Early prognostication of septic shock in Korean adults aged 80 years and over: serum albumin combined with Sequential Organ Failure Assessment score
Article information
Abstract
Background
Septic shock requires careful risk stratification in patients aged 80 and over. Serum albumin levels reflect host status and, when combined with the severity of organ dysfunction, may enhance early prognostication. Here, we evaluated whether combining albumin levels with the Sequential Organ Failure Assessment (SOFA) score upon emergency department (ED) arrival can improve risk classification.
Methods
We conducted an observational study from October 2015 to May 2022 using a prospective multicenter registry of septic shock from 12 EDs in the Korean Shock Society. We included 1,300 (18.8%) patients with septic shock aged ≥80 years. The SOFA score was calculated in the ED at the time of septic shock recognition. Patients were divided into three groups according to albumin levels (cutoff, 3.5 g/dl) and the initial SOFA score (cutoff, 7): A, normal albumin with low SOFA score; B, high SOFA with normal albumin or low SOFA with low albumin; and C, low albumin with high SOFA score. The primary outcome was 28-day mortality.
Results
Groups A, B, and C included 175, 592, and 533 patients, with 28-day mortality rates of 12.0%, 26.9%, and 49.3%, respectively. Adjusted odds ratios (95% CI) for mortality were 3.02 (1.76–5.18) for group B and 5.80 (3.38–9.97) for group C vs. group A.
Conclusions
Combining the SOFA score with the albumin level allows early classification of patients aged ≥80 years with septic shock in the ED, facilitating treatment discussions, particularly decision-making with patients and their families.
INTRODUCTION
Despite recent advancements in diagnosis and treatment, sepsis remains a considerable global public health concern [1]. Advanced age, especially beyond 80 years, is strongly linked to a heightened risk of mortality in patients with septic shock [2,3]. The admission of older patients to intensive care units (ICUs) has been questioned in limited settings owing to concerns regarding unnecessary invasive care and avoidable health care costs [4-6]. Although older patients are known to have a higher mortality rate than younger patients, identifying those with a favorable prognosis is essential for appropriate intervention and ICU admission.
The Sequential Organ Failure Assessment (SOFA) score offers a straightforward approach to evaluating and tracking organ dysfunction in critically ill patients and plays a central role in the current definition of sepsis [7,8]. Although the SOFA score can be used as a prognostication tool for patients, there are growing concerns about using it alone as a triage criterion [9,10]. During the coronavirus disease pandemic, in addition to disease severity, previous medical conditions were also considered in the triage of ICU admissions [11]. Accordingly, it is reasonable to predict the prognosis of intensive care for elderly patients based on both disease severity and baseline host conditions.
Albumin, a plasma protein, is essential for preserving oncotic pressure and facilitating the transport of numerous substances in the bloodstream. In addition, albumin possesses anti-inflammatory and antioxidant properties. Reduced serum albumin levels have been linked to increased mortality in clinical settings [12]. Moreover, serum albumin levels may serve as indicators of baseline host status, given their susceptibility to changes in nutritional state and chronic comorbidities [13-16].
We hypothesized that combining albumin levels, reflecting host status, and the SOFA score, reflecting the severity of organ dysfunction, could provide early risk stratification in very elderly patients aged 80 years and over with septic shock. Accordingly, we evaluated differences in mortality rates between groups defined by serum albumin levels and initial SOFA score.
MATERIALS AND METHODS
Ethics Statement
This study was approved by the Institutional Review Boards (IRBs) of each participating institute: (1) Asan Medical Center (No. 2015-1253), (2) Gangnam Sacred Heart Hospital (No. 2015-11-142), (3) Gangnam Severance Hospital (No. 3-2015-0227), (4) Hanyang University Hospital (No. HYUH 2015-11-013-022), (5) Korea University Anam Hospital (No. HRPC2016-184), (6) Korea University Kuro Hospital (No. KUGH15358-001), (7) Samsung Medical Center (No. SMC2015-09-057-057), (8) Seoul National University Hospital (No. J-1408-003-599), (9) Seoul National University Bundang Hospital (No. B-1409/266-401), (10) Severance Hospital (No. 4-2015-0929), and (11) Seoul National University Boramae Hospital (No. IRB-16-2014-36). Informed consent was obtained before data collection. The original IRB approvals and consents covered not only the prospective collection of data but also its future use for research purposes, including retrospective analyses such as this study. Therefore, a separate IRB approval for this secondary analysis was not required. The study was performed in accordance with the ethical standards outlined in the 1964 Declaration of Helsinki and its later amendments.
Study Design and Patients
This study was a retrospective analysis of data prospectively collected in the Korean Shock Society (KoSS) septic shock registry, a multicenter database maintained from October 2015 to May 2022 [17]. The KoSS is a collaborative research network dedicated to enhancing the diagnosis and management of sepsis. Since October 2015, KoSS investigators have prospectively gathered data on patients with septic shock from the emergency departments (EDs) of 12 university-affiliated hospitals across South Korea.
Inclusion criteria were as follows: (1) aged 19 years or older; (2) presented to the ED with a suspected or confirmed infection; (3) showed evidence of either refractory hypotension or hypoperfusion [18,19]. (a) Hypotension was defined as a systolic blood pressure (SBP) < 90 mm Hg, a mean arterial pressure < 65 mm Hg, or a drop in SBP of < 40 mm Hg from the patient’s baseline. [20]. “Refractory hypotension” was defined as persistent hypotension that did not resolve after an initial fluid resuscitation (20–30 ml/kg or at least 1 L of crystalloid over 30 minutes) and required the administration of vasopressors to maintain adequate perfusion. (b) Hypoperfusion was defined as serum lactate levels of ≥ 4 mmol/L.
The exclusion criteria were as follows: (1) presence of a documented do-not-resuscitate (DNR) order at the time of ED arrival; (2) failed to meet the inclusion criteria within 6 hours of ED presentation; (3) transferred from another hospital without fulfilling the inclusion criteria upon ED arrival; (4) direct transfer from the ED to another healthcare facility. In this study, patients aged 80 years or older were classified as the oldest age group, or very elderly.
Data Collection
The KoSS septic shock registry applies standardized definitions to over 200 variables, encompassing clinical characteristics, treatment modalities, and patient outcomes. Trained hospital coordinators entered and anonymized data using a standardized case report form within a web-based electronic system, which automatically flagged outliers and erroneous values. To ensure data integrity, a quality control committee was established, consisting of emergency physicians, regional research coordinators, and investigators from each participating ED. This committee conducted routine reviews of data quality and provided feedback to site coordinators and investigators. Any discrepancies or inquiries were addressed through the system’s query function or direct phone communication. Demographic and clinical variables were extracted from the registry, including age, sex, past medical history, suspected infection focus, initial vital signs, admission laboratory values, severity scores, and ED interventions. The initial SOFA score was determined at the time of septic shock diagnosis in the ED. The primary outcome of interest in this study was 28-day all-cause mortality. Missing data were handled using a complete case analysis. For each analysis, cases with missing values in the relevant variables were excluded. Patients with unrelated missing data were retained in other analyses to maximize the sample size. Given that the proportion of missing data was low, its impact on the results was deemed non-significant.
Statistical Analysis
Continuous variables were presented as either mean±standard deviation or median with interquartile range, depending on their distribution. Categorical variables were summarized as frequencies and percentages. For comparisons among groups, continuous variables were analyzed using one-way analysis of variance for normally distributed variables and the Kruskal-Wallis test for non-normally distributed variables. Categorical variables were compared using the chi-square or Fisher’s exact tests. The initial SOFA score and serum albumin level were used to categorize the patients into three groups as follows: group A (normal albumin with a low SOFA score); group B (high SOFA with normal albumin or low SOFA with low albumin); and group C (low albumin with a high SOFA score). Serum albumin levels were dichotomized using a cutoff of 3.5 g/dl, which represents the standard lower limit of the normal range used in most clinical laboratories. A sensitivity analysis using the Youden index suggested an optimized cutoff of 3.0 g/dl. However, the difference in predictive performance (area under the curve [AUC], 0.669 vs. 0.660; P=0.3037) was non-significant. A cutoff value of 3.5 g/dl was used to prioritize clinical interpretability and consistency with routine laboratory reference ranges, given the absence of a significant difference in discriminative performance compared with an outcome-optimized threshold. This cutoff is commonly used to define hypoalbuminemia in studies involving older adults, supporting its relevance in geriatric populations [21-23]. A threshold of 7 was selected for the initial SOFA score, as determined by receiver operating characteristic (ROC) curve analysis using the Youden index to optimize sensitivity and specificity. Multivariate logistic regression analysis was performed to identify independent predictors of the primary outcome, 28-day mortality, incorporating variables that were found to be significant in the univariate analysis. Variables with a P<0.1 in the univariate analysis were entered into the multivariable logistic regression model, including demographic factors, relevant comorbidities, infection sites, severity scores (SOFA and APACHE II), initial vital signs, key laboratory parameters, and initial lactate levels. The results are presented as odds ratios (ORs) with 95% CIs. A two-sided P-value ≤0.05 was considered statistically significant. All statistical analyses were conducted using SPSS for Windows version 21.0 (IBM Corp.).
RESULTS
Baseline Characteristics and Laboratory Findings
Of the 6,901 patients with septic shock enrolled in the KoSS septic shock registry, a total of 1,300 patients aged ≥80 years were included in the present analysis, covering the period from October 2015 to May 2022 (Figure 1). Patients were categorized into three groups based on serum albumin levels and initial SOFA scores: group A (n=175; normal albumin with a low SOFA score), group B (n=592; normal albumin with a high SOFA score or low albumin with a low SOFA score), and group C (n=533; low albumin with a high SOFA score).
Patient flow diagram. SOFA: Sequential Organ Failure Assessment.
Table 1 presents the baseline characteristics of the study population. The 28-day mortality rate of the overall population was 34.1%. The mean age of the overall population was 84.3±3.7 years; the non-survivor group was older than the survivor group. The non-survivor group comprised a greater proportion of males. Among the medical histories, chronic pulmonary disease and malignancy were more frequent in the non-survivor group than in the survivor group. Non-survivors had a higher prevalence of pulmonary infections than survivors. In contrast, genitourinary, gastrointestinal, and hepatobiliary infections were more frequent in the survivor group than in the non-survivor group. Among the total study population, 768 patients (59.1%) presented with an SBP <90 mm Hg on arrival. Refractory hypotension, defined as persistent hypotension requiring vasopressor support after adequate fluid resuscitation, was observed in 1,018 patients (78.3%), including 695 survivors (81.1%) and 323 non-survivors (72.9%). The non-survivor group had a significantly higher initial SOFA score than the survivor group.
Baseline characteristics of the study population
Table 2 presents the clinical and laboratory characteristics of the study population. The non-survivor group exhibited significantly lower levels of hemoglobin, blood urea nitrogen, and creatinine than the survivor group. The serum albumin level of the overall population was 3.0±0.6 g/dl, with the survivor group presenting a significantly higher albumin level. The levels of C-reactive protein and initial lactate were significantly higher in the non-survivor group than in the survivor group. In terms of hypoperfusion, 647 patients (49.8%) had a serum lactate level ≥4 mmol/L, with proportions differing by outcome group: 41.2% among survivors and 66.4% among non-survivors. Vasopressors were administered to 86.8% of patients, and their use was significantly more frequent among non-survivors (91.0%) than among survivors (84.7%) (P=0.001). Norepinephrine was the primary agent used, and the use of vasopressin differed significantly between outcome groups (28.2% vs. 9.2%, P<0.001). The time from ED arrival to antibiotic administration did not differ significantly between the survivor and non-survivor groups (2.4±1.3 hours vs. 2.4±1.4 hours, P=0.951). However, the non-survivor group had a significantly shorter time to vasopressor initiation than the survivor group (2.0±1.5 hours vs. 2.4±1.6 hours, P<0.001).
Clinical and laboratory findings from the study population
Mortality Rate in the Initial SOFA Score and Albumin Level-Defined Groups, and a Comparison of Group Characteristics
Figure 2 presents a comparison of mortality rates according to groups defined by albumin levels and the initial SOFA score. The 28- and 90-day mortality rates were significantly higher in group C, followed by groups B and A (28-day: 49.3% vs. 26.9% vs. 12.0%, P<0.001; 90-day: 60.2% vs. 39.9% vs. 16.0%, P<0.001, respectively).
Mortality rate by group. Group A, normal albumin with low SOFA score; Group B, high SOFA with normal albumin or low SOFA with low albumin; Group C, low albumin with high SOFA score. SOFA: Sequential Organ Failure Assessment.
A comparison of the characteristics between the groups is shown in Figure 3. A past medical history of chronic renal disease (15.4% vs. 9.5% vs. 9.1%, P=0.004) and liver cirrhosis (6.6% vs. 2.7% vs. 2.3%, P=0.002) was more frequent in group C, followed by groups B and A. The frequency of malignancy did not differ significantly among the groups. The laboratory findings revealed that hemoglobin levels were significantly higher in group A, followed by groups B and C (12.4% vs. 11.1% vs. 10.6%, P<0.001). Conversely, blood urea nitrogen (48.7 vs. 35.8 vs. 32.1, P<0.001) and creatinine (2.3 vs. 1.7 vs. 1.7, P<0.001) levels were significantly higher in group C, followed by groups B and A. A detailed comparison of the characteristics between the groups is presented in Supplementary Table 1.
Baseline and laboratory characteristics by group. (A) Past medical history. (B) Laboratory findings. Group A, normal albumin with low SOFA score; Group B, high SOFA with normal albumin or low SOFA with low albumin; Group C, low albumin with high SOFA score. BUN: blood urea nitrogen; SOFA: Sequential Organ Failure Assessment.
Risk Factors Associated with 28-Day and 90-Day Mortality
Using variables identified as significant in the univariate analysis, multivariate analysis was conducted to determine potential risk factors associated with 28- and 90-day mortality (Table 3). Compared with group A, the adjusted ORs (95% CI) for the 28- and 90-day mortality rates were 3.02 (1.76–5.18) and 3.70 (2.26–6.06) for group B and 5.80 (3.38–9.97) and 6.34 (3.80–10.58) for group C.
Crude and adjusted OR for predicting 28-day and 90-day mortality across the study groups in patients with septic shock
Comparison of ROC Curves between the Albumin with SOFA Grouping Model and The Severity Scoring Systems
To further evaluate the performance of the proposed risk stratification model, we compared the prognostic accuracy of the albumin with SOFA model with severity scoring systems using ROC curve analysis. When compared with the APACHE II score, the albumin with SOFA model demonstrated comparable discrimination for predicting 28-day mortality (AUC, 0.659 vs. 0.665) (Supplementary Figure 1). When compared with the SOFA score alone, the AUC for the SOFA score alone was 0.696, whereas that for the albumin with SOFA model was 0.660, indicating a statistically significant difference in discriminative performance (P=0.006) (Supplementary Figure 2).
DISCUSSION
In this study, approximately 20% of the patients with septic shock in the registry were aged ≥80 years, and their overall 28-day mortality rate was 34.1%. Based on the serum albumin level and initial SOFA score, we divided the patients into three groups and found that the 28-day mortality rate differed significantly among the three groups (49.3% vs. 26.9% vs. 12.0%, P<0.001). Compared with group A, the adjusted OR (95% CI) for 28-day mortality was 3.016 (1.756–5.180) for group B and 5.800 (3.375–9.967) for group C.
With the gradual growth in the aging population, advanced age has been associated with the development of sepsis and an increased risk of mortality [24-27]. According to data from clinical trials, the mortality rate of ED patients with septic shock was 18%–20% [18,28]. The mortality rate in the current study was 34.2%, which is higher than rates reported in previous studies that included all age groups. Our finding is consistent with the results of a study that reported mortality rates of 32.3%, 25.8%, and 24.8% in very elderly, elderly, and non-elderly patients, respectively [29].
Typically, lower serum albumin levels have been associated with increased mortality rates. Critically ill patients, including those with septic shock, frequently exhibit a state of malnutrition and increased inflammation [30]. These factors can contribute to decreased albumin synthesis and increased albumin breakdown, leading to lower serum albumin levels. Alternatively, this association can be explained by the capillary leak syndrome induced by septic shock, in which fluid and proteins, including albumin, leak out of blood vessels and accumulate in the tissues [31]. This can contribute to reduced serum albumin levels and worsening outcomes.
Although sepsis is characterized by infection-induced organ dysfunction, individual responses tend to vary. Multiple comorbidities and reduced organ reserve function are known to occur in the aging population [32-34]. Owing to a lack of functional reserve, a less severe injury may result in organ failure [35,36]. Accordingly, it would be reasonable to predict prognosis in the geriatric population using not only a disease severity score but also by considering baseline status. In this study, we observed significant differences in medical history and laboratory findings between groups defined by serum albumin levels and the initial SOFA score. Chronic kidney disease and liver cirrhosis can lead to hypoalbuminemia and are also associated with poor outcomes in sepsis [37,38]. Significant intergroup differences were also observed in laboratory markers such as hemoglobin, blood urea nitrogen, and creatinine, which are known to be associated with higher sepsis-related mortality [38,39].
Here, we have demonstrated that in very elderly patients aged 80 years or older with sepsis, early risk stratification is feasible using a simple two-step approach based on the serum albumin level and initial SOFA score. Although group B comprised two phenotypically distinct subgroups based on serum albumin levels and initial SOFA scores, both demonstrated intermediate mortality rates, and further subdivision did not meaningfully improve risk discrimination, supporting the use of a simplified three-tier risk stratification model. Accurate prognostication during the early phase of sepsis is particularly important in very elderly patients, in whom ICU admission decisions are often complex and challenging. Previous studies have shown that timely recognition and management of sepsis can improve outcomes in this population, underscoring the need for practical tools to support early clinical decision-making [40,41].
In a supplementary analysis, the albumin with SOFA model demonstrated discrimination comparable to that of the APACHE II score for predicting 28-day mortality, despite its substantially simpler structure. However, when compared with the SOFA score alone, the albumin with SOFA model showed lower discriminative performance. Accordingly, this albumin–SOFA-based approach should be interpreted as a complementary tool to support clinical decision-making, rather than as a standalone prognostic model. Its simplicity may facilitate rapid risk assessment in time-sensitive emergency settings and help guide discussions regarding ICU admission, escalation of care, or enrollment in future clinical studies. Supplementary Figure 3 presents a conceptual illustration of the potential clinical application of the albumin–SOFA-based risk stratification.
The strength of our study is that we have focused on a large number of very old patients (≥80 years) with septic shock compared with the populations enrolled in previous studies. Nevertheless, our study had some limitations. First, as this was a multicenter study, differences in enrollment periods and case volumes across hospitals were present, and institutional characteristics were not accounted for in the analysis. Second, the exclusion of patients with documented DNR orders may have led to the omission of individuals with more advanced or chronic underlying conditions. This constitutes selection bias, which may have reduced the differences between the groups, thereby understating the primary finding. Third, although this study primarily focused on initial prognostication, longitudinal data, such as previous nutritional status, can be used to make accurate predictions. Fourth, detailed vasopressor dosing data, including norepinephrine equivalent doses, were not collected, which may have limited the interpretation of the treatment intensity. Fifth, missing data were handled using complete case analysis. Although the proportion of missing data was low, this approach may introduce bias if missingness was not completely at random. Sixth, frailty status, identified as a notable prognostic factor in elderly patients with sepsis, was not assessed in our study owing to limitations in the available registry data. The absence of frailty measures, such as the Clinical Frailty Scale, may limit the comprehensiveness of risk stratification in this population. Finally, our study used a predefined albumin cutoff value of 3.5 g/dl based on the standard laboratory reference range, rather than an outcome-optimized threshold. Although a sensitivity analysis using the Youden index suggested an alternative cutoff (3.0 g/dl), the model performance did not differ significantly. Nonetheless, different albumin thresholds may yield different risk classifications, and future studies are warranted to validate the robustness of our findings across various cutoff values.
Serum albumin levels combined with the SOFA score may facilitate the early prognostication of very elderly patients with septic shock. Patients with lower albumin levels and higher SOFA scores may be at a higher risk of mortality. Early prognostication facilitates more informed discussions about treatment plans with patients and their families.
KEY MESSAGES
▪ Patients aged ≥80 years with septic shock have high mortality; early risk stratification is crucial for guiding appropriate treatment and intensive care unit admission decisions.
▪ A simple combination of serum albumin level (cutoff, 3.5 g/dl) and the initial Sequential Organ Failure Assessment score (cutoff, 7) stratified patients into three prognostic groups, showing markedly different 28-day mortality rates (12.0%, 26.9%, and 49.3%).
▪ This easy-to-use model can support early prognostication in the emergency department, facilitating timely decision-making and treatment discussions with patients and families.
Notes
CONFLICT OF INTEREST
No potential conflict of interest relevant to this article was reported.
FUNDING
None.
ACKNOWLEDGMENTS
None.
AUTHOR CONTRIBUTIONS
Conceptualization: SMK, SMR, WYK, KK, THL, SPC, SHC, WYK. Methodology: SMK, SMR, WYK, KK, THL, SPC, SHC, WYK. Formal analysis: SMK, SMR, WYK, KK, THL, SPC, SHC, WYK. Data curation: SMK, SMR, WYK, KK, THL, SPC, SHC, WYK. Visualization: SMK. Project administration: SMK, WYK. Writing - original draft: SMK. Writing - review & editing: SMK, SMR, WYK, KK, THL, SPC, SHC, WYK. All authors read and agreed to the published version of the manuscript.
SUPPLEMENTARY MATERIALS
Supplementary materials can be found via https://doi.org/10.4266/acc.005625.
Characteristics of the groups defined by the initial SOFA score and albumin level
acc-005625-Supplementary-Table-1.pdfReceiver operating characteristic curves comparing the prognostic performance of the Sequential Organ Failure Assessment (SOFA) and albumin model and the Acute Physiology and Chronic Health Evaluation (APACHE) II score for predicting 28-day mortality in very elderly patients with septic shock. AUC: area under the curve.
acc-005625-Supplementary-Figure-1.pdfReceiver operating characteristic curves comparing the discriminative performance of the SOFA score alone and the albumin with SOFA model. SOFA: Sequential Organ Failure Assessment; ICU: intensive care unit.
acc-005625-Supplementary-Figure-2.pdfConceptual illustration of the clinical application of SOFA + albumin model risk stratification. SOFA, sequential organ failure assessment; ICU, intensive care unit.
acc-005625-Supplementary-Figure-3.pdf