Volume 16, Issue 10 p. 1748-1757
ARTICLE
Open Access

Risk factors for the development of sepsis in patients with cirrhosis in intensive care units

Yan-qi Kou

Yan-qi Kou

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Search for more papers by this author
Yu-ping Yang

Yu-ping Yang

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Department of Gastroenterology, Affiliated Hospital of Guangdong Medical University, Guangdong Medical University, Zhanjiang, China

Search for more papers by this author
Shen-shen Du

Shen-shen Du

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Search for more papers by this author
Xiongxiu Liu

Xiongxiu Liu

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Search for more papers by this author
Kun He

Kun He

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Search for more papers by this author
Wei-nan Yuan

Wei-nan Yuan

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Search for more papers by this author
Biao Nie

Corresponding Author

Biao Nie

Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, Guangzhou, China

Correspondence

Biao Nie, Department of Gastroenterology, The First Affiliated Hospital of Jinan University, Jinan University, 613 Huangpu Avenue West, Tianhe District, Guangzhou City, Guangdong, Province, China.

Email: [email protected]

Search for more papers by this author
First published: 25 May 2023

Yan-qi Kou, Yu-ping Yang, Shen-shen Du, and Xiongxiu Liu contributed equally to this work.

Abstract

Sepsis is a serious complication of liver cirrhosis. This study aimed to develop a risk prediction model for sepsis among patients with liver cirrhosis. A total of 3130 patients with liver cirrhosis were enrolled from the Medical Information Mart for Intensive Care IV database, and randomly assigned into training and validation cohorts in a 7:3 ratio. The least absolute shrinkage and selection operator (LASSO) regression was used to filter variables and select predictor variables. Multivariate logistic regression was used to establish the prediction model. Based on LASSO and multivariate logistic regression, gender, base excess, bicarbonate, white blood cells, potassium, fibrinogen, systolic blood pressure, mechanical ventilation, and vasopressor use were identified as independent risk variables, and then a nomogram was constructed and validated. The consistency index (C-index), receiver operating characteristic curve, calibration curve, and decision curve analysis (DCA) were used to measure the predictive performance of the nomogram. As a result of the nomogram, good discrimination was achieved, with C-indexes of 0.814 and 0.828 for the training and validation cohorts, respectively, and an area under the curve of 0.849 in the training cohort and 0.821 in the validation cohort. The calibration curves demonstrated good agreement between the predictions and observations. The DCA curves showed the nomogram had significant clinical value. We developed and validated a risk-prediction model for sepsis in patients with liver cirrhosis. This model can assist clinicians in the early detection and prevention of sepsis in patients with liver cirrhosis.

Study Highlights

  • WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Cirrhosis of the liver is a common condition caused by chronic liver disease, and sepsis is a serious complication of liver cirrhosis. According to this study, a risk prediction model for sepsis has been developed and validated in patients with liver cirrhosis, suggesting that early detection and prevention of sepsis is important for improving the prognosis of patients with liver cirrhosis.

  • WHAT QUESTION DID THIS STUDY ADDRESS?

A risk-prediction model for sepsis in patients with liver cirrhosis was developed and validated. In patients with liver cirrhosis, this model may be useful for early detection and prevention of sepsis.

  • WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

The study developed and validated a risk prediction model for sepsis in patients with liver cirrhosis using a large dataset. The model was based on several independent risk variables, including gender, base excess, bicarbonate, white blood cells, potassium, fibrinogen, systolic blood pressure, mechanical ventilation, and vasopressor use. Nomograms derived from the model showed excellent predictive performance, with high C-indices and area under the curve values for both training and validation cohorts, which were significantly better than Sequential Organ Failure Assessment score, Systemic Inflammatory Response Syndrome score, and Model for End-Stage Liver Disease scores. The calibration and decision curve analysis also demonstrated the clinical value of the model.

  • HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

The risk prediction model developed in this study may have significant implications for clinical practice. It can help clinicians identify patients with liver cirrhosis who are at high risk of sepsis and implement early interventions to prevent the development of sepsis. This model may also facilitate the design of clinical trials by providing a standardized approach to patient selection and risk stratification. Overall, the study highlights the importance of using data-driven approaches to develop clinically relevant risk prediction models for complex diseases.

BACKGROUND

Liver cirrhosis is a leading cause of death and morbidity worldwide in a wide range of chronic liver diseases,1 characterized by diffuse fibrosis, disruption of intrahepatic venous flow, and portal hypertension, which may result in liver failure.2 There are nearly two million deaths caused by liver cirrhosis each year, making liver cirrhosis one of the top 10 leading causes of death in the world.3, 4 Liver cirrhosis accounts for ~2% to 4% of all deaths worldwide.5 A person who suffers from cirrhosis is at a greater risk of developing liver complications on an immediate basis.6 The majority of patients with liver cirrhosis are prone to acute decompensation with organ failure, which may require admission to the intensive care unit (ICU).4 The prognosis remains poor for these patients, with in-hospital mortality rates ranging from 39% to 83%.7

Sepsis is a common complication in patients hospitalized with liver cirrhosis and is associated with high mortality and morbidity, especially in critically ill patients.8 Sepsis is defined as the dysfunction of vital organs caused by the dysregulation of the host's response to an infection.9 Whereas the liver plays an essential role during sepsis, it is also a target for sepsis-related injury, which can quickly deteriorate liver function in patients with liver cirrhosis.9, 10 Although the care of critically ill patients with sepsis is well-established in the general population, studies in patients with liver cirrhosis are limited.11 Specifically, sepsis-related mortality is four times higher in patients with liver cirrhosis, and ICU mortality is as high as 65%.12 To reduce the risk of death in critically ill patients with liver cirrhosis, it is essential to carefully assess the risk of complications related to sepsis during the diagnostic and therapeutic process.

It is difficult to predict sepsis in patients with liver cirrhosis, data on predictors of sepsis in liver cirrhosis are limited, and a new tool is needed to detect sepsis early in patients with liver cirrhosis.13 We aimed to identify the significant risk factors for sepsis in patients with liver cirrhosis from a large database along with creating and validating an independent predictive tool-nomogram for predicting sepsis in patients with liver cirrhosis.

MATERIALS AND METHODS

Data source

This is a retrospective cohort study based on the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, a large, single-center, open-access database.14 MIMIC-IV has been approved by the ethics committees of Beth Israel Deaconess Medical Center and Massachusetts Institute of Technology. We completed an online course and examination to access the database (Record ID: 40107541). The informed consent requirement was waived because the information used in the study was derived from publicly available de-identified databases.

Study population

From the MIMIC-IV database, 3256 patients admitted to the ICU with a diagnosis of liver cirrhosis were extracted. The patients included here suffer from various kinds of liver cirrhosis, including alcoholic cirrhosis, biliary cirrhosis, toxic liver disease with fibrosis and cirrhosis of the liver, and unspecified liver cirrhosis. Exclusion criteria included patients younger than 18 years of age, patients who had died within 24 h of admission to the ICU, patients who had been admitted to the ICU more than once, and patients with missing data greater than 20%. In the end, a total of 3130 patients with liver cirrhosis were selected for the study. Further, they were randomly assigned to the training cohort (n = 2141) and the validation cohort (n = 939) in a ratio of 7:3.

Data and variables

The variables extracted were demographic features, duration of hospital stay, and ICU stay, vital status at discharge, basic vital signs, laboratory parameters, severity scoring system, sepsis, renal replacement therapy (RRT) use, mechanical ventilation, and vasopressor use. The patients’ blood and biochemical tests were conducted on the first day in the ICU. Whenever there were multiple test results for a variable, the average of these values was considered.

Statistical analysis

The study used multiple imputation to account for missing data, with variables whose missing ratio was less than 20% being included. Continuous variables were expressed as the median and interquartile range, categorical variables were expressed as total number and percentage. To compare the two groups, we used t-tests, Chi-square (χ2) tests, or Wilcoxon rank-sum tests. Data dimension reduction and predictive variable selection were conducted using least absolute shrinkage and selection operator (LASSO) regression. A nomogram was constructed according to the multivariate logistic regression analysis results. Using the training cohort as the basis for the construction of the nomogram prediction model, the validation cohort was used to verify its accuracy. Nomogram discriminative ability, calibration degree, and clinical application value were validated in training and validation cohorts. A consistency index (C-index) and receiver operating characteristic curve were used to evaluate discriminative ability. The calibration curve was plotted to analyze the correlation between the observed incidence rate and the predicted probability rate. The decision curve analysis (DCA) was also used to assess the clinical net benefit of the prediction model. SPSS version 26 and R software were used for the statistical analysis. Any p < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics between training cohort and validation cohort

Based on the inclusion and exclusion criteria, a total of 3130 patients with liver cirrhosis were enrolled and randomly assigned to the training (n = 2141) and validation cohorts (n = 939) in a ratio of 7:3 (Figure 1). There were no significant differences between the two cohorts in gender, mortality, age, hospital admission time, ICU admission time, vital signs, laboratory test results, Sequential Organ Failure Assessment (SOFA) score, Systemic Inflammatory Response Syndrome (SIRS) score, Model for End-Stage Liver Disease (MELD), sepsis, RRT use, mechanical ventilation, and vasopressor use (p > 0.05). As a result, all the characteristics were equally distributed between the two cohorts (Table 1).

Details are in the caption following the image
Flow diagram of patient selection. ICU, intensive care unit; MIMIC-IV, Medical Information Mart for Intensive Care IV database.
TABLE 1. The baseline characteristics of the training and validation cohorts.
Characteristics Training (n = 2191) Validation (n = 939) p value
Demographic features
Male, n (%) 1436 (45.9%) 622 (19.9%) 0.736
Death, n (%) 435 (13.9%) 173 (5.5%) 0.380
Age, years, median (IQR) 60.16 (52.77, 68.25) 59.98 (52.74, 67.57) 0.547
Hospital admission time, days, median (IQR) 8.88 (4.95, 17.22) 9.16 (4.94, 17.02) 0.861
ICU admission time, days, median (IQR) 2.3 (1.3, 4.79) 2.55 (1.36, 5.03) 0.275
Vital signs, median (IQR)
Heart rate, bpm 85.88 (74.78, 97.21) 87 (76.62, 98.43) 0.071
SBP, mmHg 111.66 (102.68, 124.52) 112.46 (102.39, 124.31) 0.761
DBP, mmHg 61.19 (54.96, 68.94) 61.08 (54.96, 69.25) 0.944
Respiratory rate, bpm 18.16 (15.92, 20.76) 18.1 (15.83, 21.01) 0.856
Temperature, °C 36.79 (36.56, 37.04) 36.8 (36.59, 37.04) 0.354
Laboratory parameters, median (IQR)
Lactate, mmol/L 2.3 (1.43, 3.5) 2.35 (1.48, 3.68) 0.188
PH 7.38 (7.33, 7.42) 7.38 (7.33, 7.43) 0.422
BE (mmol/L) −1.5 (−4.26, 0.5) −1.5 (−4, 0.5) 0.982
Bicarbonate, mmol/L 22 (19, 24.5) 22 (19, 24.5) 0.847
Hematocrit, % 28.9 (25.45, 33.2) 28.7 (25.38, 33.05) 0.398
Hemoglobin, g/dL 9.55 (8.35, 11.05) 9.5 (8.2, 11.03) 0.521
Platelet, 109/L 107 (71.5, 161.5) 104 (70, 157.75) 0.381
WBCs, 109/L 9.2 (6.25, 13.7) 8.9 (5.97, 12.7) 0.072
BUN, mg/dL 24 (15, 43) 24 (14.5, 40.5) 0.420
Creatinine, mg/dL 1.15 (0.78, 1.95) 1.1 (0.75, 1.9) 0.605
Calcium, mg/dL 8.25 (7.75, 8.75) 8.2 (7.75, 8.75) 0.227
Chloride, mmol/L 103 (98.5, 107) 103 (98.5, 107) 0.498
Glucose, mg/dL 129.83 (107, 167.55) 131.2 (106.71, 169.17) 0.473
Sodium, mmol/L 137 (133.5, 140) 137 (134, 140) 0.806
Potassium, mmol/L 4.15 (3.75, 4.7) 4.2 (3.8, 4.75) 0.164
Albumin, g/dL 3.05 (2.6, 3.5) 3 (2.6, 3.5) 0.959
ALP, IU/L 102 (69.5, 151.62) 104 (70, 160.5) 0.401
AST, IU/L 78 (41, 189.75) 80 (42, 209.75) 0.466
Total bilirubin, mg/dL 2.6 (1.2, 6.3) 2.85 (1.2, 6.73) 0.238
CK, U/L 1557.07 (328.17, 3509.6) 1714.85 (346.82, 3552.48) 0.415
CK-MB, ng/mL 12.36 (5, 23.9) 11.86 (4.86, 24.78) 0.993
LDH, U/L 323 (206, 619.81) 329.5 (212.75, 644.58) 0.386
Fibrinogen, mg/dL 219 (157.69, 306) 227.22 (154.44, 309.94) 0.611
INR 1.6 (1.3, 2.05) 1.55 (1.3, 2) 0.080
PT, s 17.4 (14.62, 21.95) 17.25 (14.43, 21.48) 0.117
PTT, s 37.2 (31.4, 47.65) 36.55 (30.73, 47.8) 0.192
Scoring systems, median (IQR)
SOFA 7 (5, 11) 8 (5, 11) 0.629
SIRS 3 (2, 3) 3 (2, 3) 0.182
MELD 21.32 (14.00, 29.58) 21.62(14.86, 29.23) 0.903
Sepsis 0.887
No 1724 (55.1%) 736 (23.5%)
Yes 467 (14.9%) 203 (6.5%)
Therapies, n
Renal replacement therapy 183 (5.8%) 77 (2.5%) 0.944
Mechanical ventilation 330 (10.5%) 154 (4.9%) 0.371
Vasopressor use 855 (27.3%) 368 (11.8%) 0.962
  • Abbreviations: ALP, alkaline phosphatase; AST, aspartate transaminase; BE, base excess; BUN, blood urea nitrogen; CK, creatine kinase; CK-MB, creatine kinase myocardial band; DBP, diastolic blood pressure; ICU, intensive care unit; INR, international normalized ratio; IQR, interquartile-range; LDH, lactate dehydrogenase; MELD, Model for end-stage liver disease; PH, potential of hydrogen; PT, prothrombin time; PTT, partial prothrombin time; SBP, systolic blood pressure; SIRS, Systemic Inflammatory Response Syndrome; SOFA, Spatially Oriented Format for Acoustics; WBCs, white blood cells.

Predictive nomogram for the probability of sepsis in patients with liver cirrhosis

To identify potential risk factors for sepsis in patients with liver cirrhosis, the variables were first preliminarily screened using LASSO. Among the 36 associated characteristic variables related to demographics, basic vital signs, laboratory tests, and organ support interventions, 16 potential predictors have been identified based on the data from the training cohort and have nonzero coefficients in the LASSO regression model (Figure 2). We performed multiple logistic regression analysis based on LASSO results. Independent risk factors were then identified, gender (odds ratio [OR]: 0.760, 95% confidence interval [CI]: 0.585–0.990), base excess (BE; OR: 0.715, 95% CI: 0.524–0.976), bicarbonate (OR: 0.936, 95% CI: 0.666–1.315), white blood cells (WBCs; OR: 0.517, 95% CI: 0.306–0.892), potassium (OR: 0.657, 95% CI: 0.456–0.950), fibrinogen (OR: 1.271, 95% CI: 0.954–1.697), systolic blood pressure (SBP; OR: 1.230, 95% CI: 0.601–2.638), mechanical ventilation (OR: 1.411, 95% CI: 1.013–1.959), and vasopressor use (OR: 4.756, 95% CI: 3.607–6.306) were independent factors (all p < 0.05; Table 2). The above significant factors were incorporated into a nomogram based on multiple logistic regression analysis to predict the probability of patients with liver cirrhosis developing sepsis (Figure 3).

Details are in the caption following the image
LASSO regression for variables selection. (a) The associations between the coefficients of variables and the log lambda value. With increasing log lambda, each variable's coefficient tended towards zero, with each line representing a distinct variable. (b) By verifying the optimal parameter (λ) in the LASSO model, the partial likelihood binomial deviation is plotted versus log (λ). The dotted line is determined using the minimum criterion and the one standard error of the minimum criteria (the 1 − SE criteria) when selecting the optimal values log (λ) of the features. Our study chose the lambda value using the 1 − SE criteria. LASSO, least absolute shrinkage and selection operator.
TABLE 2. Multivariate logistic analysis.
Variables OR 95% CI p value
Gender 0.760 0.585–0.990 0.041
BE 0.715 0.524–0.976 0.035
Bicarbonate 0.936 0.666–1.315 0.014
WBCs 0.517 0.306–0.892 0.015
Potassium 0.657 0.456–0.950 0.024
Fibrinogen 1.271 0.954–1.697 <0.001
SBP 1.230 0.601–2.638 0.043
Mechanical ventilation 1.411 1.013–1.959 0.040
Vasopressor use 4.756 3.607–6.306 <0.001
  • Abbreviations: BE, base excess; CI, confidence interval; OR, odds ratio; SBP, systolic blood pressure; WBCs, white blood cells.
Details are in the caption following the image
Nomogram for predicting the probability of sepsis in patients with liver cirrhosis. Every variable level was assigned a score. All of the scores for the selected variables were added to obtain the total score. The probability corresponding to the total score was the probability of that patient developing sepsis. BE, base excess; SBP, systolic blood pressure; WBC, white blood cells.

Validation of the nomogram model

The C-indexes for the training and validation cohorts were 0.814 (95% CI: 0.793–0.835) and 0.828 (95% CI: 0.798–0.858), respectively. The area under the curve (AUC) values of the nomogram in the training and validation cohorts were 0.849 (95% CI: 0.830–0.867) and 0.821 (95% CI: 0.789–0.852), respectively, which were significantly better than the SOFA, SIRS, and MELD scores (Figure 4). The training cohort optimal cutoff point was 0.173 and its corresponding sensitivity and specificity were 0.833 and 0.709. In the validation cohort, the optimal cutoff point was 0.177, for which the sensitivity and specificity were 0.823 and 0.723, respectively. According to the results, our nomogram demonstrated better discrimination abilities and was superior to the commonly used scoring methods. Calibration curves for the training and validation cohorts indicated a high degree of agreement between predictions and observations (Figure 5). DCA demonstrated that the threshold probability for the prediction model was 0.04–0.74 in the training cohort and 0.04–0.76 in the validation cohort (Figure 6). Within the wide and practical range of threshold probabilities, the nomogram demonstrated a superior overall net benefit, suggesting high clinical potential for the nomogram. Further, our study found that the nomogram had some predictive value for mortality in patients with liver cirrhosis and sepsis, with an AUC of 0.639 (95% CI: 0.563–0.715) in the training group and 0.648 (95% CI: 0.563–0.715) in the validation group. The optimal cutoff point for the training cohort was 0.186, with a sensitivity and specificity of 0.286 and 0.929; the optimal cutoff point for the validation cohort was 0.163, with a sensitivity and specificity of 0.712 and 0.516, respectively (Figure S1).

Details are in the caption following the image
ROC curves of the nomogram, SOFA, SIRS, and MELD scores. (a) Training cohort; (b) validation cohort. AUC, area under the curve; MELD, Model for End-Stage Liver Disease; ROC, receiver operating characteristic; SIRS, Systemic Inflammatory Response Syndrome SOFA, Sequential Organ Failure Assessment score.
Details are in the caption following the image
Calibration curves for the nomogram. (a) Training cohort and (b) validation cohort.
Details are in the caption following the image
Decision curve analysis for the nomogram. (a) Training cohort and (b) validation cohort.

DISCUSSION

This study revealed that gender, BE, bicarbonate, WBC, potassium, fibrinogen, SBP, mechanical ventilation, and vasopressor use were independent risk factors for sepsis in patients with liver cirrhosis. Based on these indicators, a nomogram was constructed to assess the risk of sepsis in patients with liver cirrhosis during hospitalization. For the first time, this study provided a relatively accurate tool to predict the risk of sepsis among patients with liver cirrhosis, which demonstrated good discrimination and calibration power. There were three noteworthy features in our study: (1) a large number of patients were included in the study; (2) the variables used in the model can easily be obtained in the clinical setting; and (3) the nomogram demonstrated greater predictive performance in both training and validation cohorts.

Many experimental and clinical studies have demonstrated that sepsis and infectious diseases differ according to gender.15 Female patients were twice as likely to have concomitant sepsis as male patients,16 and a Kenyan study revealed that bacteremia occurred at significantly higher rates in female patients than in male patients.17 Furthermore, women with cirrhosis had more complications than men.18 The difference in immune response may be contributed to the greater risk of infectious complications in women.19 Although data were demonstrating that trauma-injured men are more likely to develop sepsis when compared with women, owing to hormonal differences between the sexes.20 Hence, gender should be a valuable factor to predict sepsis risk among patients with liver cirrhosis.

The new predictive scoring system also included five laboratory test indicators for BE, bicarbonate, WBC, potassium, and fibrinogen that can be easily accessed. A dysfunctional liver can result in a variety of complex metabolic acid–base disorders.21 This fragile equilibrium is often tilted toward metabolic acidosis when patients with liver cirrhosis complicate sepsis, which was consistent with our findings.22 Researchers have demonstrated that BE and bicarbonate can be used to assess the risk of death and morbidity in critically ill patients with suspected sepsis in previous studies.23, 24 In addition, WBC was an indicator of inflammation that predicted poor outcomes as well as dysregulated leukocyte response that contributes to sepsis pathophysiology.25 Liver cirrhosis severity has been shown to be correlated with serum potassium concentration.26 As part of the prediction model, potassium levels were also included, and elevated or decreased levels indicate severe abnormalities within the body. Thus, serum potassium may also be a valuable predictor of sepsis in patients with liver cirrhosis. Fibrinogen is a glycoprotein produced by the liver and circulating in the blood, which is used to diagnose coagulopathy in critically ill patients.27 As an acute-phase reactant, fibrinogen is elevated in patients with infection and/or inflammation, and increased baseline fibrinogen levels are associated with an increased risk of sepsis.28 As can be seen from the nomogram, SBP and vasopressor use are weighted more heavily, and the risk of sepsis in patients with cirrhosis gradually increases with the decrease in SBP and the use of vasopressor. It suggests that the patient has experienced a decrease in blood pressure and needs pharmacological treatment to improve vascular function and microcirculatory perfusion.29 The majority of patients with sepsis suffer from hypotension and inadequate tissue perfusion.30 Mechanical ventilation is a major indication of the need to admit patients with liver cirrhosis to the ICU and is an independent risk factor for ICU mortality in patients with cirrhosis.31, 32

Sepsis affects millions of people each year globally, and it is of clinical importance to develop a predictive model predicting sepsis in patients with liver cirrhosis.33 Based on the patients’ laboratory findings, vital signs, and treatment in ICU, we constructed the first model that predicts the likelihood of sepsis in patients with liver cirrhosis. The purpose of this is to provide a basis for the professional treatment of patients suffering from liver cirrhosis, help patients understand the severity of the disease, and improve their cooperation in order to minimize the chances of sepsis occurring. Although the models developed in this study to predict sepsis in patients with liver cirrhosis demonstrated good performance and applicability, there are inevitable limitations to our study. First, because MIMIC-IV is a public database of single centers in the United States, our study is limited in its generalizability due to selection bias. Second, our study is based on a retrospective cohort study, thus the nomogram will need to be prospectively validated before it can be considered for clinical use.

We conclude that independent risk factors for sepsis in patients with liver cirrhosis have been identified based on LASSO regression and multivariate logistic analysis, including gender, BE, bicarbonate, WBC, potassium, fibrinogen, SBP, mechanical ventilation, and vasopressor use, and used them to develop a risk prediction model for sepsis in patients with liver cirrhosis. This study may provide clinical reference information for the prevention of sepsis in patients with liver cirrhosis.

AUTHOR CONTRIBUTIONS

Y.Q.K., Y.P.Y., S.S.D., X.X.L., and B.N. wrote the manuscript and designed the research. Y.Q.K., Y.P.Y., S.S.D., X.X.L., K.H., and W.N.Y. performed the research and analyzed the data.

ACKNOWLEDGMENTS

Gratitude is extended to all the staff members involved in the creation of the MIMIC-IV database.

    FUNDING INFORMATION

    No funding was received for this work.

    CONFLICT OF INTEREST STATEMENT

    The authors declared no competing interests for this work.