Study design and population
In this cohort study, the data of 3059 participants with cerebral infarction were collected from MIMIC database, including 1568 in the MIMIC-III and 1491 in the MIMIC-IV. The MIMIC-III is a large, single-center open database comprising the electronic health records including demographic characteristics, monitoring vital signs, laboratory and microbiological examination, imaging examination, observation and recording of intake and output, drug treatment, length of stay, survival data, and discharge or death records of more than 60,000 individuals admitted to an ICU at the Beth Israel Deaconess Medical Center between 2001 and 2012 (Johnson et al. 2016). The MIMIC-IV database is an updated version of the MIMIC-III and improvements have been made including simplifying the structure, adding new data elements, and improving the usability of previous data elements. The MIMIC-IV involves the comprehensive and high-quality electronic health records of patients admitted to the ICU or emergency department of the Beth Israel Deaconess Medical Center from 2008 to 2019 (Tao et al. 2021). The database got the approval from the institutional review boards of the Massachusetts Institute of Technology (Cambridge, Massachusetts) and the Beth Israel Deaconess Medical Center (Boston, Massachusetts). Cerebral infarction with ICU admission was diagnosed based on International Classification of Diseases, Ninth Revision (ICD-9) code and the Tenth Revision (ICD-10) code. ICD-9: 43301, 43311, 43321, 43331, 43381, 43391, 43401, 43411 and 43491; ICD-10: I63. In our study, patients aged < 18 years were excluded. Those admitted to ICU < 24 h were also excluded. Finally, the data of 2778 participants were analyzed. All people were divided into the alive group (n = 2083) and the dead group (n = 695). After propensity score matching (PSM) on age and gender, 2085 people were involved in and divided into the alive group (n = 1390) and the dead group (n = 695).
Variables
The main variable investigated was BUN/Cr. Covariables analyzed in the study including comorbidities [congestive heart failure (CHF) (yes or no), atrial fibrillation (AF) (yes or no), diabetes mellitus (yes or no), respiratory failure (yes or no), renal failure (yes or no), malignant cancer (yes or no), hypertension (yes or no), and liver disease (yes or no)], medication use [thrombolytic (yes or no), and anticoagulation (yes or no)], laboratory data [heart rate (time/min), systolic blood pressure (SBP) (mmhg), diastolic blood pressure (DBP) (mmhg), mean arterial pressure (MAP) (mmhg), respiratory rate (time/min), temperature (℃), white blood cell (WBC) (K/uL), platelets (PLT) (K/uL), hemoglobin (g/dL), red cell distribution width (RDW) percent, hematocrit percent, Cr (mg/dl), International Normalized Ratio (INR), BUN (mg/dl), glucose (mg/dl), bicarbonate (mEq/l), sodium (mEq/l), and potassium (mEq/l)], and the Sequential Organ Failure Assessment (SOFA) Score, the Simplified Acute Physiology Score II (SAPSII), the Oxford Acute Severity of Illness Score (OASIS), and charlson comorbidity index.
Outcome Variable
The outcome variable in the present study was the in-hospital death of participants with cerebral infarction in ICU. All subjects in the ICU were followed up until death or discharge. The median follow-up time was 10.5 days. Among 2778 participants, 695 were dead at the end of follow-up.
Sensitivity Analysis
The missing values of all variables were shown in Supplementary Table 1. The results of sensitivity analysis revealed that no statistical difference was observed in the data before and after multi-interpolation. As exhibited Supplementary Table 2, the age was statistically different between the alive group and the dead group (66.42 years vs 71.51 years). To make the baseline data equilibrium between the alive group and the dead group, PSM was applied. After PSM, the age and gender showed no statistical difference between the two groups.
Statistical analysis
The continuous variables were in the forms of mean ± standard deviation (SD) if the data were normally distributed or M (Q1, Q3) if the data were not normally distributed. Student’s t test was used to compare the difference between groups. The categorical variables were displayed as n (%), and chi-square and Wilcoxon rank sum test were applied to judge the differences between groups. The data with missing value < 10% were multi-interpolated, and with missing value ≥ 10% were excluded. PSM was performed in the data. Sensitivity analysis was performed between the data before multi-interpolation and after multi-interpolation as well as before PSM and after PSM. Multivariate logistic analyses were applied for identifying the confounders of the association between BUN/Cr and mortality of cerebral infraction. Model 1 included all variables with statistical difference between the alive group and the dead group. Model 2 included BUN/Cr, respiratory failure, malignant cancer, anticoagulation, liver disease, temperature, WBC, RDW percent, glucose and bicarbonate. Subgroup analysis was performed to assess the association between BUN/Cr and the mortality of cerebral infraction in different groups of people concerning age, gender, and whether the patient was complicated with respiratory failure, malignant cancer or liver disease, or whether the whether the patient received anticoagulation treatments. The risk ratio (RR) was employed to evaluate the association between BUN/Cr and the mortality of cerebral infraction in ICU. The confidence level was set as α = 0.05. All statistical analysis was conducted via SAS 9.4, and R 4.0.3.