Metoprolol use is associated with improved outcomes in patients with sepsis-induced cardiomyopathy: an analysis of the MIMIC-IV database

Abstract

Background

Metoprolol is commonly administered to critically ill patients; however, its effect on mortality in patients with sepsis-induced cardiomyopathy (SICM) remains uncertain. This study aimed to investigate the relationship between metoprolol use and mortality in patients with SICM.

Methods

Adults with SICM were identified from the MIMIC-IV database. The exposure of interest was metoprolol treatment. The outcomes assessed were 30-day mortality, 1-year mortality, and in-hospital mortality. Kaplan–Meier survival analysis evaluated the effect of metoprolol on these outcomes. Multivariable Cox proportional hazards and logistic regression analyses were performed to determine the correlation between metoprolol treatment and mortality in patients with SICM.

Results

1163 patients with SICM were identified, with 882 receiving metoprolol treatment (MET group) and 281 not receiving metoprolol treatment (NOMET group). Overall, the 30-day, 1-year, and in-hospital mortality rates were 10.2%, 18.2%, and 8.9%, respectively. Significant differences in mortality existed between the groups. Multivariable Cox analysis revealed that patients in the NOMET group had a higher risk of 1-year mortality (adjusted hazard ratio [HR] 2.493; 95% confidence interval [CI] 1.800–3.451; P < 0.001) and 30-day mortality (adjusted HR 4.280; 95%CI 2.760–6.637; P < 0.001). Metoprolol treatment was associated with lower in-hospital mortality (odds ratio [OR] 5.076; 95% CI 2.848–9.047; P < 0.001). Subgroup analysis supported these findings.

Conclusion

Metoprolol treatment is associated with reduced all-cause mortality in patients with SICM. Prospective studies are required to validate these findings.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-024-04271-6.

Introduction

Sepsis-induced cardiomyopathy (SICM) is characterized by reversible myocardial depression occurring in the early stages of septic shock. The incidence of SICM varies widely, from 10 to 70%, with risk factors including younger age, male gender, and a history of heart failure [1–3]. Despite its reversible nature, Blanco et al. [4] demonstrated that myocardial dysfunction in patients with sepsis exhibits a mortality rate exceeding 50% compared to those without cardiac dysfunction [5–7]. Patients typically exhibit an ejection fraction (EF) < 50%, with a decrease of > 10% compared to baseline, frequently recovering within two weeks [1, 8]. Unlike coronary artery obstructive disease, SICM does not result in extensive myocardial necrosis and presents as myocardial depression due to disturbances in myocardial cell energy metabolism or direct injury.

Although numerous studies have focused on SICM, specific clinical management guidelines remain sparse. Catecholamines are critical in myocardial inflammatory injury and apoptosis during septic shock [9, 10]. Treatment for SICM focuses on cardiac protection within the broader context of septic shock management, as excessive activation of both exogenous catecholamines and the endogenous sympathetic system can lead to sympathetic storms, causing direct myocardial damage [11, 12]. Combining milrinone (a non-catecholamine cardiotonic agent) and oral metoprolol can enhance cardiac index and reduce arrhythmias. β-adrenergic blockers (β-blockers) are particularly valuable in patients who are critically ill for their cardioprotective properties [13]. β-blockers benefit inflammatory response, cardiac dysfunction, and immune dysfunction in septic shock [14–18]. β-blockers can mitigate myocardial damage during sepsis by reducing heart rate (HR) and oxygen consumption, inhibiting inflammatory responses, and preventing myocardial apoptosis [19]. However, the impact of β-blockers on patients with SICM is complex due to their negative inotropic effect on the myocardium. While β-blockers can reduce the incidence of myocardial injury, they do not significantly improve survival rates [20].

The long-term prognosis of patients with SICM treated with β-blockers remains uncertain. This study aimed to investigate the relationship between β-blocker use and outcomes in patients with SICM by analysing data from the Medical Information Mart for Intensive Care (MIMIC-IV) database. β-blocker treatment was hypothesized to be an independent protective factor for patients with SICM.

Methods and materials

Data source and ethics

The dataset was sourced from the publicly available MIMIC-IV (Version 2.2, from PhysioNet) database [21, 22]. The database has received approval from the Institutional Review Board (IRB) at the Beth Israel Deaconess Medical Centre (BIDMC) and the Massachusetts Institute of Technology (MIT) [23]. The primary investigator (Yuanjun Zhou) completed the Collaborative Institutional Training Initiative (certification number. 39149215) and signed the data use agreement to access the database. As this study is retrospective, with all information de-identified, the IRB approved data sharing without requiring informed consent.

Study population

All adult patients with SICM were identified from the database between 2008 and 2019. SICM identification was based on criteria from previous studies [2, 3, 24–29]: [1] Diagnosis of sepsis or septic shock [30–32]; [2] Left ventricular (LV) systolic dysfunction (LVEF < 50%) or global LV hypokinesis/systolic dysfunction reported in echocardiographic data; [3] Exclusion of concomitant cardiac diseases.

When patients had multiple intensive care unit (ICU) admissions, only data from the first admission were extracted. To avoid the confounding factors associated with long-term ICU stays and ensure consistency in baseline characteristics, only patients who started metoprolol use within 6 h before ICU admission were included. Patients were segregated into the metoprolol treatment group (MET group) and the non-metoprolol treatment group (NOMET group).

The exclusion criteria were: [1] ICU stay < 24 h; [2] Pre-existing severe cardiac diseases, including primary cardiomyopathy, heart failure, or coronary artery disease, identified through discharge diagnoses based on the International Classification of Diseases (version 9 and 10) codes; [3] Treatment with other β-receptor blockers; [4] Metoprolol treatment outside the pre-defined time range; [5] Missing medical records or LOS of hospital < 2 days.

Data collection

Baseline information included the following aspects: [1] Prescription information of metoprolol usage; [2] Demographic characteristics, such as age, gender, and race; [3] Disease severity scores at admission, including Sequential Organ Failure Assessment (SOFA) score and Simplified Acute Physiology Score (SAPS) II; [4] Comorbidities included hypertension, atrial fibrillation (AF), cerebrovascular disease, diabetes, renal disorder, hepatic disorder, and cancer; [5] Initial vital signs included heart rate (HR), mean blood pressure, body temperature, and oxygen saturation; [6] Initial laboratory results included white blood cell count (WBC), haemoglobin, red blood cell distribution width-to-platelet ratio (RPR), bicarbonate, creatinine, and blood glucose levels; [7] Initial therapies included the use of vasopressors, renal replacement therapy (RRT), mechanical ventilation (MV), and dexmedetomidine. Vasopressors included dopamine, dobutamine, epinephrine, norepinephrine, and milrinone; [8] Culture results. Data were extracted from MIMIC-IV v2.2 using Structured Query Language (SQL) and PostgreSQL 15.0. All baseline data were collected within the first 24 h after ICU admission.

Outcomes

The primary outcomes were 1-year and 30-day mortality. The secondary outcome was in-hospital mortality.

Statistical analysis

Categorical variables were expressed as numbers (percentages) and compared using chi-square tests. All continuous variables were reported as median [interquartile range] due to non-normal distribution (Kolmogorov–Smirnov test) and compared using the Mann–Whitney U test. Baseline information was compared based on drug use and outcome grouping. Kaplan–Meier survival analysis (log-rank test) was conducted to calculate the 30-day and 1-year mortality incidence for the MET and NOMET groups. Binary logistic regression and Cox proportional hazards models, both unadjusted and adjusted, were employed to determine the odds ratio (OR) and hazard ratio (HR) with 95% confidence intervals (CI) between metoprolol exposure and outcomes. Variables with a p-value < 0.05 in the univariate analysis and those related to prognosis were included as confounding variables in the multivariate model. Model 1 was the unadjusted model; Model 2 included age, gender, race, and comorbidities based on Model 1; Model 3 was further adjusted for disease scores, laboratory tests, and treatment measures. Furthermore, we investigated the association between the timing of metoprolol use and outcomes, including two subsets of patients: one subset that received metoprolol immediately after admission to ICU and another subset that started using metoprolol 48 h after ICU admission. Subgroup analysis was conducted to assess the consistency of metoprolol’s effect on the primary outcome across different populations, including gender, age (≤ 65 years and > 65 years), history of hypertension, history of AF, and SOFA score (≤ 5 and > 5). Interactions between drug exposure and stratification variables were also tested. A two-tailed p < 0.05 was considered statistically significant. Statistical analyses were performed using R software (version 4.0.2).

Report guideline

The study adhered to the REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) Statement [33].

Results

Population

A total of 1163 patients with SICM were included (Fig. 1) in this study. The overall median age was 67 years (interquartile range [IQR]: 58–76 years), with 372 individuals (32.0%) being women (Table 1). The median SOFA score was 5 (IQR: 4–8). Of the participants, 882 individuals (75.8%) received metoprolol therapy (MET group), while 281 (24.2%) did not receive metoprolol or any other β-blockers (NOMET group).

Fig. 1.

The flowchart of the patient selection process. ICU: intensive care units; LOS: length of stay; MI: myocardial infarction; HF: heart failure; SICM: sepsis-induced cardiomyopathy. All data process codes were provided in Additional Files 1

Table 1.

Baseline characteristics and outcomes of SICM patients

Variables	Overall(n = 1163)	NOMET group(n = 281)	MET group(n = 882)	p
Age(years)	67[58, 76]	64.00 [50, 75]	68.00 [60, 76]	< 0.001
Gender (woman)	372 (32.0)	132 (47.0)	240 (27.2)	< 0.001
Ethnicity				0.015
White	823 (70.8)	178 (63.4)	645 (73.1)
Black	62 (5.3)	21 (7.5)	41 (4.7)
Asian	38 (3.3)	12 (4.3)	26 (3.0)
Others	240 (20.6)	70 (24.9)	170 (19.3)
Disease scores
SOFA	5 [4, 8]	6 [4, 9]	5 [4, 7]	< 0.001
SAPSII	37 [30, 46]	39 [31, 49]	36 [29, 45]	0.019
Comorbidities
Hypertension	649 (55.8)	116 (41.3)	533 (60.4)	< 0.001
Atrial fibrillation	407 (35.0)	43 (15.3)	364 (41.3)	< 0.001
Cerebrovascular disease	167 (14.4)	33 (11.7)	134 (15.2)	0.151
COPD	287 (24.7)	92 (32.7)	195 (22.1)	< 0.001
Diabetes	342 (29.4)	57 (20.3)	285 (32.3)	< 0.001
Renal disease	162 (13.9)	45 (16.0)	117 (13.3)	0.246
Liver disease	111 (9.5)	57 (20.3)	54 (6.1)	< 0.001
Cancer	112 (9.6)	46 (16.)	66 (7.5)	< 0.001
Initial vital signs
Heart Rate (beats/min)	83 [76, 93]	87 [75, 101]	82 [76, 91]	0.002
MBP (mmHg)	75 [70, 79]	75 [70, 81]	75 [71, 79]	0.264
Temperature (℃)	36.8 [36.6, 37.1]	36.9 [36.6, 37.2]	36.8 [36.6, 37.0]	< 0.001
SpO2	98 [96, 99]	97 [95, 98]	98 [97, 99]	< 0.001
Initial laboratory results
WBC (109/L)	12.4 [9.3, 16.1]	11.70 [7.8, 17.4]	12.60 [9.8, 16.0]	0.030
Hemoglobin (g/dl)	10.5 [9.4, 11.7]	10.70 [9.4, 12.7]	10.40 [9.43, 11.6]	0.014
RPR	0.09 [0.07, 0.11]	0.08 [0.06, 0.12]	0.09 [0.07, 0.11]	0.042
Bicarbonate (mmol/L)	23.0 [21.0, 25.0]	21.7 [19.0, 24.5]	23.3 [21.5, 25.0]	< 0.001
Creatinine (mg/dl)	0.9 [0.8, 1.3]	1.00 [0.7, 1.6]	0.90 [0.8, 1.2]	0.016
Blood glucose (mg/dl)	124 [108, 149]	135 [114, 176]	121 [106, 143]	< 0.001
Initial therapies
RRT	51 (4.4)	19 (6.8)	32 (3.6)	0.025
MV	653 (56.2)	142 (50.5)	511 (57.9)	0.029
Vasopressors	452 (38.9)	157 (55.9)	295 (33.5)	< 0.001
Dexmedetomidine	84 (7.2)	7 (2.5)	77 (8.7)	< 0.001
Culture-positive	106 (9.1)	47 (16.7)	59 (6.7)	< 0.001
Outcomes
In-hospital mortality	103 (8.9)	66 (23.5)	37 (4.2)	< 0.001
1-year mortality	212 (18.2)	96 (34.2)	116 (13.2)	< 0.001
30-day mortality	118 (10.2)	72 (25.6)	46 (5.2)	< 0.001
LOS of Hospital	7.3 [5.1, 12.7]	8.1 [5.2, 15.3]	7.1 [5.1, 11.8]	0.122
LOS of ICU	2.5 [1., 5.1]	3.9 [2.2, 7.2]	2.2 [1.3, 4.2]	< 0.001

Count data are presented as numbers(percentage) and median[quartile] for non-normal continuous data. SICM: sepsis-induced cardiomyopathy; MET group: patients with metoprolol exposure; NOMET group: patients without metoprolol exposure; SOFA: Sequential Organ Failure Assessment; SAPSII: Simplified Acute Physiology Score; COPD: chronic obstructive pulmonary diseases; WBC: white blood cells; RPR: red blood cell distribution width to platelet ratio; MBP: mean blood pressure; RRT: renal replacement therapies; MV: mechanical ventilation; Vasopressors included dopamine, dobutamine, epinephrine, norepinephrine, and milrinone; LOS: length of stay

Baseline characteristics

Table 1 shows significant between-group differences in most baseline characteristics. The NOMET group consisted of younger patients and had higher SOFA and SAPS II scores, with a more substantial proportion being female. Conversely, patients in the MET group had a higher incidence of hypertension and AF. Fewer patients in the MET group received vasopressors and RRT; however, more patients received MV. There was no significant difference in cerebrovascular disease between the groups. The NOMET group exhibited higher in-hospital, 30-day, and 1-year mortalities than the MET group (23.5% vs. 4.2%, 34.2% vs. 13.2%, 25.6% vs. 5.2%, respectively, all p < 0.001). Moreover, patients in the MET group had shorter ICU and total hospital stays (7.1 days vs. 8.1 days, 2.2 days vs. 3.9 days, respectively, p < 0.001).

Baseline information between survivors and non-survivors

Additional File 2 illustrates an imbalance in the distribution of confounding variables among survivors and non-survivors, stratified by different mortality outcomes. The survival group had a higher rate of metoprolol treatment than the non-survival group, with treatment rates of 80.0% vs. 39.0% for 30-day survivors and non-survivors; and 80.6% vs. 54.7% when stratified by 1-year mortality. Overall, individuals in the survival group had lower severity scores, RPR, creatinine, and blood glucose levels, a lower incidence of comorbidities, higher haemoglobin levels, and a higher usage rate of dexmedetomidine. However, a higher proportion of patients in the survival group had a history of hypertension compared to the non-survival group. Additional details regarding the baseline information of in-hospital survivors and non-survivors are provided in Additional File 3.

Outcomes

The Kaplan–Meier analysis revealed significant differences in the primary outcomes between the MET and NOMET groups (Fig. 2). The NOMET group had higher 30-day and 1-year mortality rates (both log-rank p < 0.001). The association between metoprolol treatment and outcomes was further evaluated using Cox proportional hazard analyses (Table 2). The unadjusted, partially adjusted, and fully adjusted models consistently indicated an independent association between metoprolol use and higher 30-day and 1-year survival rates, demonstrating robust results. Logistic regression analysis showed an independent association between metoprolol use and lower in-hospital mortality (Table 2). Furthermore, two patient subsets were analyzed: one subset received metoprolol treatment immediately after ICU admission, and the other started using metoprolol 48 h after ICU admission. The results showed that in both subsets, patients in the NOMET group exhibited higher in-hospital mortality, 30-day mortality, and 1-year mortality than patients in the MET group (Additional File 4).

Fig. 2.

Kaplan-Meier analysis for metoprolol exposure and 30-day/1-year mortality. Figure 2a Cumulative incidence of 30-day mortality and the number at risk. Figure 2b Cumulative incidence of 1-year mortality and the number at risk. MET group: patients with metoprolol exposure; NOMET group: patients without metoprolol exposure

Table 2.

The association between all-cause mortality and metoprolol exposure

	Model 1		Model 2		Model 3
Primary outcomes	HR(95%CI)	p	HR(95%CI)	p	HR(95%CI)	p
1-year mortality
MET group	Reference	< 0.001	Reference	< 0.001	Reference	< 0.001
NOMET group	3.143(2.398–4.120)		2.270(1.688–3.051)		2.493(1.800-3.451)
30-day mortality		< 0.001
MET group	Reference		Reference	< 0.001	Reference	< 0.001
NOMET group	5.627(3.886–8.147)		4.211(2.813–6.304)		4.280(2.760–6.637)
Secondary outcome	OR(95%CI)	p	OR(95%CI)	p	OR(95%CI)	p
In-hospital mortality
MET group	Reference	< 0.001	Reference	< 0.001	Reference	< 0.001
NOMET group	7.011(4.563–10.771)		7.169(4.455–11.536)		5.076(2.848–9.047)

SICM: sepsis-induced cardiomyopathy; MET group: patients with metoprolol exposure; NOMET group: patients without metoprolol exposure; HR: hazard ratio; OR: odds ratio; 95%CI: 95%confidence interval; Model 1 was the unadjusted model; Model 2 included age, gender, race, and disease scores; Model 3 included age, gender, race, disease scores, comorbidities, vital signs, laboratory tests, treatment measures, and culture results

Subgroup analysis

Most subgroup analysis results were consistent with the primary analysis (Fig. 3). Metoprolol was significantly associated with higher 1-year and 30-day survival rates across various subgroups. However, no significant association between metoprolol and 1-year mortality was found in the culture-positive and SOFA ≤ 5 groups. Most subgroup analyses did not demonstrate considerable interaction effects except for the gender subgroups regarding 1-year mortality (p for interaction = 0.030).

Fig. 3.

Subgroup analysis and interactive effect for 1-year/30-day mortality. Patients were grouped regarding age, gender, hypertension, AF, and SOFA score. Figure 3a Subgroup analysis for 30-day mortality. Figure 3b Subgroup analysis for 1-year mortality. SICM: sepsis-induced cardiomyopathy; MET group: patients with metoprolol exposure; NOMET group: patients without metoprolol exposure; HR: hazard ratio; 95%CI: 95% confidence interval; p for int: p for interactive effect; AF: atrial fibrillation; SOFA score: Sequential Organ Failure Assessment score

Discussion

This study explored the impact of metoprolol treatment on mortality in patients with SICM. Our findings demonstrate an association between metoprolol administration and lower mortality in patients with SICM. Our findings hold significant implications, suggesting that β-blocker treatment in patients with SICM is not an absolute contraindication. To the best of our knowledge, limited studies have explored the effects of β-blockers on patients with SICM.

SICM results from multifactorial and multi-pathway involvement, predominantly characterized by functional changes leading to myocardial suppression. While most patients gradually recover normal cardiac function after disease control, the specific mechanisms remain unclear. Excessive catecholamines and sympathetic nervous system stimulation during sepsis are essential factors, adversely affecting immune response, energy metabolism, cardiotoxicity, and myocardial structural integrity. Overstimulation of β-adrenergic receptors can exacerbate these harmful effects [34]. Therefore, the use of β-blockers in sepsis is proposed to mitigate excessive adrenergic responses, thereby alleviating cardiac and metabolic impairments [35]. The present study demonstrated a substantially higher survival rate in patients with SICM treated with metoprolol, consistent with previous studies. Metoprolol and esmolol can improve short-term survival in patients with sepsis without impairing myocardial contractile function [36–39]. However, there is controversy regarding the effectiveness of β-blockers in treating SICM. Although β-blockers can alleviate sepsis-related myocardial damage, they do not necessarily reduce in-hospital mortality [20]. This inconsistency may be due to the high heterogeneity of the patients regarding sepsis aetiology, infection severity, and treatment regimens. Further research is needed to clarify the optimal use of β-blockers in treating SICM. Although most subgroup results are consistent with the primary analysis, the interpretation of post-hoc subgroup analyses should be approached with caution due to the small sample sizes within subgroups. When evaluating the impact of metoprolol treatment on the prognosis of different patients, demographic characteristics (such as gender), severity of illness, comorbidities, and other treatments can not be overlooked, as these factors may influence the effect of metoprolol treatment.

Many studies have shown that β-blockers are essential in improving cardiac function. Metoprolol, a highly selective β1-blocker, has been found to exert cardioprotective and anti-inflammatory effects. β1-blockers can mitigate myocardial injury [6, 40], reduce myocardial sensitivity to catecholamines, counteract catecholamine-induced high metabolic turnover, and improve left ventricular remodelling [41]. Furthermore, β1-blockers can prolong diastolic time, inhibit myocardial contractility, and increase coronary artery perfusion, thereby reducing the risk of myocardial ischemia, improving cardiac output and blood lactate levels, and reducing damage to other organs [42–45]. Metoprolol does not block β2-adrenergic receptors, thereby preserving the cardioprotective effects of β2-adrenergic receptors [46, 47]. Selective stimulation of the β2-adrenergic receptors may enhance anti-apoptotic effects and cardiac contractility. Conversely, β1-blockers can inhibit the activation of multiple inflammation and apoptosis signalling pathways and the expression of inflammatory mediators [48–50], thereby protecting the cardiac function of patients with SICM. Moreover, metoprolol has been shown to enhance coagulation function and immune function in patients with sepsis [11] and improve respiratory and central nervous system scores in the SOFA [39].

Despite the cardioprotective effects of β-blockers, clinicians should be aware of the safety concerns associated with their adverse myocardial effects. Balancing HR reduction and myocardial contractility impairment caused by β-blockers with tissue microcirculation perfusion is essential in clinical practice. Microcirculation perfusion warrants more significant attention than hypotension [51]. Schmittinger et al. found that combining metoprolol with milrinone can significantly increase stroke volume and reduce blood lactate levels in patients with sepsis, dramatically improving cardiac function and prognosis [52]. Moreover, the combined use of β-blockers and vasoconstrictors does not increase the demand for vasoconstrictors or harm microcirculation [53, 54]. Clinicians should consider individualized adjustments of β-blocker therapy based on varying patient conditions when treating SICM.

Although the main analysis suggests that metoprolol is independently associated with a better prognosis for patients with SICM, the interpretation of results should be approached cautiously. Patients in the NOMET group generally have more severe conditions and require more frequent use of vasopressors, which may cause doctors to avoid β-blockers due to their adverse effects on cardiac contractility and HR. This medical decision could affect the representativeness of the cohort. The concordance between the subset analysis and the main analysis results indicates that metoprolol is associated with a better prognosis regardless of whether treatment is initiated early or slightly later. Given the retrospective nature of this study, however, we can not definitively determine the optimal timing for metoprolol initiation. Well-designed randomized controlled trials are needed to investigate the therapeutic timing, appropriate dosage, and duration of metoprolol treatment in patients with SICM. The potential benefits of β-blockers for critically ill patients warrant further exploration.

Limitations

Firstly, patients were not randomized, as physicians decided whether to administer β-blockers based on patient conditions, leading to potential patient selection bias and imbalanced baseline characteristics. Hence, multivariable-adjusted models were used to evaluate the independent effect of metoprolol on the outcomes. Secondly, unmeasured confounding factors, such as the use of levosimendan, may still influence the prognosis. We should comprehensively consider the potential combined effects when using β-blockers concomitantly with other medications. Thirdly, patients without echocardiographic data were excluded. While this approach ensured the standardization and consistency of the study population, it may have resulted in data loss from some potential patients. Therefore, further randomized controlled trials are necessary to ensure the reliability of these findings.

Conclusion

Administration of metoprolol is associated with decreased all-cause mortality in patients with SICM. Further prospective trials are required to provide high-level evidence regarding the mechanisms, optimal dosing, and timing of β1-blocker administration.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

SICM

Sepsis-induced cardiomyopathy

EF

Ejection fraction

MIMIC-IV

Medical Information Mart for Intensive Care IV

IRB

Institutional Review Board

LV

Left ventricular

ICU

Intensive care unit

SOFA

Sequential Organ Failure Assessment

SAPS II

Simplified Acute Physiology Score II

AF

Atrial fibrillation

RPR

Red blood cell distribution width-to-platelet ratio

RRT

Renal replacement therapy

MV

Mechanical ventilation

WBC

White blood cell count

HR

Hazard ratio

SQL

Structured Query Language

OR

Odds ratio

CI

Confidence interval

Author contributions

Zh.LP: study design, data collection and examination, data analysis, and manuscript drafting, manuscript revision; Zh.YT: data examination and analysis, manuscript drafting, and supervision of the study process, manuscript revision; L.YL: data examination and analysis and the supervision of the study process, manuscript revision; ZH.YJ: study design, data collection, and examination, data analysis, manuscript drafting, manuscript revision and supervision of the study process.

Funding

The research did not receive any external funding.

Data availability

The dataset was based on the publicly available, open-access MIMIC-IV (Version 2.2). And, the corresponding author can grant data access to this study upon request.

Declarations

Ethics approval and consent to participate

The Institutional Review Boards of both the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC) have approved this study and the sharing of the research resource (the NO. of official certification 39149215) and waived the requirement of informed consent due to retrospective design.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Patient consent for publication

Informed consent was not required as the study design was retrospective and data was anonymized.