This was a retrospective study based on data retrieved from the Medical Information Mart for Intensive Care IV (MIMIC-IV) v3.0 database, which includes critically ill patients admitted to an intensive care unit (ICU) or the emergency department of the Beth Israel Deaconess Medical Center (Boston, MA, USA) between 2008 and 2019. One of the authors passed the Collaborative Institutional Training Initiative program course required to access the database and obtained study approval from the Institutional Review Board of Beth Israel Deaconess Medical Center.

This study included adult patients admitted to the ICU after achieving return of spontaneous circulation (ROSC) following cardiac arrest (12). We included patients in the MIMIC-IV v2.2 database who were aged ≥18 years, admitted to ICU for the first time, and diagnosed with cardiac arrest at hospital admission according to diagnostic codes of the International Classification of Diseases Ninth Revision and Tenth Revision. The exclusion criteria were as follows: (I) not the first ICU admission; (II) age <18 years; (III) pregnancy or in the postpartum period; (IV) missing data for blood urea nitrogen (BUN) and serum albumin (ALB); (V) admission >24 h earlier; (VI) multiple admissions.

We used PostgreSQL version 10.13 to extract data from MIMIC-IV v2.2 regarding patients' baseline characteristics, including age, sex, comorbidities, and laboratory test results.

We extracted information from the first document containing data for vital signs and laboratory test data for patients with cardiac arrest who were admitted to the hospital. Demographic data included age, sex, and race. Vital signs data included body temperature (T), mean blood pressure (MBP), heart rate (HRT), respiratory rate (RR), and pulse oximetry-derived oxygen saturation (SPO₂). Comorbid conditions included the Charlson Comorbidity Index (CCI), myocardial infarct, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, diabetes, paraplegia, renal disease, malignant cancer, liver disease, metastatic solid tumor, and AIDS. Laboratory test results within the first 24 h were extracted from the time of admission to the ICU and included white blood cell count, hemoglobin (Hb), platelet count (PLT), alkaline phosphatase (ALP), alanine aminotransferase, aspartate aminotransferase, total bilirubin (TBIL), creatinine (Cr), glucose, sodium, potassium (K⁺), chloride (CI⁻), calcium, lactate (Lac), pH, partial pressure of carbon dioxide, partial pressure of oxygen/fraction of inspired oxygen, base excess (BE), prothrombin time (PT), and partial thromboplastin time (PTT). We also recorded scoring systems applied in the ICU including the Sequential Organ Failure Assessment (SOFA) score and Simplified Acute Physiology Score II (SAPS II). Therapy included the use of vasoactive drugs, mechanical ventilation, and renal replacement therapy (RRT) during hospitalization. The length of stay in the ICU and 28-day all-cause mortality were also recorded.

To minimize bias from missing data imputation, variables with ≥25% missing values were excluded; the remaining missing values were handled using multiple imputation (13). The number of imputations was five. Continuous variables were imputed using linear regression models through an iterative computational process to ensure stability. The final results were obtained by pooling estimates across multiply imputed datasets, providing unbiased statistical inference under the missing-at-random (MAR) assumption.

Continuous variables that exhibited a normal distribution are expressed as mean ± standard deviation; otherwise, these are expressed as median (interquartile range). Categorical variables are reported as frequency and percentage. To explore the association between different BAR levels and 28-day all-cause mortality, enrolled patients were categorized into three groups according to the tertile distribution of BAR values within the study cohort (with the 33rd and 66th percentiles used as cutoff points).

We used the t-test for continuous variables with a normal distribution, the Mann–Whitney U test for continuous variables with a skewed distribution, and the chi-squared test for categorical variables to analyze differences between the two groups.

Univariable and multivariable Cox regression analyses were performed to examine the relevance between an increased serum BAR and risk of 28-day all-cause mortality in the ICU. Restricted cubic splines analysis was applied to detect the association between the BAR and risk of 28-day all-cause mortality. Kaplan–Meier analysis was conducted to compare the cumulative survival rate between the high and low BAR groups using the log-rank test. The survival rates among different groups were then compared using log-rank tests.

Variables that were considered potential confounders based on the existing literature and clinical judgment were included. Five regression models were built, and the results are presented as hazard ratio (HR) with 95% confidence interval (CI). Model 1 was unadjusted. Model 2 adjusted for age and sex. Model 3 further adjusted for the Charlson Comorbidity Index, myocardial infarct, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, diabetes, paraplegia, renal disease, malignant cancer, liver disease, metastatic solid tumor, and AIDS, based on Model 2. Model 4 further adjusted for white blood cell count, Hb, platelet count, ALP, alanine aminotransferase, aspartate aminotransferase, TBIL, Cr, glucose, sodium, K⁺, CI⁻, calcium, Lac, pH, partial pressure of carbon dioxide, partial pressure of oxygen/fraction of inspired oxygen, BE, PT, partial thromboplastin time, vasoactive drug use, RRT, mechanical ventilation, and length of ICU stay, based on Model 3. Model 5 further adjusted for HRT, MBP, RR, T, SPO₂, SAPS II, and SOFA, based on Model 4.

Subgroup analyses were performed according to age, sex, renal disease, liver disease, vasoactive drugs, rrt, ventilation, race, aids, malignant cancer, paraplegia, diabetes, peptic ulcer disease, rheumatic disease, chronic pulmonary disease, cerebrovascular disease, peripheral vascular disease, congestive heart failure and myocardial infarct. Data analyses were completed using Stata version 17.0 and IBM SPSS version 26 (IBM, Armonk, NY, USA). Statistical significance was defined as a two-tailed P-value <0.05.

A total of 793 patients were ultimately enrolled in the study (Figure 1). The enrolled patients were divided into three groups according to tertiles of the BAR, the Q1, Q2, and Q3 groups. The mean age of study participants was 65.1 ± 16.9 years, 498 (62.8%) participants were men, and the 28-day all-cause mortality rate among patients with cardiac arrest was 51.6% (409/793). Patients in the group with a higher BAR(Q3) were older, predominantly men, and had higher values for SAPS II, SOFA, the Charlson Comorbidity Index, ALP, TBIL, Cr, BUN, K⁺, Lac, BE, and PT, as well as a higher prevalence of renal disease, congestive heart failure, peripheral vascular disease, diabetes, renal disease, and liver disease. Additionally, the use of vasoactive drugs and RRT was more common. With increased BAR, the values for MBP, SPO₂, CI⁻, PO₂/FiO₂, Hb, and ALB showed a tendency to decrease. The 28-day all-cause mortality showed a significant incremental trend between BAR groups. The difference in mortality rates between the three groups (Q1, Q2, Q3) was statistically significant (37.5% vs. 53.4% vs. 63.8%, respectively; P < 0.001), suggesting that the BAR is associated with the risk of 28-day all-cause mortality (Table 1).

When the BAR was treated as a continuous variable, for each 1-unit increase in the BAR, the risk of mortality increased by 1% (HR = 1.010, 95% CI: 1.010–1.020, P < 0.001). Patients were divided into two equal groups (Q1, Q2) according to BAR value; the HR (Q1 for reference) was 1.79 (95% CI: 1.470–2.180, P < 0.001). Moreover, when patients were divided into three groups according to BAR value (Q1, Q2, Q3), the HR (Q2 vs. Q1) was 1.6 (95% CI: 1.240–2.070, P < 0.001) and further increased (Q3 vs. Q1) to 2.07 (95% CI: 1.610–2.650, P < 0.001). This indicates that the higher the BAR, the higher the risk of 28-day all-cause mortality among patients (Table 2).

The effect of the BAR on 28-day all-cause mortality was assessed using a Cox proportional hazard model. When the BAR was treated as a continuous variable, it was positively associated with 28-day all-cause mortality. As shown in Table 3, in the unadjusted Model 1, each unit increase in the BAR was associated with a 1% increase in the risk of 28-day all-cause mortality (HR = 1.010, 95% CI: 1.010–1.020, P < 0.001). In Model 2 (adjusted for age and sex), the results remained significant (P < 0.001). In Model 3 (Model 2, further adjusted for myocardial infarct, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, diabetes, paraplegia, renal disease, malignant cancer, liver disease, metastatic solid tumor, and AIDS), the results remained significant (P < 0.001). In Model 4 (Model 3, further adjusted for white blood cell count, Hb, platelet count, ALP, alanine aminotransferase, aspartate aminotransferase, TBIL, Cr, glucose, sodium, K⁺, CI⁻, Lac, pH, partial pressure of carbon dioxide, partial pressure of oxygen/fraction of inspired oxygen, BE, PT, partial thromboplastin time, vasoactive drug use, RRT, mechanical ventilation, and length of ICU stay), the results remained significant (P < 0.001). In Model 5 (Model 4, further adjusted for HRT, MBP, RR, T, SPO₂, SAPS II, and SOFA), the results showed significant differences (HR = 1.010, 95% CI: 1.000–1.020, P = 0.003).

Patients were divided into equal groups based on BAR values (Q1, Q2), where the Q1 group was used as the reference group; the results remained significant in the different models.

To further explore the effect of the BAR, patients were divided into three groups according to BAR values (Q1, Q2, Q3), with the Q1 group used as the reference. The results of Model 1 showed that higher initial admission in patients was associated with an increased 28-day all-cause mortality, with a gradual increase in the HR starting from Q2 (HR = 1.600, P < 0.001) to Q3 (HR = 2.070, P < 0.001). Overall, the adjusted Models 2, 3, 4, and 5 showed consistency with Model 1.

To further investigate the impact of other risk factors on the correlation of the BAR with 28-day all-cause mortality, subgroup analyses were carried out using the following variables in stratification: age, sex, renal disease, liver disease, vasoactive drugs, rrt, ventilation, race, aids, malignant cancer, paraplegia, diabetes, peptic ulcer disease, rheumatic disease, chronic pulmonary disease, cerebrovascular disease, peripheral vascular disease, congestive heart failure and myocardial infarct. Additive interactions between the BAR and 28-day all-cause mortality were observed for rrt and paraplegia (P for interaction <0.05). However, significant interactions were not found for age, sex, renal disease, liver disease, vasoactive drugs, ventilation, race, aids, malignant cancer, diabetes, peptic ulcer disease, rheumatic disease, chronic pulmonary disease, cerebrovascular disease, peripheral vascular disease, congestive heart failure and myocardial infarct.

In the sensitivity analyses, among patients aged ≤65 years, each one-unit increase in the BAR was positively associated with mortality risk, although this relationship did not reach statistical significance (HR = 1.013, p = 0.228). In contrast, among patients aged >65 years, the association was significant (HR = 1.017, p < 0.05). A significant positive correlation between BAR and mortality was observed in male patients, whereas no such association was found in female patients. In patients without liver disease, AIDS, malignant cancer, Diabetes, Peptic ulcer disease, Cerebrovascular disease, Peripheral vascular disease, Myocardial infarct and in those not receiving vasoactive drugs—higher BAR correlated significantly with increased mortality (p < 0.05). Similarly, patients undergoing mechanical ventilation and those with congestive heart failure exhibited a significant increase in mortality risk per unit increase in BAR (p < 0.05); In the non-RRT group, each one-unit increase in BAR was associated with a significant rise in mortality risk, whereas in the RRT group, each one-unit increase in BAR was associated with a significant reduced in mortality risk both subgroup differences reached statistical significance (p < 0.05). The findings in the paraplegia and rheumatic disease subgroups were also consistent with those observed in the RRT subgroup. In the Chronic pulmonary disease subgroup analysis, BAR was significantly associated with increased mortality risk in both patients without Chronic pulmonary disease (HR = 1.014, p < 0.05) and those with Chronic pulmonary disease (HR = 1.034, p < 0.05) (Table 4).

After adjusting for the covariates in Model 5, the results of restricted cubic spline analysis showed a nonlinear relationship between the BAR and 28-day all-cause mortality (Figure 2). As shown in Table 5, with BAR ≤ 17.981, the 28-day all-cause mortality risk increased by 5.7% (95% CI: 1.012–1.105, P < 0.05) for every 1-unit increase in the BAR. With BAR > 17.981, there was a 1.7% (95% CI: 0.994–1.041, P > 0.05) increase in the 28-day all-cause mortality risk for every 1-unit increase in the BAR; although there was an upward trend, the data did not reach statistical significance.

The cumulative mortality rate was compared using Kaplan–Meier curves. Following restricted cubic splines analysis, patients were divided into tertiles by BAR values (Q1, Q2, Q3). The results showed that a higher BAR (Q3) was significantly associated with an enhanced risk of 28-day all-cause mortality, compared with other BAR groups (Figure 3).