Blood urea nitrogen to serum albumin ratio: a novel mortality indicator in intensive care unit patients with coronary heart … – Nature.com

Database and study population

The data in this study were all obtained from a large publicly accessible MIMIC-IV database version 2.2 (https://mimic.physionet.org/about/mimic/), which is the latest version that encompasses extensive clinical data from patients admitted to the Beth Israel Deaconess Medical Center between 2008 and 2019. MIMIC-IV contains more than 50,000 admissions information for adult patients including demographic characteristics, laboratory tests results, medication treatment, vital signs and other comprehensive information39. To obtain access, one author (ZLZ) completed web-based training courses and exams and gained full access to the database (certification number: 53158939). The collection of patient information and creation of the MIMIC-IV data was reviewed by the Institutional Review Board at the Beth Israel Deaconess Medical Center, who granted a waiver of informed consent and approved the data sharing initiative.

This study enrolled patients who had been diagnosed with CHD, which was defined as the occurrence of myocardial infarction (MI), acute coronary syndrome, ischemic heart disease, or those who had undergone percutaneous coronary intervention or coronary artery bypass grafting40. These diagnosis were based on the International Classification of Diseases (ICD) codes, specifically ICD-9 code 410, 411, 413, 414 and ICD-10 code I20, I21, I22, I24, I25. The exclusion criteria were as follows: (1) patients who were admitted to the ICU multiple times for CHD, only the data of their initial admission were retained for analysis (n=16,145); (2) patients aged<18years at the time of the first admission; (3) patients with an ICU length for less than 24h; (4) patients without recorded (blood urea nitrogen and albumin) within 24h of admission. Finally, 4254 patients were enrolled in our study and categorized into four groups according to the quartiles of the BAR index.

The PostgreSQL software (version 15.2) were used to extract data by running Structured Query Language (SQL) from the MIMIC-IV database. The extraction of potential confounders could be categorized into following primary groups, including (1) Demographics: age, gender, race, weight and height; (2) Vital Signs: heart rate, systolic blood pressure, diastolic blood pressure, mean arterial pressure and respiratory rate; (3) Comorbidities: myocardial infarct, heart failure, acute kidney injury (AKI), sepsis, respiratory failure, hypertension, diabetes and obesity; (4) Disease Severity Scores: the sepsis-related organ failure assessment score (SOFA), simplified acute physiology score II (SAPSII), systemic inflammatory response syndrome (SIRS), Oxford acute severity of illness score (OASIS) and Acute Decompensated Heart Failure National Registry (ADHERE score, calculate by utilizing blood urea nitrogen, systolic blood pressure and creatinine)41; (5) Laboratory Indicators: hemoglobin, platelets, white blood cells (WBC), anion gap, bicarbonate, serum sodium, serum calcium, serum potassium, serum chloride, total bilirubin, serum glucose, serum creatinine, international normalized ratio (INR), prothrombin time (PT), partial thromboplastin time (PTT), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and aspartate aminotransferase (AST). The value of the BAR index was the ratio of bun and albumin. Furthermore, the primary outcomes of this study included in-hospital mortality, all-cause mortality within 28-day and 1-year after ICU admission. These outcomes were calculated based on the hospital stay data, like the hospital expire flag, and the follow-up data, such as the recorded out-of-hospital date of death. All vital signs, laboratory variables and disease severity scores were obtained from the data collected within the initial 24h after the patient's admission to the ICU.

In order to avoid potential bias, variables were excluded in the analysis if they had missing values exceeding 15% of the total observations. Variables with missing values less than 15% were processed by multiple imputation using the Predictive Mean Matching (PMM) method through the mice package in R42.

Continuous variables were presented as median (Interquartile range), whereas categorical variables were expressed as frequencies and proportions. The former were compared by Wilcoxon test or KruskalWallis test, and the latter were compared by Pearsons chi-squared test. KaplanMeier survival analysis was used to assess the relationship between the incident of CHD during follow-up and the quartiles of the BAR index, differences among groups were assessed by log-rank tests.

Cox proportional hazard models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the higher quartiles compared to the lowest quartile, as well as for each 1-standard deviation increment in the BAR index for all the study endpoints. The P values for liner trend were calculated based on the median value of each category of BAR index as a continuous variable. Three models were constructed to sequentially adjust for potential confounders of CHD: model 1: unadjusted; model 2: adjusted for age, sex, and race; model 3: adjusted for age, sex, race, hemoglobin, platelet, glucose, WBC, bicarbonate, creatinine, ALT, heart rate, SBP, MBP, heart failure, obesity, hypertension, diabetes, AKI, sepsis, SOFA and OASIS. The proportional hazards assumption was assessed relying on Schoenfeld residuals, the results revealed that there were no violations with a P value>0.05. Additionally, we investigated the possible nonlinear relationship between the baseline BAR index and the risk of all-cause mortality using a restricted cubic spline (RCS) regression model with four knots. Subgroup analyses were conducted to explored whether the associations varied based on sex, age (<75, 75years), hypertension, diabetes, obesity, heart failure, acute heart failure, CS, AKI and sepsis. Tests for the interactions between the factor of interest and each potential effect modifier were performed by fitting an interaction term.

We conducted Receiver Operating Characteristic (ROC) analysis to assess the predictive power of BUN, albumin, BAR, SOFA, OASIS, and ADHERE regarding in-hospital mortality, as well as mortality within 28 days and 1 year following ICU admission. To evaluate the discriminative performance of these parameters, we employed the DeLong test to compare their areas under the curves (AUCs). Additionally, the optimal cut-off value for BAR was identified using the Youden index, optimizing the balance between sensitivity and specificity.

All statistical analyses were performed by using R software (version 4.2.3). A two-sided P value of <0.05 was considered statistically significant.

Originally posted here:
Blood urea nitrogen to serum albumin ratio: a novel mortality indicator in intensive care unit patients with coronary heart ... - Nature.com