Analysis of Risk Factors For Sepsis And Indicators of Prognosis In Patients With Sepsis: A Retrospective Observational Study

Background: To investigate the independent risk factors for sepsis and the prognostic indicators of sepsis-related mortality to guide clinical practice. Methods: Adult patients diagnosed with sepsis in the respiratory intensive care unit (RICU), emergency ICU (EICU), and surgical ICU (SICU) of Beijing Chao-Yang Hospital, Capital Medical University, from January 2016 to April 2021 were enrolled. Comorbidities, complications, and laboratory indicators were retrospectively analyzed. Variables with a p value < 0.05 in the univariate analysis were entered into multivariate logistic regression analysis to identify the independent risk factors for sepsis. Receiver operating characteristic curve (ROC) analysis was used for those variables with P < 0.05 in multivariate regression to evaluate the t of the predictive model and its prognostic ecacy. Results: A total of 123 adult patients with sepsis were enrolled, with 80 males and 43 females and a mean age of 61.56 ± 17.12 years. Acute respiratory distress syndrome (ARDS) occurred in 84 patients (68.3%), acute kidney injury (AKI) occurred in 28 patients (22.8%), acute myocardial injury (AMI) occurred in 6 patients (4.9%), disseminated intravascular coagulation (DIC) occurred in 14 patients (11.4%), septic shock occurred in 40 patients (32.5%), and 41 patients (33.3%) died. Multivariate logistic regression analysis showed that mean arterial pressure (MAP), acute physiology and chronic health evaluation II (APACHE II) score, albumin level, and the presence of DIC were independent risk factors for sepsis (P < 0.05). The area under the ROC curve for the model including MAP, albumin, and APACHE II score was the highest at 0.890. Conclusion: The MAP, APACHE II score, albumin level, and DIC were independent risk factors for sepsis. The inclusion of the MAP, albumin level, and APACHE II score in the model yielded the most accurate prediction of the risk of mortality. The ROC of MAP, ALB, and APACHE II for ability to evaluate the prognose . The AUC is 0.747 (95% CI 0.654-0.839) for MAP, 0.641 (95% CI 0.541-0.741) for ALB, 0.839 (95% CI 0.760-0.917) for APACHE II, and 0.890 (95% CI 0.832-0.948) for the combination of them. The AUC value of APACHE II is signicantly higher than of both MAP (p < 0.001) and ALB (p = 0.011). The AUC value of the combination of them is the highest. Abbreviations: MAP, mean arterial pressure; APACHE II, acute physiology and chronic health evaluation; ALB, albumin; ROC, receiver operating characteristic curve; AUC, area under the curve; CI, condence interval.


Introduction
Sepsis is a common and often fatal syndrome characterized by physiological, biological, and biochemical abnormalities caused by dysregulated in ammatory responses to infection. [1,2] Sepsis remains the leading cause of death in patients admitted to the intensive care unit (ICU). [3] The related mortality rate remains high worldwide, [4] especially in cases of septic shock. [5] Carolin et al performed a meta-analysis of 27 studies in high-income countries and found that the incidence of sepsis was 288 cases per 100,000 person-years. A multicenter cohort study in Korea reported that the incidence of sepsis in the emergency department was 1.5%, and 36.5% of those patients developed septic shock. [6,7] Several studies have shown that sepsis-related mortality has decreased every year due to the launch of the Surviving Sepsis Campaign (SSC), but there is still a high mortality rate of up to 80% in low-middleincome countries.
[8] Thus, the diagnosis and treatment of sepsis need more attention.
According to numerous studies, the perception of sepsis has gradually changed from an in ammatory response syndrome to organ dysfunction. Previous studies have focused on the value of 258 different biomarkers, such as interleukin-27 (IL-27), interleukin-6 (IL-6), and inter-alpha inhibitor proteins, for the diagnosis, treatment and prognostic prediction of sepsis. These biomarkers have some value, but most of them are expensive and cannot be widely used in clinical practice. In addition, the role of most biomarkers in the management of patients with sepsis has not been clari ed, and thus far, the most accurate combination of the commonly used clinical indicators has yet to be determined.
Considering the high incidence and mortality rate of sepsis, coupled with the limitations of the existing biomarkers, it is necessary to identify the early independent risk factors for sepsis and explore the speci c prognostic indicators, which can be used to evaluate the therapeutic effect. Therefore, this study identi ed the independent risk factors for sepsis and the prognostic indicators to determine biomarkers that can provide valuable, clinically relevant information and guide clinical practice.

Study design and participants
We performed a retrospective cohort analysis in which patients who met the inclusion criteria and were admitted to ICU were enrolled. Adult patients diagnosed with sepsis in the respiratory intensive care unit (RICU), emergency intensive care unit (EICU) and surgical intensive care unit (SICU) of Beijing Chao-Yang Hospital, Capital Medical University from January 2016 to April 2021 were enrolled.
The inclusion criteria were as follows: (1) subjects aged above 18 years; (2) subjects who met the Sepsis 3.0 diagnostic criteria. [5] The exclusion criteria were as follows:(1) subjects who died within 24 hours after admission; (2) subjects whose clinical data were missing or who did not undergo the relevant examinations during hospitalization; (3) subjects who were pregnant or lactating; and (4) subjects who were discharged automatically without a record of the clinical outcome.
This study was approved by the Institutional Review Board of Beijing Chaoyang Hospital (project approval number 2016-Ke-115). Informed consent from each patient was waived since we collected and analyzed all of the data retrospectively.

Data collection
Parents were divided into the survival group and the death group according to the 28-day outcome. The demographic and clinical data of the patients were entered into an electronic case report form included the following variables: basic characteristics, such as age, sex, comorbidities (e.g., hypertension, diabetes, etc.), infection site (respiratory system, urinary system, digestive system, skin and soft tissue, blood, catheter); relevant clinical indicators, such as the oxygenation index (PaO 2 /FiO 2 ), mean arterial pressure (MAP), arterial blood lactate level, white blood cell count (WBC), platelet count, neutrophil to lymphocyte ratio, D-dimer level, brinogen level, activated partial thromboplastin time (APTT), prothrombin time (PT), thrombin time (TT), international normalized ratio (INR), serum albumin level, total bilirubin level, serum creatinine level, blood sodium level, blood potassium level, procalcitonin level, C-reactive protein (CRP) level, Glasgow Coma Scale (GCS) score, sequential organ failure assessment (SOFA) score, acute physiology and chronic health evaluation (APACHE II) score; and outcomes, namely, septic shock, acute respiratory distress syndrome (ARDS), acute kidney injury (AKI), acute myocardial injury (AMI), acute liver injury, and disseminated intravascular coagulation (DIC).

Assessment measures
Sepsis (Sepsis 3.0): Patients with sepsis were de ned as patients with infections or strongly suspected infections in the ICU with an increase in their SOFA score ≥ 2 points from baseline.
Septic shock: Septic shock is a clinically de ned subset of sepsis cases wherein, despite adequate uid resuscitation, patients have hypotension requiring vasopressors to maintain a MAP above 65 mmHg and have an elevated serum lactate concentration greater than 2 mmol/L. SOFA: The SOFA score separately evaluates the condition of the following organ systems: respiration, coagulation, liver, cardiovascular, central nervous, and renal. [9] APACHE II: The APACHE II score is the sum of three items: the Acute Physiology Score (APS) (0-60 points), the age score (0-6 points), and Chronic Health Evaluation (CHS) score (0-5 points). [10] Statistical analysis In this study, SPSS 26.0 software was used for the statistical analysis. Categorical variables were summarized using frequencies and percentages, and continuous data are presented as the means ± SDs for normally distributed variables and the medians (interquartile ranges [IQRs]) for nonnormally distributed variables. The Mann-Whitney U test was used for continuous variables, and the χ2 test or Fisher's exact test was used for categorical variables. Variables with a p value < 0.05 in the univariate analysis were entered into multivariate logistic regression analysis to identify the independent risk factors for sepsis. Receiver operating curve (ROC) analysis was used for those variables with P < 0.05 in multivariate regression to evaluate the t of the prediction model and the prognostic e cacy of the independent risk factors. The prognostic value was evaluated according to the area under the curve (AUC), and those variables with AUC > 0.5 were jointly evaluated. All statistical tests were two-sided.

Results
A total of 135 consecutive patients were included from January 2016 to April 2021, 12 subjects did not meet the inclusion criteria. Thus, 123 patients were eligible for the analysis. Among them, there were 80 males and 43 females, the patients were had a mean age of 61.56 ± 17.12 years.
Basic characteristic of the patients with sepsis The basic characteristics are summarized in Table 1. There were no signi cant differences in comorbidities, sex, neutrophil to lymphocyte ratio, D-D, procalcitonin level, CRP level, creatinine level, and total bilirubin level between the two groups (P > 0.05). The APACHEII score, age, MAP, oxygenation index, platelet count, WBC count, arterial blood lactate level, SOFA score, albumin level, APTT, PT, TT, and INR were signi cantly different between the two groups (P < 0.05). In terms of complications, there were signi cant differences in the incidences of ARDS, AKI, ACI, and DIC between the two groups (P < 0.05). The infection sites were signi cantly different with regard to infections originating from the respiratory and urinary systems between the two groups (P < 0.05).

Prognostic value of clinical parameters in patients with sepsis
In terms of the prognostic value of the individual indicators, the area under the ROC curve for the predication of mortality in patients with sepsis based on the MAP was 0.747 (95% CI, 0.654-0.839, P < 0.001), that based on the APACHE II score was 0.839 (95% CI, 0.760-0.917, P < 0.001), and that based on the albumin level was 0.641 (95% CI, 0.541-0.741, P = 0.011). (Table 3). The above indicators with an AUC greater than 0.6 were combined to improve the predictive value. The predictive model incorporating the MAP, APACHE II score and albumin level had the highest predictive power, with an AUC of 0.890 (95% CI, 0.832-0.948, P < 0.001). (Fig. 1) The optimal cut-off value of MAP to evaluate prognose was 73mmHg. The sensitivity, speci city, and positive likelihood ratio and negative likelihood ratio for predict the outcome of patients with sepsis at optimal cut-off points of MAP, ALB, and APACHE II are listed in Table 4

Discussion
At present, the main controversial topics regarding the early management of sepsis are the lack of a robust de nition and reliable markers that can assist in the early identi cation of sepsis and the development of a therapeutic strategy. In this retrospective analysis, we found that the arterial blood lactate level, MAP, albumin level, SOFA score, and APACHE II score were risk factors for mortality in the univariate analysis, of which the MAP, albumin level, and APACHE II score had high prognostic value. Therefore, the above clinical parameters were analyzed in detail.
Blood lactate levels indicate organ dysfunction. Septic patients have decreased tissue perfusion and cellular hypoxia, in this context, the energy supply depends on anaerobic glycolysis, and the production of lactate gradually increases. Normally, the liver can further convert this lactic acids, but impaired liver function reduces lactate clearance. [11,12] In critically ill patients, lactate can be used as an indicator of the cumulative oxygen load and as a prognostic indicator. Mikkelsen et al reported that the initial lactate level was independently associated with mortality in patients with severe sepsis (OR 1.34, 95% CI 1.30-1.40, P = 0.001). [13] Our study showed that the lactate level was a risk factor for mortality in patients with sepsis (OR 1.482, 95% CI 1.105-1.988), but it was not an independent factor. The difference between the two studies may be due to the larger sample size and fewer confounders in the former study. To date, the e cacy of lactate for the guidance of resuscitation has been reported because of its prognostic value, [14] but multicenter randomized trials have shown that uid resuscitation is not bene cial in septic shock patients with or without hyperlactatemia. [15,16] In addition, when hyperlactatemia is not caused by hypoperfusion, uid resuscitation and lactate removal may be harmful. Therefore, high lactate levels should be interpreted on a patient-by-patient basis.
The main physiological functions of serum albumin are the regulation of the plasma colloid osmotic pressure and capillary membrane permeability, as well as the binding and transport of ligands. As a reservoir and carrier of many endogenous and exogenous compounds with antioxidant and circulatory protective properties, albumin is involved in free radical scavenging. [17] Current studies have shown that protein metabolism can indicate not only liver function but also prognosis in patients with critical illness.
In a comprehensive meta-analysis of 90 cohort studies including 291,433 patients with critical illness, the e cacy of hypoalbuminemia as a predictor of the outcome was evaluated by multivariate analysis, which showed that a 10g/L decrease in the serum albumin concentration was associated with a signi cant 137% increase in mortality.
[18] Similarly, our study showed that the serum albumin level was an independent risk factor for sepsis-related mortality (adjusted OR 0.755, 95% CI 0.583-0.979), and the AUC was 0.641. All of the above ndings indicate that albumin, as a nutritional indicator, can predict the later outcome of the disease. To ameliorate hypoalbuminemia, several albumin replacement studies have been conducted, such as the SAFE, CRISTAL, and ALBIOS projects, but these interventions have not reduced mortality of septic patients. One thing can be sure is that nutritional therapy is recommended in patients with sepsis as early as possible, as it may help improve hypoalbuminemia, maintain normal vascular permeability, and reduce mortality to some extent. [19][20][21] The international guidelines de ne severe sepsis as a subset of sepsis cases wherein, despite adequate uid resuscitation, patients have hypotension requiring vasopressors to maintain an MAP above 65 mmHg. This hemodynamic indicator is a simple, convenient, and more accurate re ection of the perfusion level. Decreased tissue perfusion worsens clinical outcomes by affecting organ function. [5] In a similarly designed study involving 274 patients with sepsis and septic shock, the results were adjusted for disease severity. Patients with one or more hypotensive episodes during which the MAP was below 60 mmHg for at least 2 min had higher mortality at day 28 than those without hypotension. [22] However, the SEPSISPAM trial, a large prospective randomized controlled study that compared septic shock patients with high and low MAP, showed no difference in mortality between the two groups. [23] Our study showed that MAP was an independent risk factor for mortality in patients with sepsis (adjusted OR 0.889, 95% CI 0.822-0.961), and the area under the ROC curve was 0.747, (95% CI 0.654-0.839), indicating that it could be used as a prognostic indicator and resuscitation goal for patients with sepsis. It has been shown that sepsis is associated with altered microcirculation. [24,25] The relationship between systemic hemodynamics and microcirculatory blood ow is complex. An increase in MAP levels above 65 mmHg may not alter microvascular perfusion per se, and conditions in which MAP has been corrected while microcirculatory abnormalities still persist often occur. However, it is important to note that the timely initiation of interventions may play a major role. DeBacker et al reported that microcirculation could be improved by uid resuscitation in the early stage of sepsis. [25] Fluid resuscitation is the preferred method of MAP enhancement, and the SSC advocates for the use of crystalloids. In fact, uid overload can lead to the worsening of myocardial contractility, acute pulmonary edema, and possibly death. Hence, patients with septic shock and left-sided ventricular-arterial uncoupling may require more rigorous uid management. In addition, vasoactive agents are required when uid resuscitation is ineffective in patients with septic shock. Norepinephrine is recommended by the SSC as the most commonly used vasopressor in patients with sepsis.
[26] However, to date, no precise MAP target has been identi ed for septic shock, and more research is needed. [27] In conclusion, MAP affects the prognosis of patients, and the targeted treatment goal needs to be individualized. More attention should be given to the appropriate use of vasoactive drugs for sepsis.
Multiple organ dysfunction is associated with a high risk of mortality. The SOFA and APACHE II scores are two comparatively comprehensive scores used to assess organ function, both of which correlate well with the outcome of sepsis.
The APACHE II score is widely used for the evaluation of the systemic functioning in critically ill patients to predict the severity of the disease and to facilitate risk strati cation. In 1985, the University of Washington Medical Center proposed the APACHE II scoring system in a study of 5030 cases in 13 ICUs across the United States performed by Knaus et al. [10] This score provides an initial picture of the patient's general condition in the ICU and helps predict the clinical outcome. One study showed that APACHE II score could be used to identify disease severity, and there were signi cant differences in mortality among patient groups strati ed by the APACHE II score. Regression analysis showed that an APACHE II score ≥ 15 and soluble urokinase-type plasminogen activator receptor (suPAR) ≥ 10.82 ng/mL were independently associated with adverse outcomes.
[28] Our study also showed that this score was an independent risk factor for mortality in patients with sepsis, (adjusted OR 1.356, 95% CI 1.119-1.643), and it possessed some prognostic value, with an AUC of 0.839. Thus, the APACHE II score should be calculated daily for patients with sepsis in the ICU to accurately monitor the overall condition of these patients in a timely manner. This study also indicated that the predictive value of the combination of the APACHE II score, MAP, and albumin level was signi cantly better than those of the individual indicators, with an AUC of 0.890. We recommend the use of these indicators in clinical practice.
The SOFA score can be used as one of the key indicator for the diagnosis of sepsis and is also included by the European Medicines Agency (EMA) as a treatment e cacy indicator and clinical endpoint in phase II clinical trials on new treatment options. [5] It can re ect the severity of the disease to a certain extent and has some suggestive value with regard to prognosis. In the past, the improvement in 28-day mortality has been used as a primary endpoint in sepsis trials, which may result in the underestimation of some of the potential bene cial effects of the initial intervention. [29] Pocock et al [30] proposed that the SOFA score could be selected as a surrogate endpoint when the primary outcome of the study failed. The usefulness of serial evaluation of the SOFA score for the prediction of the outcome in critically ill patients has been previously reported many times. However, those studies only provided a xed SOFA score, which is not able to comprehensively re ect the condition of patients with sepsis. Then, DeGrooth et al put forward the idea of using the change in the SOFA score (i.e., the score at a xed date after randomization minus the baseline score, or the maximum score during the ICU stay minus the baseline score) as an endpoint in studies evaluating the treatment effect in critically ill patients. Based on the 87 randomized controlled studies included in the analysis, the change in the SOFA score appeared to be reliably and consistently associated with mortality. [31] Additionally, a series of observational studies have shown a signi cant effect on mortality of a small changes in the SOFA scores. [32] Studies have shown that changes in the SOFA scores within 48 hours in patients with sepsis are associated with DIC, sudden cardiac arrest, and other serious illnesses in the ICU and emergency room. [33,34] In our study, univariate regression results showed that the SOFA score was a risk factor for sepsis-related mortality, although the relationship was nonsigni cant after accounting for confounders, probably because we used a xed SOFA score. Therefore, the change in the SOFA score is suggested is suggested as a useful predictor, in agreement with the above studies.
There are some limitations in this study. First, it was a retrospective study including 123 patients.
Prospective randomized trials with larger sample sizes are needed to more accurately analyze the factors predictive of mortality in patients with sepsis. Second, the in uential factors included in this study are not exhaustive. Finally, some patients were in critical condition, died within 24 hours or were discharged automatically, leading to insu cient clinical data and possibly affecting the nal results.

Conclusion
Sepsis is associated with a high mortality rate and seriously affects the long-term patient quality of life, which is an important consideration. In this study, we found that the combination of the MAP, albumin level, and APACHE II score was the best predictor of the risk of mortality, and we proposed regression formula calculation to more precisely re ect the risk of death. Chronic kidney disease; suPAR: Soluble urokinase-type plasminogen activator receptor; COPD: Chronic obstructive pulmonary disease; NLR: Neutrophil-to-lymphocyte ratio; AMI: Acute myocardial injury.

Declarations
Ethics approval and consent to participate This study was reviewed and approved by the Ethics Committee of Beijing Chao-Yang Hospital (2016-Ke-115), and all methods were performed in accordance with the relevant guidelines. Waiver of consent was granted by the Institutional Review Board since this was a retrospective study.

Consent for publication
Not applicable.

Availability of data and materials
All the data and materials are available from the corresponding author by request.

Competing interests
The authors declare that they have no competing interests.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.