Blood urea nitrogen and in-hospital mortality in critically ill patients with MI or cardiogenic shock: Analysis of the MIMIC-IV database

Abstract

Background

Blood urea nitrogen (BUN) reflects renal function and metabolic catabolism and thus can be a potential biomarker in severe illness. The purpose of this investigation was to assess whether independent associations exist between admission BUN concentrations and hospital mortality in critically ill adults with acute myocardial infarction (MI) or cardiogenic shock (CS). Associations were assessed using tertiles of mean BUN during the hospital admission and mean BUN within 24 h of ICU entry.

Methods

Using MIMIC-IV (v3.1), we conducted a retrospective cohort study of first ICU admissions (2008–2019) in adults identified by ICD-9/10 codes for acute myocardial infarction and/or cardiogenic shock. The primary exposure was mean BUN during hospitalization, categorized into tertiles; first-24-h BUN was analyzed as a sensitivity exposure. Analyses used complete cases.

Results

We included 6719 patients. In-hospital mortality climbed from 5.2% to 25.8% across average BUN tertiles. In multivariable logistic regression, higher mean BUN was associated with increased odds of in-hospital mortality across tertiles and per standard-deviation increase. Findings were directionally consistent when first-24-h BUN was used as the exposure. Model diagnostics supported acceptable collinearity and fit.

Conclusions

Among critically ill adults with acute myocardial infarction and/or cardiogenic shock, higher BUN, particularly mean BUN during hospitalization, was independently associated with in-hospital mortality. Early BUN (first 24 h) showed a similar signal, supporting BUN as a simple adjunct for bedside risk assessment.

Keywords

Blood urea nitrogen hospital mortality acute myocardial infarction cardiogenic shock MIMIC-IV critical care prognosis

Introduction

Acute myocardial infarction (AMI) is responsible for nearly one-third of all fatalities and stands as the primary cause of death on a global scale.¹ The occurrence of AMI accompanied by cardiogenic shock (CS) is a critical syndrome marked by a swift deterioration in cardiac function.² Shock is a state of circulatory failure that leads to an insufficient supply of oxygen, resulting in potentially life-threatening tissue hypoxia.³ Cardiogenic shock (CS) is particularly complex and critical; it arises from inadequate cardiac output, resulting in poor tissue perfusion and limited oxygen delivery. This deficiency can ultimately cause damage to vital organs.^4,5 Among AMI patients, CS occurs in 5% to 10% of cases.^6–8 Despite significant reductions in AMI incidence and mortality,^9,10 outcomes in patients with CS remain poor.¹¹

Even with advancements in reperfusion, in-hospital mortality rates for CS remain alarmingly high, ranging from 27% to 51%, with nearly half of deaths occurring within the first 24 h.^5,12,13 This implies that the risk of death is time-sensitive. Therefore, it is important to early recognize patients at high risk of myocardial infarction and cardiogenic shock.¹³ Furthermore, research has shown that long-term survival in AMI-CS patients is poor with little improvement over time, and survivors are destined to develop significant morbidity and also an increased risk for readmission and mortality.¹⁴ Hence, there exists a critical need for simple and effective methods to identify and stratify at-risk patients for increased mortality.

Blood urea nitrogen (BUN) is a biochemistry indicator that portrays protein metabolism and renal excretion processes primarily occurring in the liver and kidneys. It is also utilized to calculate protein intake and evaluate kidney function.^15,16 Increased BUN levels not only reflect impaired cardiorenal function but are also associated with neurohormonal activation, and therefore BUN is a valid adverse prognostic indicator in various clinical conditions.^17,18 In cardiogenic shock, impaired left ventricular function reduces cardiac output, and decreased coronary perfusion activates the neurohormonal system, which initiates vasoconstriction, increased heart rate, and fluid retention, all of which exacerbate myocardial ischemia. This pattern creates a self-perpetuating cycle of decreasing shock and compromise in myocardial function.^19,20 Therefore, BUN concentration in CS may serve as a marker of the severity of underlying pathophysiology and provide valuable prognostic information.

Increased BUN concentrations have also been seen by other studies to independently predict mortality in critically ill patients, which provides additional evidence for its use as a prognostic marker.²¹ However, most of these studies were limited by small sample sizes, and few had examined the association between BUN levels and in-hospital mortality in cardiogenic shock patients using large-scale databases. To address this gap, the present study is the first to utilize the MIMIC-IV database, offering a more recent and comprehensive dataset of critically ill CS patients. By leveraging this updated resource, our research aims to provide current evidence on the prognostic value of BUN, contributing to improved risk stratification and supporting clinical decision-making in the management of cardiogenic shock.

Methods

Study design and data source

This was a retrospective cohort study based on data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, version 3.1. MIMIC-IV is a large, de-identified, publicly accessible database of extensive health-related information on patients admitted to critical care units in Beth Israel Deaconess Medical Center from 2008 to 2019.²²

Study population and inclusion/exclusion criteria

This retrospective cohort study used routinely collected intensive care unit (ICU) data. We first identified all ICU stays and retained only the first ICU admission per patient (ranked by ICU admission time) to avoid within-patient clustering. The index stay was required to have a recorded initial ICU type (first care unit). We then restricted the population to adult patients (age ≥18 years) with evidence of acute myocardial infarction and/or shock syndromes during the index hospital admission, defined using ICD-9 and ICD-10 diagnosis codes. Eligible admissions included acute myocardial infarction (ICD-9 410* or ICD-10 I21*), cardiogenic shock (ICD-9 78551 or ICD-10 R57.0), or other shock categories (ICD-10 R57* excluding R57.0).

To focus on clinically comparable critical illness and ensure meaningful early ICU exposure, we required an ICU length of stay of at least 1 day. We further required at least one blood urea nitrogen (BUN) laboratory measurement during the hospital admission, because BUN was the principal exposure. Patients were excluded if they had diagnosis codes consistent with severe liver disease or metastatic cancer, given the strong, non-modifiable influence of these conditions on BUN metabolism and short-term mortality. Finally, we excluded patients who received renal replacement therapy (RRT) during the hospital admission, identified from ICU charted therapy indicators and procedure codes, to reduce confounding by treatment-driven changes in nitrogen balance. The analytic cohort additionally required non-missing values for the primary exposure and the in-hospital mortality outcome.

Variable definitions and data extraction

The data were accessed using Google BigQuery by running a custom SQL script. Demographics included age (years), sex (coded as male vs non-male), and ethnicity as recorded at hospital admission. ICU characteristics included the first care unit and ICU length of stay. The primary outcome was in-hospital mortality, derived from the hospital expiration flag.

The primary exposure, average BUN during the hospital admission (bun_mean_all), was calculated as the mean of all available blood BUN values. A secondary exposure, BUN in the first 24 h after ICU admission (bun_first24), was computed using BUN measurements whose timestamps fell within 24 h of ICU entry. BUN was identified from blood urea nitrogen laboratory item identifiers. Additional laboratory covariates (electrolytes, glucose, creatinine, anion gap, hematology indices, and coagulation tests) were extracted from blood-based measurements and summarized as admission-level means. Vital signs were obtained from ICU chart events within the first 24 h after ICU admission and summarized as the maximum recorded value for each parameter over that window.

Statistical analysis

We documented cohort assembly using the selection flags so that each eligibility step could be reproduced directly from the dataset. Baseline characteristics were described overall and compared across tertiles of bun_mean_all, using cut points at the 33.3rd and 66.7th percentiles. Continuous variables were summarized as mean ± standard deviation or median (interquartile range), depending on distributional shape, while categorical variables were summarized as counts and percentages. Between-tertile comparisons used standard tests matched to data type and distribution (e.g. ANOVA or non-parametric alternatives for continuous variables, and chi-square tests for categorical variables).

The association between BUN and in-hospital mortality was evaluated using logistic regression and reported as odds ratios with 95% confidence intervals. BUN was modeled in two complementary ways: (1) per one standard deviation increase (standardized BUN) to provide a scale-free effect estimate, and (2) as tertiles (T2 and T3 compared with T1) to support clinical interpretability. We fitted sequential models to show how adjustment changed the estimate: unadjusted models; age- and sex-adjusted models; and multivariable models additionally adjusting for available early hemodynamic and laboratory covariates. Each model was fit on complete cases for its included predictors; covariates with no usable values were omitted from the corresponding model.

To support interpretability and model stability, we performed diagnostic checks that were aligned with our research question. We screened multicollinearity using variance inflation factors, and we probed departures from linearity in the log-odds by adding a quadratic term for standardized BUN. Discrimination was assessed with receiver operating characteristic analysis, reporting the area under the curve with a 95% confidence interval and an optimal predicted-probability cutoff based on Youden's index. Sensitivity analyses repeated the same modeling framework using bun_first24 (per standard deviation and tertiles) to test robustness to exposure timing. Finally, we evaluated potential non-linearity using a restricted cubic spline for bun_mean_all (four knots at the 5th, 35th, 65th, and 95th percentiles) and visualized the adjusted predicted probability of in-hospital mortality with 95% confidence intervals across the observed BUN range. All analyses were conducted with R 4.5.2.

Results

Characteristics of study population

After applying the prespecified inclusion and exclusion criteria, we identified 6719 adult ICU admissions with myocardial infarction and/or shock and at least one blood urea nitrogen (BUN) measurement for the analytic cohort (Figure 1). Patients were divided into tertiles of average BUN during the hospital stay.

Figure 1.

CONSORT flow diagram—sample selection.

Baseline characteristics differed substantially across BUN tertiles (Table 1). Patients in higher tertiles were older and more often male. The distribution of ethnicity and ICU type also shifted with rising BUN, with a higher proportion of White patients and a greater share of non-coronary ICU admissions in the upper tertile.

Table 1.

Baseline characteristics of the analytic cohort according to tertiles of average BUN during the hospital stay.

Variable	T1	T2	T3	P value
Number of patients, n	2257	2222	2240	-
Mean BUN	14.14 (11.75–16.09)	22.60 (20.19–25.56)	43.27 (35.11–58.72)	-
BUN (first 24 h)	13.00 (10.54–16.00)	21.00 (17.00–26.00)	40.60 (29.67–58.33)	-
Age	62.66 ± 13.92	69.74 ± 12.85	73.28 ± 12.71	<0.0001
Gender (male)	1326 (58.8%)	1430 (64.4%)	1458 (65.1%)	<0.0001
Heart rate	101.32 ± 19.64	102.38 ± 20.05	103.59 ± 22.07	0.0011
SBP	137.00 (127.00–150.00)	138.00 (128.00–152.00)	137.00 (125.00–152.00)	0.2058
DBP	75.00 (68.00–85.00)	73.00 (65.00–82.00)	70.00 (62.00–81.00)	<0.0001
MBP	97.00 (89.00–109.00)	95.00 (87.00–107.00)	93.00 (84.00–106.00)	<0.0001
RR	27.00 (24.00–31.00)	27.00 (24.00–31.00)	28.00 (25.00–32.00)	<0.0001
Temperature	98.80 (98.50–99.40)	98.80 (98.40–99.50)	98.70 (98.30–99.30)	<0.0001
SpO2	100.00 (100.00–100.00)	100.00 (100.00–100.00)	100.00 (99.00–100.00)	<0.0001
Urine output	175.30 (149.60–205.00)	179.50 (150.00–212.10)	176.30 (150.70–209.00)	0.0488
Sodium	138.10 ± 3.17	138.47 ± 3.45	138.94 ± 4.55	<0.0001
Potassium	4.08 (3.91–4.26)	4.15 (3.96–4.33)	4.23 (4.00–4.51)	<0.0001
Creatinine	0.79 (0.66–0.93)	1.03 (0.86–1.24)	1.71 (1.31–2.35)	<0.0001
Hemoglobin	10.83 ± 1.86	10.60 ± 1.78	9.86 ± 1.71	<0.0001
Glucose	117.44 (107.10–136.25)	126.39 (111.63–151.00)	141.39 (120.26–175.74)	<0.0001
Anion gap	12.22 (10.75–13.71)	12.80 (11.29–14.33)	14.52 (12.60–16.62)	<0.0001
Chloride	103.17 ± 3.82	102.84 ± 4.27	102.40 ± 5.81	<0.0001
Hematocrit	32.64 ± 5.19	32.21 ± 5.12	30.45 ± 5.14	<0.0001
Bicarbonate	24.66 ± 2.69	24.89 ± 3.45	23.90 ± 4.50	<0.0001
Platelet	206.00 (163.82–256.43)	201.66 (158.42–256.82)	196.86 (146.00–252.72)	<0.0001
WBC	10.15 (8.28–12.32)	10.64 (8.60–13.01)	11.25 (8.89–14.13)	<0.0001
aPTT	39.70 (29.98–54.28)	41.12 (30.59–54.76)	44.42 (31.66–60.34)	<0.0001
PT	13.55 (12.47–14.90)	13.90 (12.75–15.70)	14.75 (13.05–18.43)	<0.0001
INR	1.23 (1.13–1.36)	1.27 (1.16–1.44)	1.34 (1.19–1.70)	<0.0001
Hospital mortality	118 (5.2%)	191 (8.6%)	577 (25.8%)	<0.0001
ethnicity	5 (0.2%)	4 (0.2%)	3 (0.1%)	0.0023
- ASIAN	30 (1.3%)	20 (0.9%)	17 (0.8%)	-
- ASIAN - ASIAN INDIAN	8 (0.4%)	4 (0.2%)	5 (0.2%)	-
- ASIAN - CHINESE	10 (0.4%)	13 (0.6%)	19 (0.8%)	-
- ASIAN - KOREAN	4 (0.2%)	2 (0.1%)	1 (0.0%)	-
- ASIAN - SOUTH EAST ASIAN	3 (0.1%)	5 (0.2%)	8 (0.4%)	-
- BLACK/AFRICAN	3 (0.1%)	9 (0.4%)	10 (0.4%)	-
- BLACK/AFRICAN AMERICAN	143 (6.3%)	114 (5.1%)	135 (6.0%)	-
- BLACK/CAPE VERDEAN	9 (0.4%)	7 (0.3%)	4 (0.2%)	-
- BLACK/CARIBBEAN ISLAND	11 (0.5%)	15 (0.7%)	8 (0.4%)	-
- HISPANIC OR LATINO	9 (0.4%)	9 (0.4%)	6 (0.3%)	-
- HISPANIC/LATINO - CENTRAL AMERICAN	1 (0.0%)	0 (0.0%)	0 (0.0%)	-
- HISPANIC/LATINO - COLUMBIAN	1 (0.0%)	0 (0.0%)	0 (0.0%)	-
- HISPANIC/LATINO - CUBAN	4 (0.2%)	4 (0.2%)	0 (0.0%)	-
- HISPANIC/LATINO - DOMINICAN	29 (1.3%)	13 (0.6%)	18 (0.8%)	-
- HISPANIC/LATINO - GUATEMALAN	5 (0.2%)	3 (0.1%)	1 (0.0%)	-
- HISPANIC/LATINO - HONDURAN	0 (0.0%)	2 (0.1%)	3 (0.1%)	-
- HISPANIC/LATINO - MEXICAN	0 (0.0%)	2 (0.1%)	1 (0.0%)	-
- HISPANIC/LATINO - PUERTO RICAN	20 (0.9%)	18 (0.8%)	19 (0.8%)	-
- HISPANIC/LATINO - SALVADORAN	6 (0.3%)	3 (0.1%)	2 (0.1%)	-
- MULTIPLE RACE/ETHNICITY	2 (0.1%)	0 (0.0%)	2 (0.1%)	-
- NATIVE HAWAIIAN OR OTHER PACIFIC ISLANDER	3 (0.1%)	4 (0.2%)	3 (0.1%)	-
- OTHER	93 (4.1%)	68 (3.1%)	60 (2.7%)	-
- PATIENT DECLINED TO ANSWER	11 (0.5%)	19 (0.9%)	9 (0.4%)	-
- PORTUGUESE	2 (0.1%)	6 (0.3%)	2 (0.1%)	-
- SOUTH AMERICAN	1 (0.0%)	0 (0.0%)	2 (0.1%)	-
- UNABLE TO OBTAIN	107 (4.7%)	89 (4.0%)	81 (3.6%)	-
- UNKNOWN	401 (17.8%)	377 (17.0%)	340 (15.2%)	-
- WHITE	1243 (55.1%)	1319 (59.4%)	1373 (61.3%)	-
- WHITE - BRAZILIAN	4 (0.2%)	4 (0.2%)	0 (0.0%)	-
- WHITE - EASTERN EUROPEAN	6 (0.3%)	9 (0.4%)	9 (0.4%)	-
- WHITE - OTHER EUROPEAN	66 (2.9%)	55 (2.5%)	59 (2.6%)	-
- WHITE - RUSSIAN	17 (0.8%)	25 (1.1%)	40 (1.8%)	-
ICU type	880 (39.0%)	830 (37.4%)	438 (19.6%)	<0.0001
- Coronary Care Unit (CCU)	658 (29.2%)	634 (28.5%)	849 (37.9%)	-
- Intensive Care Unit (ICU)	0 (0.0%)	2 (0.1%)	3 (0.1%)	-
- Medical Intensive Care Unit (MICU)	253 (11.2%)	305 (13.7%)	445 (19.9%)	-
- Medical/Surgical Intensive Care Unit (MICU/SICU)	150 (6.6%)	155 (7.0%)	239 (10.7%)	-
- Medicine	0 (0.0%)	3 (0.1%)	1 (0.0%)	-
- Neuro Intermediate	37 (1.6%)	44 (2.0%)	17 (0.8%)	-
- Neuro Stepdown	6 (0.3%)	6 (0.3%)	3 (0.1%)	-
- Neuro Surgical Intensive Care Unit (Neuro SICU)	30 (1.3%)	29 (1.3%)	26 (1.2%)	-
- PACU	1 (0.0%)	1 (0.0%)	2 (0.1%)	-
- Surgery/Trauma	0 (0.0%)	0 (0.0%)	1 (0.0%)	-
- Surgery/Vascular/Intermediate	0 (0.0%)	8 (0.4%)	13 (0.6%)	-
- Surgical Intensive Care Unit (SICU)	121 (5.4%)	102 (4.6%)	107 (4.8%)	-
- Trauma SICU (TSICU)	121 (5.4%)	103 (4.6%)	96 (4.3%)	-

P-values reflect overall differences across BUN tertiles based on appropriate statistical tests (ANOVA or Kruskal–Wallis for continuous variables and chi-square tests for categorical variables).

Hemodynamic variables showed modest but consistent differences. Compared with T1, patients in T3 had lower diastolic and mean arterial blood pressures and slightly higher respiratory rates, while heart rate remained broadly similar. Temperature and oxygen saturation were marginally lower in the highest tertile, although absolute differences were small.

Laboratory values displayed a clear pattern compatible with more advanced renal and systemic dysfunction at higher BUN levels. Creatinine increased steeply across tertiles. Electrolytes and acid–base parameters also changed: potassium and anion gap rose, whereas chloride and bicarbonate tended to fall with higher BUN. Markers of anemia and coagulopathy were more abnormal in T3, with lower hemoglobin and hematocrit and higher aPTT, PT, and INR. Platelet counts and white blood cell counts showed small but statistically significant shifts, suggesting a more inflamed and coagulopathic phenotype in patients with elevated BUN.

In-hospital mortality increased sharply across BUN tertiles. The crude mortality rate was 5.2% in T1, 8.6% in T2, and 25.8% in T3 (p < 0.0001), indicating that patients with the highest average BUN experienced nearly a five-fold higher absolute risk of death than those in the lowest tertile.

Relationship of BUN to hospital mortality

In unadjusted logistic regression, each one–standard deviation (SD) increase in average BUN was associated with an odds ratio (OR) for in-hospital death of 1.88. When BUN was modeled categorically, T2 and T3 were strongly associated with mortality compared with T1 (OR 1.70 and 6.29, respectively; both p < 0.0001). Adjustment for age and sex produced only modest attenuation of these associations.

In the fully adjusted model, controlling for demographic factors, vital signs, laboratory variables, and ICU type, the strength of the association remained pronounced, although the gradient became more concentrated in the highest tertile. Each SD increase in average BUN conferred an OR of 2.48. Compared with T1, T3 retained a robust association with mortality OR 3.01, whereas the association for T2 was no longer statistically significant (OR 1.20, 95% CI 0.69–2.11; p = 0.52).

The discriminative performance of the fully adjusted BUN model was excellent. The area under the ROC curve was 0.906 (Figure 2). At the optimal predicted-probability cutoff derived from Youden's index (0.071), sensitivity was 86.3% and specificity was 80.2%, indicating that the model correctly identified the majority of patients who died while maintaining good specificity among survivors (Table 2).

Figure 2.

Receiver operating characteristic (ROC) curve for the adjusted BUN model. The AUC ROC was 0.906 (95% CI 0.884–0.928). Using Youden's index, the optimal predicted-probability cutoff was 0.071, with sensitivity 86.3% and specificity 80.2%.

Table 2.

Association between average BUN during the hospital stay and in-hospital mortality.

Model	N	Deaths	Exposure	OR (95% CI)	P-value
Unadjusted	6719	886 (13.2%)	BUN (per SD)	1.88 (1.77–1.99)	<0.0001
	6719	886 (13.2%)	T2 vs T1	1.70 (1.35–2.17)	<0.0001
	6719	886 (13.2%)	T3 vs T1	6.29 (5.13–7.78)	<0.0001
Age- and sex-adjusted	6719	886 (13.2%)	BUN (per SD)	1.85 (1.74–1.97)	<0.0001
	6719	886 (13.2%)	T2 vs T1	1.64 (1.29–2.09)	0.0001
	6719	886 (13.2%)	T3 vs T1	5.93 (4.79–7.40)	<0.0001
Fully adjusted	2205	226 (10.2%)	BUN (per SD)	2.48 (1.87–3.33)	<0.0001
	2205	226 (10.2%)	T2 vs T1	1.20 (0.69–2.11)	0.5237
	2205	226 (10.2%)	T3 vs T1	3.01 (1.65–5.59)	0.0004

This table presents odds ratios (with 95% confidence intervals) for in-hospital mortality associated with average BUN during the hospital stay (bun_mean_all), modeled both per one standard deviation increase and as tertiles (T2 and T3 compared with T1). Unadjusted, age- and sex-adjusted, and fully or fallback adjusted models are shown. Each model reports the complete-case sample size (N) and the number of deaths to aid interpretation.

Sensitivity and spline analyses

Sensitivity analyses using BUN measured during the first 24 h after ICU admission yielded consistent findings (Table 3). In the fully adjusted model, each SD increase in first-24-h BUN was associated with an OR of 1.56. The highest tertile of early BUN remained significantly associated with mortality compared with the lowest tertile (OR 1.89), whereas the middle tertile again showed no clear excess risk.

Table 3.

Sensitivity analysis: association between BUN in the first 24 h and in-hospital mortality.

Model	N	Deaths (%)	Exposure	OR (95% CI)	P-value
Unadjusted	6694	884 (13.2%)	BUN first 24 h (per SD)	1.62 (1.52–1.71)	<0.0001
	6694	884 (13.2%)	T2 vs T1 (first 24 h)	1.97 (1.57–2.47)	<0.0001
	6694	884 (13.2%)	T3 vs T1 (first 24 h)	5.41 (4.42–6.68)	<0.0001
Age- and sex-adjusted	6694	884 (13.2%)	BUN first 24 h (per SD)	1.58 (1.48–1.67)	<0.0001
	6694	884 (13.2%)	T2 vs T1 (first 24 h)	1.87 (1.49–2.35)	<0.0001
	6694	884 (13.2%)	T3 vs T1 (first 24 h)	4.97 (4.03–6.18)	<0.0001
Fully adjusted	2199	226 (10.3%)	BUN first 24 h (per SD)	1.56 (1.23–1.98)	0.0003
	2199	226 (10.3%)	T2 vs T1 (first 24 h)	0.89 (0.54–1.48)	0.6525
	2199	226 (10.3%)	T3 vs T1 (first 24 h)	1.89 (1.09–3.28)	0.0230

This table presents sensitivity analyses in which the exposure is BUN measured during the first 24 h after ICU admission (bun_first24). BUN is modeled both per standard deviation increase and in tertiles (T2 and T3 compared with T1), using unadjusted, age- and sex-adjusted, and fully or fallback adjusted logistic regression. Model-specific complete-case sample size (N) and deaths are shown to facilitate interpretation.

Finally, restricted cubic spline analyses did not provide evidence of a strongly non-linear relationship between average BUN and mortality after adjustment for covariates (Figure 3, Table 4). Adding linear BUN to the covariate-only model significantly improved model fit (χ² = 39.09, p < 0.0001), whereas allowing for spline terms did not materially reduce the AIC or significantly enhance fit. These findings support a largely linear increase in the odds of in-hospital death across the observed range of BUN, with the greatest absolute risk concentrated among patients in the highest tertile.

Figure 3.

Non-linear association between average BUN and in-hospital mortality.

Table 4.

Metrics for spline model.

Comparison	df	Chi square	P value	AIC covariates only	AIC linear BUN	AIC spline BUN
Covariates-only vs + linear BUN	1	39.09	<0.0001	957.2	920.1	923
Linear BUN vs spline BUN	4	5.14	0.2728		920.1	923

Discussion

Summary of findings

The current study, based on the large dataset of MIMIC-IV, investigated the correlation between BUN levels of ICU-admitted patients with CS, both categorical and quantitative, and in-hospital mortality. The outcomes of multivariate logistic regression analysis showed that higher levels of BUN were significantly related to in-hospital mortality and worse prognosis in CS patients.

Comparison with prior research

In this study, we stratified patients by BUN tertile and found that patients in the highest tertiles had a higher death rate than those in the lowest tertiles, which suggests that the BUN prognostic value as a continuous variable may be underestimated. These findings are consistent with the increasing evidence linking renal dysfunction markers to adverse outcomes in critical care settings.²³ Furthermore, the association was maintained after adjustment for confounders such as age, sex, vital signs, and laboratory parameters included in the fully adjusted model, suggesting that BUN plays an independent role in stratification.

Aligned with our results, Zhang et al. used the MIMIC-IV and eICU-CRD databases to develop an explicable machine learning (ML) model that identified BUN as a key predictor of hospital mortality in patients with AMI.²⁴ In a similar study of AMI cases and prediction of in-hospital mortality by ML models, BUN and CS were both important predictors and related to inpatient mortality.²⁵ This study's findings are consistent with the previous literature in the MIMIC-III database,^26,27 e.g. lower BUN or higher BUN-to-Cr ratios were associated with a reduced risk of mortality in patients admitted to ICU with CS.^26,27 Moreover, supporting evidence has evaluated the threshold value of BUN for the mortality risk of CS patients.^28,29 In the study by Zhu et al., BUN levels above 8.95 mmol/L significantly increased the 30-day mortality in AMI-CS patients.²⁸ In addition, Yamga et al. have developed a checklist-based risk score including a BUN limit of 25 mg/dL (≈8.93 mmol/L), which is consistent with our findings and improves its clinical utility.²⁹ These consistent results across different datasets and methodologies reinforce the evidence for the usefulness of BUN as a reliable prognostic marker.

Several possible mechanisms could explain the elevated BUN levels and higher risk of adverse events in patients with MI or CS. First, during acute myocardial infarction—especially when complicated by cardiogenic shock—a sudden hypoperfusion reduces Glomerular Filtration Rate (GFR), increases filtration parameters, such as BUN or serum Cr, and contributes to kidney injury.^21,30 Consequently, Acute Kidney Injury (AKI) as a complication of CS has been shown to correlate with a higher mortality risk.³¹ Second, high BUN levels may also be related to low blood volume in CS cases. Diuretic use, vomiting, diaphoresis, and poor intake in MI patients, combined with dehydration, can further increase mortality in critically ill patients.^20,28 Third, reduced cardiac output during MI or CS activates neurohormonal mechanisms, which include sympathetic nervous system (SNS) activation, renin-angiotensin-aldosterone system (RAAS) stimulation, and arginine vasopressin (AVP) release.³² These may raise BUN by vasoconstriction, sodium and water retention, and urea reabsorption in renal tubules.^19,32,33 In addition, Jujo et al. demonstrated that persistent high BUN levels are associated with increased cardiovascular death in acute heart failure, a finding that is relevant for our CS cohort, where sustained BUN elevations after admission predicted a lower survival.³³ Together, these mechanisms highlight the multifunctional role of BUN in reflecting both renal and hemodynamic derangement.³⁴

Some studies have evaluated biochemical markers in CS patients after administration of treatment modalities such as medication or dialysis.^35,36 Wang et al. examined the effect of intravenous levosimendan in acute HF patients.³⁵ Levosimendan significantly decreased B-type natriuretic peptide (BNP) and increased urine output without a meaningful reduction in BUN or lactate levels of the patients after administration.³⁵ Another study performed peritoneal dialysis in patients with cardiorenal syndrome type 1, of whom 68% had CS.³⁶ However, according to the results of multivariate analysis, BUN and Creatinine (Cr) levels at dialysis initiation were not correlated with in-hospital mortality.³⁶ These different outcomes compared with our results might be attributed to the limited sample size, different study settings or interventions, and focus on baseline BUN.^35,36 Future studies should look at the longitudinal BUN trends after treatment to reconcile these differences.

In addition, prospective studies should investigate whether targeted interventions based on BUN levels, such as individualized fluid therapy and renal protection strategies, can improve patient outcomes. As BUN is influenced by factors such as diet and medication use, the integration of BUN with other biomarkers such as albumin and inflammatory markers may provide a more comprehensive assessment in patients with critical disease.²¹

In this study, BUN was associated with mortality in critically ill patients in a fully adjusted model, which is in line with previous studies such as Biere et al., which concluded that elevated BUN in patients with normal creatinine was independently associated with increased mortality in critically ill patients.²¹ This adjustment minimizes the confounding effects of demographic and physiological variables, indicating that BUN independently contributes to the adverse outcomes.³⁷

Strengths and limitations strengths

Large, High-Resolution Database: Utilization of the MIMIC-IV database offers access to a sizable cohort of real-world critically ill patients, with detailed, time-stamped clinical and laboratory data, which improves statistical power and generalizability in comparable intensive care environments.

Robust Patient Selection: Our analysis used comprehensive inclusion and exclusion criteria using ICD codes and clinical variables to provide a targeted and clinically meaningful patient population of MI and CS patients and reduce confounding due to conditions such as chronic kidney disease, metastatic malignancy, or brief ICU stays.

Complete Adjustment: The logistic regression models included a broad array of demographic, vital sign, and laboratory covariates, enabling vigorous adjustment of potential confounders and reinforcing the inference of an independent relationship of BUN with mortality.

Transparency: The process, right down to the exact SQL queries used for data extraction and Python scripts used for analysis, is explained in the declarations, fostering reproducibility and transparency.

Limitations:

Retrospective design: This retrospective observational analysis can demonstrate association but cannot establish causality. Although we adjusted for a broad set of demographic, haemodynamic, and laboratory covariates, residual confounding from unmeasured factors (e.g. fluid balance, vasoactive medication dosing, and clinician decision-making) may persist. Future prospective studies and external validation cohorts are needed to confirm clinical utility and to test whether BUN-guided strategies improve outcomes.

Single-center data and generalisability: MIMIC-IV reflects practice at a single tertiary center, and our eligibility criteria (e.g. excluding very short ICU stays and patients with conditions that markedly distort BUN interpretation) may limit generalisability. Therefore, the magnitude of effect estimates may differ in other health systems or patient mixes. We emphasize the need for multi-center validation before applying BUN thresholds for decision-making.

Data completeness and case definition: As with many large electronic health record datasets, some variables were missing or inconsistently recorded. In the primary models, we used complete-case analyses and reported the analysed sample size for transparency; if missingness is not completely at random, bias is possible. Additionally, diagnoses were identified using ICD codes, which may introduce misclassification—particularly for cardiogenic shock. Where feasible, future work should incorporate physiologic criteria (e.g. hypotension, lactate, vasopressor requirements) or validated phenotyping algorithms.

BUN measurement timing and clinical implementation: Using the average BUN across the hospital admission may capture overall illness trajectory but is less directly actionable at the bedside. To address this, we performed a sensitivity analysis using BUN within the first 24 h after ICU admission (a proxy for admission-time assessment), which showed consistent direction and statistical significance. This supports the potential role of early BUN for practical risk stratification, while recognizing that prospective validation is required.

Clinical implications

The results indicate that BUN, a simple and inexpensive laboratory test, can be an important adjunct for early risk stratification in critically ill patients with acute MI or cardiogenic shock. Clinicians should be alerted by elevated levels of BUN to view the patient's global physiology, not solely renal function, including volume status, catabolic load, and systemic perfusion. Incorporating BUN into standard risk assessment algorithms for such patients may assist in the identification of patients at increased risk of hospital mortality, with the potential for more intensive monitoring, earlier institution of supportive therapies, or the consideration of advanced interventions. This may lead to better patient outcomes through more focused and timely interventions.

Future research

It is critical to perform future prospective studies to validate these findings in external, multi-center populations to enhance generalizability. Investigations that study the temporal variation in BUN levels and their association with mortality can provide further insight into its prognostic value. Additionally, mechanistic studies are needed to delineate the precise pathophysiological processes that link high BUN levels with adverse outcomes in patients with MI and CS, beyond conventional markers of renal impairment. Determination of the impact of targeted therapies aimed at normalizing BUN levels (e.g. vigorous fluid resuscitation and nutritional support) on clinical outcomes in this group would also be of paramount significance. Finally, the determination of the predictive value of BUN when combined with other novel biomarkers or advanced risk prediction tools may increase prognostic precision.

Conclusion

Increased blood urea nitrogen levels are a strong and independent predictor of hospital mortality in critically ill adult patients with acute myocardial infarction or cardiogenic shock. This investigation identifies BUN as an important, readily available, and inexpensive risk stratification biomarker in this high-risk group. Our results emphasize the need to interpret BUN levels not solely as a renal function marker but also as a global marker of severity of illness and compromised physiological reserve. These observations have practical implications for clinical decision-making and the design of improved patient management strategies in acute cardiac emergencies.

Supplemental Material

sj-txt-1-cvd-10.1177_20480040261434635 - Supplemental material for Blood urea nitrogen and in-hospital mortality in critically ill patients with MI or cardiogenic shock: Analysis of the MIMIC-IV database

Supplemental material, sj-txt-1-cvd-10.1177_20480040261434635 for Blood urea nitrogen and in-hospital mortality in critically ill patients with MI or cardiogenic shock: Analysis of the MIMIC-IV database by Sadaf Derakhshandeh, Sepehr Ramezanipour, Golnaz Yazdanpanah, Sahar Nasrollahi, Sina Esmaeili, Hamed Mottaghi, Atefeh Bahrambeigi, Mahmoud Abdollahi, Alaleh Alizadeh, Sepideh Hadimaleki and Danyal Yarahmadi, Niloofar Deravi in JRSM Cardiovascular Disease

Supplemental Material

sj-txt-2-cvd-10.1177_20480040261434635 - Supplemental material for Blood urea nitrogen and in-hospital mortality in critically ill patients with MI or cardiogenic shock: Analysis of the MIMIC-IV database

Supplemental material, sj-txt-2-cvd-10.1177_20480040261434635 for Blood urea nitrogen and in-hospital mortality in critically ill patients with MI or cardiogenic shock: Analysis of the MIMIC-IV database by Sadaf Derakhshandeh, Sepehr Ramezanipour, Golnaz Yazdanpanah, Sahar Nasrollahi, Sina Esmaeili, Hamed Mottaghi, Atefeh Bahrambeigi, Mahmoud Abdollahi, Alaleh Alizadeh, Sepideh Hadimaleki and Danyal Yarahmadi, Niloofar Deravi in JRSM Cardiovascular Disease

Footnotes

List of abbreviations

Acknowledgment

The authors would like to thank the researchers whose work was included in this study.

ORCID iDs

Alaleh Alizadeh

Niloofar Deravi

Danyal Yarahmadi

Ethics approval and consent to participate

MIMIC-IV contains de-identified health-related data and is made available through PhysioNet. Access requires completion of human-subjects training and acceptance of the data use agreement. Because the dataset is de-identified, this analysis was considered non-human-subjects research by the investigators’ institution; therefore, individual informed consent was not required.

Authors’ contributions

Sadaf Derakhshandeh: Conceptualization, Methodology, Data Curation, Writing - Original Draft, Visualization.

Sepehr Ramezanipour: Conceptualization, Methodology, Data Curation, Writing - Original Draft, Visualization.

Golnaz Yazdanpanah: Formal Analysis, Investigation, Writing - Review & Editing.

Sahar Nasrollahi: Formal Analysis, Investigation, Writing - Review & Editing.

Sina Esmaeili: Data Curation, Writing - Review & Editing.

Hamed Mottaghi: Data Curation, Writing - Review & Editing.

Atefeh Bahrambeigi: Data Curation, Writing - Review & Editing.

Mahmoud Abdollahi: Data Curation, Writing - Review & Editing.

Alaleh Alizadeh: Writing - Review & Editing, Supervision.

Sepideh Hadimaleki: Writing - Review & Editing.

Danyal Yarahmadi: Methodology, SQL writing, Formal Analysis, Investigation, Supervision, Project Administration, Writing - Review & Editing

Niloofar Deravi: Conceptualization, SQL writing, Methodology, Supervision, Project Administration, Writing - Review & Editing.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Availability of data and materials

The dataset(s) supporting the conclusions of this article are available in the PhysioNet repository, maintained by MIT Laboratory for Computational Physiology. The data can be accessed from:

MIMIC-IV core modules (hosp and icu):

MIMIC-IV note module (clinical notes):

Access to the dataset requires completion of a data use agreement and certification in human research.

Data citation

Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Scientific Data. 2023;10:1. doi: 10.1038/s41597-022-01899-x

Database and software information

Project name: MIMIC-IV

Project home page:

Archived version: DOI: 10.13026/s6n6-xd98

Operating system(s): Platform independent

Programming language: SQL, Python (optional for analysis)

License: PhysioNet Credentialed Health Data License

Restrictions for non-academics: Use limited to credentialed researchers; approval required

Supplemental material

Supplemental material for this article is available online.

References

Global, regional, and national age-sex specific mortality for 264 causes of death, 1980-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017; 390: 1151–1210.

Shah

Puri

Kalra

. Management of cardiogenic shock complicating acute myocardial infarction: a review. Clin Cardiol 2019; 42: 484–493.

Vincent

Ince

Bakker

. Clinical review: circulatory shock–an update: a tribute to professor Max Harry Weil. Crit Care 2012; 16: 39.

Vallabhajosyula

Dunlay

Prasad

, et al. Acute noncardiac organ failure in acute myocardial infarction with cardiogenic shock. J Am Coll Cardiol 2019; 73: 1781–1791.

van Diepen

Katz

Albert

, et al. Contemporary management of cardiogenic shock: a scientific statement from the American Heart Association. Circulation 2017; 136: e232–ee68.

Fox

Steg

Eagle

, et al. Decline in rates of death and heart failure in acute coronary syndromes, 1999-2006. Jama 2007; 297: 1892–1900.

Samsky

Krucoff

Morrow

, et al. Cardiac safety research consortium “shock II” think tank report: advancing practical approaches to generating evidence for the treatment of cardiogenic shock. Am Heart J 2020; 230: 93–97.

Thiele

Ohman

de Waha-Thiele

, et al. Management of cardiogenic shock complicating myocardial infarction: an update 2019. Eur Heart J 2019; 40: 2671–2683.

Yeh

Sidney

Chandra

, et al. Population trends in the incidence and outcomes of acute myocardial infarction. N Engl J Med 2010; 362: 2155–2165.

10.

Chi

Kanter

, et al. Trends in acute myocardial infarction by race and ethnicity. J Am Heart Assoc 2020; 9: e013542.

11.

Dalzell

. Review of cardiogenic shock after acute myocardial infarction. Jama 2022; 327: 78.

12.

Kaddoura

Elbdri

. Current evidence in the diagnosis and management of cardiogenic shock complicating acute coronary syndrome. Rev Cardiovasc Med 2021; 22: 691–715.

13.

Samsky

Morrow

Proudfoot

, et al. Cardiogenic shock after acute myocardial infarction: a review. Jama 2021; 326: 1840–1850.

14.

Sterling Lee

Fernando Shannon

Talarico

, et al. Long-Term outcomes of cardiogenic shock complicating myocardial infarction. J Am Coll Cardiol 2023; 82: 985–995.

15.

van Veldhuisen

Ruilope

Maisel

, et al. Biomarkers of renal injury and function: diagnostic, prognostic and therapeutic implications in heart failure. Eur Heart J 2016; 37: 2577–2585.

16.

You

Zheng

Zhong

, et al. Prognostic significance of blood urea nitrogen in acute ischemic stroke. Circ J 2018; 82: 572–578.

17.

Horiuchi

Aoki

Tanabe

, et al. A high level of blood urea nitrogen is a significant predictor for in-hospital mortality in patients with acute myocardial infarction. Int Heart J 2018; 59: 263–271.

18.

Feng

Zhou

Zou

, et al. Elevated blood urea nitrogen-to-Serum albumin ratio as a factor that negatively affects the mortality of patients with hospital-acquired pneumonia. Can J Infect Dis Med Microbiol 2019; 2019: 1547405.

19.

Matsue

van der Meer

Damman

, et al. Blood urea nitrogen-to-creatinine ratio in the general population and in patients with acute heart failure. Heart 2017; 103: 407–413.

20.

Brisco

Coca

Chen

, et al. Blood urea nitrogen/creatinine ratio identifies a high-risk but potentially reversible form of renal dysfunction in patients with decompensated heart failure. Circ Heart Fail 2013; 6: 233–239.

21.

Beier

Eppanapally

Bazick

, et al. Elevation of blood urea nitrogen is predictive of long-term mortality in critically ill patients independent of “normal” creatinine. Crit Care Med 2011; 39: 305–313.

22.

Johnson

AEW

Bulgarelli

Shen

, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data 2023; 10: 1.

23.

Sun

Wei

. Blood urea nitrogen to creatinine ratio is associated with in-hospital mortality among critically ill patients with cardiogenic shock. BMC Cardiovasc Disord 2022; 22: 58.

24.

Zhang

Xie

, et al. Machine learning-based prediction of mortality in acute myocardial infarction with cardiogenic shock. Front Cardiovasc Med 2024; 11: 1402503.

25.

Zhu

Xie

Chen

, et al. Machine learning in the prediction of in-hospital mortality in patients with first acute myocardial infarction. Clin Chim Acta 2024; 554: 117776.

26.

Liu

Zeng

. Blood urea nitrogen and in-hospital mortality in critically ill patients with cardiogenic shock: analysis of the MIMIC-III database. BioMed Res Int 2021; 2021: 5948636.

27.

Sun

Wei

. Blood urea nitrogen to creatinine ratio is associated with in-hospital mortality among critically ill patients with cardiogenic shock. BMC Cardiovasc Disord 2022; 22: 258.

28.

Zhu

Sasmita

, et al. Blood urea nitrogen for short-term prognosis in patients with cardiogenic shock complicating acute myocardial infarction. Int J Clin Pract 2022; 2022: 9396088.

29.

Yamga

Mantena

Rosen

, et al. Optimized risk score to predict mortality in patients with cardiogenic shock in the cardiac intensive care unit. J Am Heart Assoc 2023; 12: e029232.

30.

Beaini

Chunawala

Cheeran

, et al. Cardiogenic shock: focus on non-cardiac biomarkers. Curr Heart Fail Rep 2024; 21: 604–614.

31.

Lassus

. Kidney and liver dysfunction in cardiogenic shock. Curr Opin Crit Care 2020; 26: 417–423.

32.

Ren

Zhang

, et al. Role of blood urea nitrogen in predicting the post-discharge prognosis in elderly patients with acute decompensated heart failure. Sci Rep 2018; 8: 13507.

33.

Jujo

Minami

Haruki

, et al. Persistent high blood urea nitrogen level is associated with increased risk of cardiovascular events in patients with acute heart failure. ESC heart Failure 2017; 4: 545–553.

34.

Sciaccaluga

Mandoli

Ghionzoli

, et al. Risk stratification in cardiogenic shock: a focus on the available evidence. Heart Fail Rev 2022; 27: 1105–1117.

35.

Wang

Y-H

Chen

J-L

Tsai

C-S

, et al. Effects of levosimendan on systemic perfusion in patients with low interagency registry for mechanically assisted circulatory support (INTERMACS) score: experience from a single center in Taiwan. Acta Cardiologica Sinica 2021; 37: 512.

36.

Parapiboon

Kingjun

Wongluechai

, et al. Outcomes after acute peritoneal dialysis for critical cardiorenal syndrome type 1. Cardiorenal Med 2021; 11: 184–192.

37.

Smith

Shlipak

Havranek

, et al. Serum urea nitrogen, creatinine, and estimators of renal function: mortality in older patients with cardiovascular disease. Arch Intern Med 2006; 166: 1134–1142.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.06 MB

0.01 MB