, weekly aggregation and Newey-West standard errors. Results Ninety-five admissions were recorded (37 [39%] pediatric; 62 [65%] admitted to intensive care). No admission occurred on 279 of 365 days (Poisson expectation 283). The daily count differed by season ( p = 0.003); the incidence rate ratio (IRR) versus winter was 2.69 (95% CI 1.39–5.21) for summer and 2.64 (1.36–5.12) for autumn, and 2.61 (1.66–4.10) for May–October versus November–April. The warm-season excess consisted largely of 59 consecutive days (5 August to 2 October) with one admission each; with Newey-West standard errors the May–October estimate was 2.61 (1.26–5.42) at a 7-day and 2.61 (0.97–7.00) at a 30-day lag window. A one-SD (10.4 °C) rise in mean temperature gave an IRR of 1.58 (1.27–1.95), adjusted 1.36 (0.85–2.19). The association was larger in children (2.24 [1.52–3.30]) than in adults (1.30 [1.00–1.69]; interaction p = 0.024). July, the hottest month, had one admission. Severity of admitted patients did not differ by season. Conclusions DKA admissions were about 2.5 times more frequent in summer and autumn than in winter, more so in children, without greater severity. Season and temperature overlap within a single year, the adjusted temperature effect was not significant, and the warm-season excess formed one continuous run of admission days; one year of data therefore cannot separate a direct heat effect from a recurring seasonal pattern or a one-time cluster. Multi-year, multicenter data are needed. Trial registration Not applicable; the study is observational in nature.