ASSOCIATIONS OF THE SERUM ALBUMIN CONCENTRATION AND SEQUENTIAL ORGAN FAILURE ASSESSMENT SCORE AT DISCHARGE WITH 1-YEAR MORTALITY IN SEPSIS SURVIVORS: A RETROSPECTIVE COHORT STUDY

ABSTRACT Introduction: This study was performed to investigate the predictors of 1-year mortality at discharge in sepsis survivors. Methods: This study was a retrospective analysis of patients with sepsis and septic shock at a single center. Patients who survived hospitalization for sepsis or septic shock between January 2016 and December 2017 were included in this study. Age, sex, body mass index, laboratory results such as blood cell count, C-reactive protein (CRP) and albumin levels, the Sequential Organ Failure Assessment (SOFA) score at the time of discharge and site of infection were compared between the survivors and nonsurvivors at 1 year postdischarge. Multivariate logistic regression was performed to identify the predictors of 1-year mortality. Results: During the study period, 725 sepsis patients were included in the analysis, 64 (8.8%) of whom died within the first year. The nonsurvivors were older and had a lower body mass index and a higher SOFA score at discharge than the survivors (P < 0.05). Among the laboratory results at discharge, hemoglobin, platelet counts, and albumin concentrations were lower in the nonsurvivors than in the survivors, whereas CRP was higher in the nonsurvivors than in the survivors. In the multivariate logistic regression analysis, serum albumin <2.5 mg/dL and SOFA score ≥2 at discharge were identified as independent prognostic factors for 1-year mortality (odds ratio, 2.616; 95% confidence interval, 1.437–4.751 for albumin <2.5 mg/dL and 2.106, 1.199–3.801 for SOFA score ≥2, respectively). Conclusions: A low serum albumin concentration of <2.5 mg/dL and a high SOFA score of ≥2 at the time of discharge were prognostic factors for 1-year mortality in survivors of sepsis.


INTRODUCTION
Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection (1). It has been reported that 19.7% of all global deaths are related to sepsis, potentially accounting for 5.3 million deaths annually (2,3). Due to the risk of sepsis and the global burden, many studies have been conducted to investigate appropriate management strategies and prognostic factors.
Many studies have been performed to identify the prognostic factors for mortality in sepsis. In previous studies, age, sex, comorbidities, biomarkers, and severity of illness were identified as prognostic factors for patients with sepsis (4)(5)(6)(7)(8)(9). However, most studies have mainly focused on short-term mortality, such as 28-or 90-day mortality.
Several studies have been conducted on the long-term mortality of sepsis survivors after hospital discharge. In a previous study, age, sex, race/ethnicity, severe comorbidities, dependency, nonsurgical status, site of infection and organ dysfunction were reported to be associated with long-term mortality (10). However, this study was performed using data acquired within the patients' first 24 hours following admission and included only patients admitted to the intensive care unit (ICU). Other studies used various biomarker trajectories, such as highly sensitive C-reactive protein, interleukin-6, interleukin-10, and soluble programmed death ligand 1; however, these markers are difficult to measure clinically (11,12).
During the hospitalization period, a patient with sepsis undergoes many changes in his or her condition, and at discharge, the condition of sepsis survivors is better than that during hospitalization. However, sepsis survivors may not recover to their presepsis health condition and continue to suffer from various complications, which may lead to death after discharge (13,14). The identification of factors related to the long-term prognosis of sepsis survivors at the time of discharge would be helpful in the management and follow-up plan of patients and would improve their prognosis. Therefore, we performed this study to investigate the factors associated with the 1-year mortality of sepsis survivors at the time of discharge.

Study design
A retrospective analysis of a prospectively registered database of all consecutive patients with sepsis or septic shock was performed at an urban tertiary academic hospital. The Institutional Review Board of Seoul National University Bundang Hospital approved the study and waived the requirement for informed consent (B-1909-567-111).

Study setting and population
From January 2016 to November 2017, adult patients (age, ≥18 years) who visited the emergency department and were diagnosed with sepsis or septic shock were initially included. Patients who died in the hospital or were requiring end-oflife care were excluded. If patients were hospitalized more than once during the study period, only the first hospitalization data were used; data from subsequent hospitalizations were excluded. Patients without available data for the week before hospital discharge were also excluded.
The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) was used to define sepsis and septic shock. Sepsis was defined as a sequential organ failure assessment (SOFA) score ≥2 points related to infection. Septic shock was defined as persistent hypotension that required vasopressors to maintain the mean arterial pressure ≥65 mm Hg and the serum lactate concentration ≥2 mmoL/L after fluid resuscitation (1). Patients with sepsis or septic shock were treated according to the international guidelines for sepsis management (1,15).

Measurement
Patient data at admission were collected using standardized data collection forms (4,16,17). The forms included demographic data, including age, sex, body mass index (BMI), underlying comorbidities, site of infection, laboratory results, and severity scores. In addition to the data at the time of admission, the data at the time of discharge were collected through a hospital information system. Data at the time of discharge included body weight; laboratory results such as complete blood cell counts, albumin, and C-reactive protein, and the SOFA score. Data closest to the time of discharge within 1 week before discharge were used, and patients without available data collected during that week were excluded.
The primary outcome was 1-year mortality after discharge, and medical records or structured telephone follow-ups were used to check survival. If both methods were impossible, we checked the expunged record of the National Health Insurance Service (NHIS). In the Republic of Korea, all citizens are obliged to subscribe to the NHIS. The NHIS record is expunged when a person dies, has been missing for more than 5 years, or loses nationality. We classified expunged patients as dead and used the expunged date as the date of death.

Statistical analysis
Continuous variables are expressed as medians with interquartile ranges, and categorical variables are expressed as numbers (%). Continuous variables were compared using Student t test or the Mann-Whitney U test, as appropriate, depending on whether they were normally distributed. Categorical variables were compared using the χ 2 test or Fisher exact test as appropriate. A linear regression analysis was performed to identify variables correlated with albumin concentration at discharge. Multivariate logistic regression was used to analyze the relationship between each variable at discharge and 1-year survival, and the results were expressed as adjusted odds ratios (ORs) and 95% confidence intervals (CIs). Age, sex, BMI, SOFA score, site of infection and variables considered to be associated with long-term prognosis, such as white blood cell count, hemoglobin, albumin, and C-reactive protein, were used in the multivariate analysis (9)(10)(11)18,19). Based on the results of the multivariate analysis with continuous variables, albumin and the SOFA score at discharge were categorized and analyzed. Survival analysis for 1 year after hospital discharge was performed using the Kaplan-Meier survival curve and multivariate Cox proportional hazards regression analysis. For the Kaplan-Meier curve, the survival probabilities of the two groups were analyzed using the log-rank test. In the multivariate Cox proportional hazards regression analysis, the results were expressed as adjusted hazard ratios (HRs) and 95% CIs. Receiver operating characteristic (ROC) analysis with albumin and SOFA scores at discharge was performed. The positive and negative predictive values of serum albumin <2.5 mg/dL and SOFA score ≥2 at discharge were calculated. All statistical analyses were performed using R software version 4.2.2 for Windows (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism 9.3.1 (GraphPad Software, La Jolla, CA). A two-tailed P value less than 0.05 was considered statistically significant.

Characteristics of the study population
During the study period, 1,125 eligible patients with sepsis were included. Among these patients, 203 (18%) died in the hospital or were requiring end-of-life care, and 922 survived to hospital discharge. Patients who were hospitalized more than once (n = 63) and those without available data from the week before discharge (n = 134) were excluded. Finally, 725 patients who survived until hospital discharge were included in the final analysis. Among them, 64 (8.8%) died within 1 year (Fig. 1).
The median age of the patients was 75 (65-81) years, and there were 396 male patients (54.6%). The baseline characteristics of the survivors and nonsurvivors at admission are presented in Table 1. The nonsurvivors were older and had a lower BMI than the survivors. Comorbidities did not differ between the survivors and nonsurvivors. The patients with respiratory tract infections had low survival rates. Laboratory results showed that the nonsurvivors had lower hemoglobin and albumin concentrations than the survivors. Microorganisms were identified from blood cultures of 189 (26.1%) patients. The most frequently isolated microorganisms in the blood cultures were Escherichia coli, Klebsiella pneumoniae, and Staphylococcus aureus. The nonsurvivors had longer hospital lengths of stay and higher SOFA scores than the survivors ( Table 1).
The patient characteristics at discharge are presented in Table 2. BMI, white blood cell count, hemoglobin, albumin, creatinine, CRP concentrations, and SOFA scores were decreased, and platelet counts were increased at the time of discharge compared with the initial values at admission (Supplemental Table 1, http:// links.lww.com/SHK/B623). At the time of discharge, BMI, hemoglobin, platelets, and albumin levels were lower, and CRP and SOFA scores were higher in the nonsurvivors than in the survivors ( Table 2).
The positive and negative predictive values of serum albumin <2.5 mg/dL at discharge were 0.21 and 0.94, respectively. Those with SOFA scores ≥2 at discharge were 0.13 and 0.95, respectively.
Of a total of 725 sepsis survivors, 245 patients (33.8%) were admitted to the ICU. Among them, 191 (78.0%) patients had septic shock. In the patients with septic shock with an albumin level <2.5 mg/dL, the 1-year mortality rate was 20.4%, and it was not significantly different from that of the patients without septic shock with an albumin level <2.5 mg/dL (18.2%).

Multivariate analysis
In the multivariate logistic regression analysis with all variables considered to be continuous variables, low albumin at discharge (OR, 0.360; 95% CI, 0.181-0.708) and a high SOFA score at discharge (OR, 1.209; 95% CI, 1.042-1.400) were identified as prognostic factors for 1-year mortality (Supplemental Table 2, http:// links.lww.com/SHK/B623). Based on these results, the patients White blood cells, 10 9 /L 11.6 (7.9-16.9) 11.7 (7.9-16.9) 11.3 (7.8-18.9) 0.793 Hemoglobin, g/dL 11.6 (10.0-13.0) 11.6 (10.  Values are expressed as medians (interquartile range) or numbers (%) as appropriate. The P value is from Student t test or the Mann-Whitney U test for continuous variables and the χ 2 test or Fisher exact test for categorical variables, as appropriate. ICU, intensive care unit; SOFA, Sequential Organ Failure Assessment.  were divided into groups based on the albumin concentration at discharge: albumin ≥2.5 mg/dL and albumin <2.5 mg/dL. In the sepsis survivors, the median albumin concentration was 2.9 (2.5-3.2) mg/dL, and a cutoff of 2.5 mg/dL albumin was selected based on clinically significant hypoalbuminemia (20). The patients were divided into two groups based on the SOFA score at discharge: SOFA score ≥2 and <2 (1). In the linear regression analysis, albumin concentration at discharge was related to age, respiratory infection, and abdominal infection (Supplemental Table 3, http://links.lww.com/SHK/B623), and these variables were also included in the multivariate logistic regression analysis. In the multivariate logistic regression model that included albumin and the SOFA score as categorical variables, albumin <2.5 mg/dL and SOFA score ≥2 were identified as prognostic factors for 1-year mortality (OR, 2.616; 95% CI, 1.437-4.751 for albumin <2.5 mg/dL and OR, 2.106; 95% CI, 1.199-3.801 for SOFA score ≥2, respectively) ( Table 3). During the 1-year follow-up period after hospital discharge, albumin <2.5 mg/dL at discharge was associated with an approximately 2.5-fold higher risk of death in a year (HR, 2.486; 95% CI, 1.439-4.295). A SOFA score ≥2 at discharge was associated with an approximately 2.0-fold higher risk of death in a year (HR 2.024, 95% CI 1.181-3.468) (Fig. 2). The Kaplan-Meier curve showed that albumin <2.5 mg/dL at discharge and a SOFA score ≥2 points at discharge increased mortality over the first year ( P < 0.05; Fig. 3, A and B).

DISCUSSION
In this study, in the sepsis survivors, inflammatory markers, such as white blood cell counts and C-reactive protein concentrations, were decreased at discharge compared to the values at admission. In addition, the albumin concentration and SOFA scores were also decreased. In the multivariate analysis, the serum albumin concentration and SOFA score at discharge were predictors of 1-year mortality in patients who survived to hospital discharge with sepsis or septic shock.
Many studies have investigated prognostic factors for sepsis, but most studies investigated prognostic factors for 1 or 3 months of follow-up (5,9,21,22). There have been some studies on long-term mortality in sepsis. In a study of sepsis survivors who were admitted to 192 critical care units, variables obtained within 24 hours of the index hospitalization were analyzed (10). In this study,  In the log-rank test, the two groups showed a significant difference ( P < 0.001, both A and B).
age, male sex, one or more severe comorbidities, prehospitalization dependency, and nonsurgical status were identified as risk factors for long-term mortality. In a study of immune responses, many biomarkers were measured at 5 time points from the index admission (11). In this study, persistent elevations of inflammatory and immunosuppressive biomarkers were associated with worse long-term outcomes (11). In patients with community-acquired pneumonia, it was reported that elevated concentrations of interleukin-6 and interleukin-10 were associated with 1-year mortality (12). In another study of sepsis survivors, laboratory data obtained within 24 hours of ICU admission were analyzed, and a lower serum albumin concentration was a predictor of short-term and long-term mortality in patients with septic shock (23).
The present study has several differences compared to previous studies. First, we focused on long-term rather than short-term survival. Sepsis survivors suffer from various organ injuries during hospitalization, which can influence the long-term prognosis after discharge. Therefore, the present study was performed because we thought that identifying the factors related to long-term outcomes at the time of discharge would be helpful for the treatment and follow-up plans of sepsis survivors after discharge. Second, although there have been a few studies on long-term mortality in sepsis survivors, most of them used variables at the time of admission rather than at discharge. However, the initial variables reflect the patient's condition at the time of admission; therefore, they are likely to be more relevant to short-term outcomes than long-term outcomes. Therefore, we used variables measured after the completion of acute treatment at the time of discharge. Third, this study included patients admitted to the ICU and general wards. The criteria for ICU admission may differ among hospitals. As this study included sepsis patients admitted to the general wards as well as the ICUs, we could include all sepsis patients compared to studies of only ICU patients. Finally, the present study did not include various biomarkers that can be measured only in the laboratory but included biomarkers that could be easily measured in clinical practice.
In the present study, the serum albumin concentration and SOFA score at discharge were identified as prognostic factors for long-term survival in patients with sepsis. In the present study, most laboratory results were improved at discharge compared to those at admission and were almost within the normal range (Supplemental Table 1, http://links.lww.com/SHK/B623). However, the albumin concentration at discharge did not improve; rather, it decreased compared with that at admission. The decreased albumin concentration might be due to the dilution effect caused by fluid therapy and albumin leakage due to increased vascular permeability that occurs during sepsis (24,25). Albumin concentration has been reported as a prognostic factor for short-and long-term mortality in patients with sepsis (23,26,27). Patients with malnutrition have a high risk of death in the ICU, and serum albumin concentration is a nutritional marker (28). In addition, albumin could act as an antioxidant and a transporter for hormones or signaling molecules and inhibit platelet aggregation (20). Therefore, hypoalbuminemia is associated with long-term survival.
The SOFA score is used to measure organ dysfunction in critically ill patients, and a high SOFA score is associated with severity and poor outcome. Sepsis-3 also uses a change in SOFA score of 2 or more as an important diagnostic criterion (1). In a previous study on the long-term survival of sepsis, the SOFA score at admission was identified as a prognostic factor, and the SOFA score at discharge was identified as a prognostic factor in the present study (10). In the present study, the median SOFA score decreased from 6 (4-8) at admission to 1 (0-3) at discharge. However, 330 patients (45.5%) had a SOFA score of 2 or higher at discharge. This suggests that organ dysfunction remains chronic in many patients with sepsis, and a SOFA score of 2 or higher at discharge is associated with long-term mortality.
The results of this study could help to establish a plan for discharge and long-term follow-up of patients with sepsis. High-risk patients, such as those with hypoalbuminemia or high SOFA scores, may be considered for delayed discharge or more frequent follow-up after discharge.
This study had several limitations. First, this was a retrospective analysis conducted at a single center and the analysis period was relatively short. Therefore, the results may not be generalizable to other institutions. Second, long-term mortality from sepsis is known to be associated with inflammation and immune function (11). However, our study did not measure biomarkers of the immune response. Instead, this study used commonly available variables. Finally, in some patients, the date of death was estimated based on the date that the National Health Insurance Service record was expunged. It is possible that health insurance was canceled not because of death but for other reasons, such as patient disappearance or loss of nationality, although this would have a negligible effect.

CONCLUSIONS
A low serum albumin concentration of <2.5 mg/dL and a high SOFA score of ≥2 at the time of discharge were prognostic factors for 1-year mortality in survivors of sepsis.

FIG. 4.
Receiver operating characteristic curves of albumin concentration and SOFA score at discharge. The area under the curve (95% confidence interval) of albumin was 0.720 (0.653-0.787) and that of the SOFA score was 0.659 (0.589-0.730), with P < 0.0001. There was no significant difference in the AUC of the two curves in the bootstrap test ( P = 0.147). AUC, area under the curve; SOFA, Sequential Organ Failure Assessment.