In this cohort study, data of a total of 2,312 ICH patients aged ≥18 years were identified from MIMIC-IV and extracted. The Medical Information Mart for Intensive Care (MIMIC) database was established in 2003 through funding from Beth Israel Deaconess Medical Center (BIDMC), Massachusetts Institute of Health, the National Institutes of Health Technology, Massachusetts General Hospital, emergency room physicians, critical care physicians, computer science experts, and other professionals in the field of critical care medicine. The MIMIC is recognized as the largest open-source and freely accessible clinical database specifically designed for research on critical care and emergency departments. MIMIC-IV encompassed data collected between 2008 and 2019 (19). Patient records were extracted from the MIMIC-IV database for this study if they met the following criteria: aged ≥18 years old, diagnosed with ICH at ICU admission, and hospitalized in the ICU for at least 24 h. Patients without data on serum osmolality or BCR calculations, those without survival information, or those without data on GCS score were excluded. The requirement of ethical approval for this study was waived by the Institutional Review Board of School of Basic Medicine, Anhui Medical University, because the data were accessed from MIMIC-IV. The need for written informed consent was waived by the Institutional Review Board of School of Basic Medicine, Anhui Medical University, due to the retrospective nature of the study.

Demographic characteristics including age (<65 year or ≥ 65 years); gender (women or men); race (white, Black, other, or unknown); insurance type (Medicare, Medicaid, or other); marital status (married, single, divorced, widowed, or unknown); complications including hypertension, neurological deterioration, acute kidney injury (AKI), and sepsis; Simplified Acute Physiology Score II (SAPSII); Sequential Organ Failure Assessment (SOFA) score; Glasgow Coma Scale (GCS) score (<13 or ≥ 13); Charlson Comorbidity Index (CCI); ICU type [Neuro ICU, surgical Intensive Care Unit (SICU), or other]; treatments including receiving mechanical ventilation, vasopressor, diuretics, mannitol, brain surgery, or cerebral drainage or not; and laboratory data including WBC (K/uL), platelet (K/uL), red cell distribution width (RDW) (%), hemoglobin (g/dL), hematocrit (%), bicarbonate (mEq/L), estimated glomerular filtration rate (eGFR) (mL/min/1.73m²); International Normalized Ratio (INR), calcium (mg/dL), chloride (mEq/L), mean arterial pressure (MAP) (mmHg), heart rate (bpm), respiratory rate (insp/min), temperature (°C), oxyhemoglobin saturation (SpO₂) (%), prothrombin time (PT) (sec), and urine output-24 h (mL) were analyzed as potential covariates.

Acute kidney injury (AKI) was determined based on the Kidney Disease: Improving Global Outcomes (KDIGO) definition. According to this definition, AKI was diagnosed as an increase of 0.3 mg/dL in the serum creatinine (SCr) level within 48 h or an increase in SCr level to 1.5 times of the baseline value, which is known or presumed to have occurred within the previous 7 days, or a urine volume of <0.5 mL/kg/h for 6 h. Sepsis was diagnosed according to the Sepsis-3 criteria (20). Patients with documented or suspected infection and an acute change in total SOFA score of ≥ 2 points were reported to have sepsis. Brain surgery included craniotomy (ICD-9 codes: 0120–0129) or minimally invasive surgery (ICD-9 codes: 0221, 0222, 0139, 0101, 0102, and 0109). Neurological deterioration was defined as a decline in the GCS score by ≥2 points during the ICU stay (21). eGFR _CKD-EPI (mL/min/1.73m²) = 141 × min (Scr/κ, 1) α × max (Scr/κ, 1) -1.029 × 0.993 age × 1.108 (if woman) × 1.159 (if Black), where κ is 0.7 for women and 0.9 for men, α is −0.329 for women and − 0.411 for men, min indicates the minimum of Scr/κ or 1, and max indicates the maximum of Scr/κ. A GCS score of <13 indicated moderate or severe damage, while a GCS score of ≥13 indicated mild damage (22).

Serum osmolality and BCR were the main variables. According to the ESPEN-recommended equation, serum osmolality is calculated using the following formula:

Serum osmolality = 1.86× (sodium + potassium) + 1.15 × glucose + BUN+14 (all in mmol/L).

A serum osmolality of ≥295 mmol/L indicated dehydration (12, 23, 24). BCR was calculated as the BUN-to-Cr ratio, and a BCR of >20 indicated dehydration (12, 25).

The in-hospital mortality was the outcome. For participants who survived, the follow-up ended when they were discharged from the hospital. For participants who died during the hospital stay, the follow-up ended at the time of death. This information was recorded by the hospital departments or the Social Security Bureau.

The normality of quantitative data was tested by skewness and kurtosis methods, and the homogeneity of variance was tested by Levene’s test. Normal distribution measurement data were described as mean and standard deviation [mean (±SD)]. Student’s t-test (indicated hereafter as t-test) was used for comparison between groups with the assumption of homogeneity of variance, and Satterthwaite’s t-test (indicated hereafter as t’-test) was used for heterogeneity of variance. Non-normally distributed quantitative data were presented as median and quartiles [M (Q₁, Q₃)], and the Wilcoxon rank-sum test was used for between-group comparison. Categorical data were presented by the number and percentage of cases n (%), and the chi-square test or Fisher’s exact test was used for within-group comparison. Missing values were dealt with multiple imputation (Supplementary Table 1). Sensitivity analysis was performed to compare the data before and after imputation, and no significant difference was found (Supplementary Table 2). Serum osmolality and BCR were the main variables used to analyze their association with the risk of in-hospital mortality in ICH patients after first ICU admission using the univariable Cox regression model. Potential covariables with a p-value of <0.05 in the univariable Cox regression analysis were considered as confounders to be adjusted. Univariable and multivariable Cox regression analyses were applied to examine the association of serum osmolality and BCR with the in-hospital mortality of ICH patients. Subgroup analyses were performed according to age, sex, ethnicity, GCS score, and AKI. Hazard ratio (HR) and 95% confidence intervals (CIs) were calculated. R (version 4.2.3) was used for missing value interpolation, covariate screening, Cox regression analysis, and subgroup analysis.

In total, data of 2,312 ICH patients aged ≥18 years were collected from MIMIC-IV. Among them, patients who were hospitalized in the ICU for less than 24 h were excluded (n = 388). Those without data on serum osmolality or BCR calculations (n = 83), without survival information (n = 1), and subjects without data on GCS (n = 3) were excluded. Finally, 1,837 participants were included, consisting of 1,427 survivors and 410 deaths at the end of the follow-up. The screening process is shown in Figure 1.

The median survival time of all participants was 8.29 (4.61–15.24) days. The median serum osmolality in the death group was higher than that in the survival group (298.63 mmol/L vs. 294.85 mmol/L). The percentage of participants with serum osmolality of ≥295 mmol/L in the death group was higher than that in the survival group (64.15% vs. 49.05). The median BCR in the death group was higher than that in the survival group (18.69 vs. 17.50). The percentage of ICH patients with BCR of >20 in the death group was higher than that in the survival group (40.24% vs. 31.39%). More information on the characteristics of survived or dead participants is shown in Table 1.

As exhibited in Table 2, age, race, insurance type, marital status, MAP, respiratory rate, urine output-24 h, AKI, SAPSII, SOFA score, GCS score, CCI, ICU type, ventilation, vasopressor, diuretic, mannitol, brain surgery, platelet, RDW, hemoglobin, hematocrit, bicarbonate, and eGFR were potential covariates associated with in-hospital mortality of ICH patients. No significant multicollinearity was observed among these variables, as the generalized variance inflation factors (GVIFs) were all <10 (Supplementary Table 3).

In the univariable Cox regression model, compared to those with serum osmolality of <295 mmol/L, the risk of in-hospital mortality might be increased in those with serum osmolality of ≥295 mmol/L (HR = 1.69, 95%CI: 1.38–2.07). After adjusting for confounding factors, serum osmolality of ≥295 mmol/L was correlated with an increased risk of in-hospital mortality in patients with ICH (HR = 1.43, 95%CI: 1.14–1.78). BCR of >20 might be a risk factor for in-hospital mortality in patients with ICH in the unadjusted model. However, BCR of >20 was not significantly associated with the risk of in-hospital mortality in patients with ICH (Table 3).

In both participants aged <65 (HR = 1.74, 95%CI: 1.17–2.61) and ≥ 65 (HR = 1.32, 95%CI: 1.00–1.74) years, serum osmolality of ≥295 mmol/L was related to an elevated risk of in-hospital mortality in ICH patients. The increased risk of in-hospital mortality was observed in female ICH patients with serum osmolality of ≥295 mmol/L (HR = 1.64, 95%CI: 1.19–2.27). Serum osmolality of ≥295 mmol/L was correlated with an increased risk of in-hospital mortality of ICH patients who belonged to white (HR = 1.44, 95%CI: 1.06–1.96) or Black (HR = 2.84, 95%CI: 1.05–7.71) race. In ICH patients with a GCS score of ≤13, the elevated risk of in-hospital mortality was found to be associated with serum osmolality of ≥295 mmol/L (HR = 2.14, 95%CI: 1.45–3.15). In ICH patients with AKI complications, we found a heightened risk of in-hospital mortality associated with serum osmolality ≥295 mmol/L (HR = 1.44, 95%CI: 1.11–1.87) (Figure 2).