IntroductionGlobal warming is progressing, the incidence of heat-related illnesses has been reported to be increasing yearly, estimated 500,000 additional deaths worldwide each year1,2. Heat-related illnesses encompass a continuum that includes heat edema, heat syncope, heat cramps, heat exhaustion, and the most severe form, heatstroke3. Clinically, heatstroke is characterized by central nervous system dysfunction, multiorgan failure, and extreme hyperthermia (usually > 40.5 °C)4.Heatstroke can be classified as either classic or exertional, depending on its cause. Both types involve an imbalance between the body’s heat production and heat dissipation, though their underlying mechanisms differ5. Classic heatstroke (CHS) results from passive exposure to environmental heat and inadequate heat-dissipation mechanisms. In contrast, exertional heatstroke (EHS) occurs due to exposure to a hot environment during physical exercise, resulting when excessive metabolic heat production overwhelms the body’s physiological heat-loss mechanisms6. EHS affects mainly athletes, military personnel, firefighters, and occupational workers. For CHS, older adults are particularly vulnerable, especially those with common age-associated chronic health conditions (e.g., cardiovascular disease, hypertension, obesity, type 2 diabetes, chronic kidney disease)7. It is a life-threatening condition ultimately progressing to life-threatening multiple organ failure and associated with a reported 28-day mortality rate of up to 58%8. Currently, the primary treatment modalities for heatstroke include hypothermia control, rehydration therapy, and hemodialysis. While significant research has been conducted on symptomatic therapies for heatstroke, most of these approaches remain at various preclinical stages3. Therefore, management of heatstroke primarily focuses on prevention and early intervention to prevent the progression of the disease.Rhabdomyolysis and heat-induced inflammatory damage both significantly elevate the risk of acute kidney injury (AKI) in heatstroke patients3. Previous studies have demonstrated that heatstroke complicated by AKI is associated with higher hospitalization costs and worse clinical outcomes9. Despite this urgency, AKI typically manifests in the later stages of the disease, and there remains a lack of studies specifically addressing the early prediction of AKI in heatstroke patients. To bridge this gap, this study introduces machine learning models designed to predict AKI incidence in heatstroke patients using clinical data obtained during the first 24 h of hospitalization.MethodsPatients and study designData were collected from 55 hospitals in China between 2008 and 2024. After applying the inclusion and exclusion criteria, a total of 290 patients with heatstroke were enrolled in the study (Fig. 1)10. The inclusion criteria for this study were as follows: (1) a history of exposure to high-temperature environments and/or participation in high-intensity manual labor; (2) an axillary temperature above 39 °C; (3) evidence of central nervous system dysfunction, including symptoms such as delirium, coma, impaired consciousness, or disorientation5,6; (4) patients with a hospital stay of more than 24 h.Fig. 1Flowchart of patient enrollment and group allocation. A total of 511 patients were enrolled from 55 hospitals between 2008 and 2024. After excluding 97 patients under 18 years old, 89 patients with pre-existing comorbidities prior to heatstroke onset, and 35 patients with more than 30% missing data, 290 patients were included in the study. These included patients were further divided into two groups: CHS and EHS.Full size imageThe exclusion criteria were as follows: (1) patients aged under 18 years; (2) patients with pre-existing comorbidities prior to heatstroke onset, including diabetes, cerebral infarction, pulmonary infection, chronic kidney disease, and dementia; and (3) patients with more than 30% missing data in their records were excluded from the analysis.Ethical considerationsThe study was conducted by the PLA General Hospital and received approval from the ethics committees of all participating institutions. Each patient underwent comprehensive, condition-specific treatment, which included body cooling, fluid administration, and anti-inflammatory measures. For those diagnosed with rhabdomyolysis and AKI, organ support was provided as needed, in accordance with clinical guidelines. This support included appropriate hydration, urine alkalization, and, when necessary, continuous renal replacement therapy (CRRT), along with other interventions9.Definitions1.AKI was defined according to the Kidney Disease: Improving Global Outcomes (KDIGO) criteria as one of the following: (1) an increase in serum creatinine (Scr) to ≥ 26.5 μmol/L (≥ 0.3 mg/dL) within 48 h; (2) an increase in Scr to ≥ 1.5 times the baseline within 7 days; (3) urine output 1000 IU/L or CK > 5 × upper limit of normal for the standard definition of rhabdomyolysis. Additionally measured myoglobinuria and AKI indicate a severe type of rhabdomyolysis12,13.3.Sequential organ failure assessment (SOFA) score: a validated tool used to quantify the extent of organ dysfunction in critically ill patients. It evaluates six organ systems—respiratory, cardiovascular, hepatic, coagulation, renal, and neurological—each assigned a score ranging from 0 (normal function) to 4 (severe dysfunction). The total SOFA score, ranging from 0 to 24, reflects the overall severity of organ failure, with higher scores associated with increased mortality (Supplementary Table 1)14.4.Disseminated intravascular coagulation (DIC): Diagnosed based on a combination of clinical manifestations and laboratory findings that reflect systemic activation of the coagulation cascade. The International Society on Thrombosis and Haemostasis (ISTH) has proposed a widely accepted scoring system to identify overt DIC. The scoring system incorporates four parameters: platelet count, prolongation of prothrombin time (PT), levels of fibrin-related markers (such as D-dimer or fibrin degradation products), and fibrinogen concentration (Fib). Each parameter is assigned a score, and a cumulative score of ≥ 5 is considered indicative of overt DIC (Supplementary Table 2)15.5.Effective cooling: Defined as the reduction of core body temperature to below 38.5 °C within 30–60 min of initiating treatment. This threshold is widely accepted to prevent irreversible neurological injury and multi-organ dysfunction. Commonly employed cooling strategies include cold-water immersion, evaporative cooling, ice blanket therapy, intravascular temperature management, and extracorporeal methods such as cold hemodialysis or high-flow continuous hemodiafiltration5,16.Statistical analysisAll statistical analyses were performed using RStudio (version 2024.12.1) running R version 4.4.2 (R Core Team, 2024, https://www.r-project.org/). Prior to any analysis, data preprocessing was carried out to handle missing values and ensure data quality. Variables with more than 30% missing values were excluded from further analysis. For variables with less than 30% missingness, multiple imputation was performed using the “mice” package (version 3.18.0) to reduce bias and maximize statistical power17,18,19.Continuous variables were first assessed for normality using the Shapiro–Wilk test. Normally distributed variables are reported as means with standard deviations (mean ± SD), while non-normally distributed variables are expressed as medians with interquartile ranges (median [Q1, Q3]). Categorical variables are presented as frequencies and percentages. For group comparisons, the two-independent-samples t-test was applied to normally distributed continuous variables, and the Mann–Whitney U test was used for non-normally distributed data. The chi-square (χ2) test was employed to assess associations between categorical variables. A two-tailed p value