Abstract
Objectives
The original Sequential Organ Failure Assessment (SOFA-1) score is widely used to quantify organ dysfunction in critically ill patients, including those with septic shock. A recently updated, data-driven SOFA score (SOFA-2) revises component thresholds and incorporates contemporary organ support modalities. While SOFA-2 has been validated in intensive care unit populations, its performance in the emergency department (ED), particularly among patients with septic shock, is uncertain. We therefore sought to validate SOFA-2 in ED patients with septic shock.
Methods
In this observational study, we analyzed two prospective septic shock registries. Adult ED patients with septic shock were included. SOFA-1 and SOFA-2 scores were calculated using the worst physiologic and laboratory values recorded within the first 24 h in the ED. We compared score distributions and examined mortality across score ranges. Discrimination for in-hospital, 28-day, and 90-day mortality was evaluated using the area under the receiver operating characteristic curve (AUROC). Calibration for in-hospital mortality was assessed using the calibration curve and Brier scores. Reclassification performance was evaluated using continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI).
Results
The study included 2669 patients in Cohort A and 1443 patients in Cohort B. SOFA-1 and SOFA-2 score distributions differed significantly. Mortality generally increased with higher SOFA-1 and SOFA-2 scores across all outcomes. SOFA-2 demonstrated significantly higher AUROCs than SOFA-1 for in-hospital, 28-day, and 90-day mortality in both cohorts (Cohort A: 0.734 vs. 0.702, 0.719 vs. 0.694, and 0.680 vs. 0.658; Cohort B: 0.719 vs. 0.671, 0.737 vs. 0.689, and 0.711 vs. 0.666; all p < 0.001). SOFA-2 showed generally closer agreement between predicted and observed in-hospital mortality and lower Brier scores. Continuous NRI (0.276–0.635) and IDI (0.024–0.062) also favored SOFA-2 across all mortality outcomes (all p < 0.001). In adjusted analyses, each 1-point increase in SOFA-2 was independently associated with higher mortality. Sensitivity analyses and pooled analyses showed consistent findings.
Conclusion
Among ED patients with septic shock, SOFA-2 showed modestly improved discrimination, closer agreement between predicted and observed in-hospital mortality with lower overall prediction error, and more favorable continuous reclassification metrics compared with SOFA-1. These findings support the external validation of SOFA-2 in this population and warrant further evaluation in diverse ED settings.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s40560-026-00921-3.
Keywords: Sepsis, Septic shock, SOFA-1, SOFA-2, Organ dysfunction, Mortality, Emergency department
Introduction
Sepsis is a life-threatening condition caused by a dysregulated host response to infection. Septic shock is the most severe form of sepsis, with high mortality [1–3]. Despite advances in sepsis management, mortality remains high, leading to a major global burden [1–5]. A substantial number of patients with sepsis and septic shock present to the emergency department (ED) [1–3,5–9]. The ED plays a critical role in early identification, risk stratification, resuscitation, and management of patients with sepsis and septic shock.
The Sequential Organ Failure Assessment (SOFA) score was developed approximately 30 years ago to assess the severity of organ dysfunction in critically ill patients [10–12]. It was not originally intended to predict mortality. The original SOFA (SOFA-1) score evaluates six organ systems (brain, respiratory, cardiovascular, hepatic, renal, and hemostasis). Each system is scored from 0 to 4, yielding a total score ranging from 0 to 24. Higher SOFA scores indicate more severe organ dysfunction. The SOFA-1 score has been widely used in epidemiological studies, assessment of disease severity, evaluation of treatment response, mortality prediction, and as a key component of the Sepsis-3 definition [1,10–15]. However, the SOFA-1 score has not changed since its original development, and therefore it does not reflect advances in modern clinical practices [1,10–12,16–21].
Recently, an updated data-driven SOFA (SOFA-2) score has been published [16,17]. The thresholds of each component (brain, respiratory, cardiovascular, liver, kidney, and hemostasis) were refined using large datasets. In addition, modern advances in clinical practices, including vasopressin use, advanced ventilatory supports, and renal replacement therapies, were incorporated into the scoring system. The SOFA-2 score showed improved prognostic performance for illness severity in large multinational intensive care unit (ICU) cohorts. However, it has been validated only in ICU populations using ICU mortality and has not yet been evaluated in ED patients, in specific subgroups such as septic shock, or against other short-term mortality outcomes.
Therefore, this study aims to externally validate the updated SOFA-2 score in ED patients with septic shock using two complementary prospective registries to evaluate performance across different ED septic shock settings. We also compared the prognostic performance of the SOFA-1 and SOFA-2 scores for survival outcomes. We hypothesized that the ED-based SOFA-2 score would provide better discrimination for mortality than the SOFA-1 score in this population.
Methods
Study design and setting
This observational study used data from two prospective septic shock registries to perform two external validations of the SOFA-2 score in ED patients with septic shock. The two registries were selected because they represent complementary ED septic shock populations with distinct enrollment strategies, clinical settings, and case mix.
First, the Korean Shock Society (KoSS) registry, an ED-based prospective multicenter septic shock registry, was used as Cohort A [14,22]. Adult patients with confirmed or suspected infections, and patients with refractory hypotension or hyperlactatemia in the ED were included in the KoSS registry. This cohort represents a multicenter ED-based population with early hypoperfusion of septic shock at ED presentation. Four participating institutions from the KoSS registry were used in this study (Samsung Medical Center, n = 1,743; Asan Medical Center, n = 820; Korea University Ansan Hospital, n = 76; CHA Bundang Medical Center, n = 46).
Second, a prospective single center registry from Korea University Ansan Hospital, a tertiary academic teaching hospital, was used as Cohort B [23]. This registry included patients with septic shock identified across the hospital, including the ED, ICU, and general wards, based on suspected infection, vasopressor requirement, and elevated lactate levels. For the present study, only patients with septic shock identified in the ED were included. This cohort was used to assess SOFA-2 performance in a broader real-world ED septic shock population, including patients without initial hypoperfusion and patients with delayed deterioration.
These two registries were intentionally selected to provide complementary perspectives on ED septic shock. The KoSS registry represents early ED septic shock identification across multiple EDs, whereas the single-center registry captures a broader and more heterogeneous real-world clinical spectrum. Evaluating the SOFA-2 score across both cohorts allowed us to assess its robustness and generalizability across different patient characteristics and clinical settings.
The study was conducted in accordance with the principles of the Declaration of Helsinki. This study was approved by the Institutional Review Board of Korea University Ansan Hospital (2026AS0001). The requirement for informed consent was waived due to the nature of the observational study.
Selection of participants
Adult patients enrolled in either registry who met the Sepsis-3 criteria for septic shock in the ED were included. For Cohort B, patients who developed or were first identified as having septic shock outside the ED were excluded. Patients with unknown outcomes were excluded.
Measurements
Sepsis was defined as an acute increase in the SOFA-1 score ≥2 from baseline due to infection [1]. Septic shock was defined as the requirement of a vasopressor to maintain a mean arterial pressure ≥65 mmHg and a serum lactate level >2 mmol/L, despite adequate fluid resuscitation. All patients were managed according to the Surviving Sepsis Campaign guidelines [2,3].
Worst values in the ED were used for SOFA score calculations. If the patients stayed more than 24 h in the ED, only values within the first 24 h in the ED were used. Accordingly, the present study evaluated the prognostic performance of SOFA-1 and SOFA-2 based on the initial ED assessment period.
Because the Alert/Verbal/Painful/Unresponsive (AVPU) scale is widely used in the ED, resulting in frequently missing Glasgow Coma Scale (GCS) across the cohorts, and because minor adjustments were made to the GCS in the brain component of the SOFA-2 score [16,17], cases with missing GCS were imputed using the AVPU scale in the ED (Supplementary Table 1) [24].
SOFA-1 and SOFA-2 scores were calculated according to previously published definitions [1,10–12,16,17]. Briefly, the SOFA-2 score is an updated SOFA score that evaluates six organ systems (brain, respiratory, cardiovascular, liver, kidney, and hemostasis), each ranging from 0 to 4 points, yielding a total score ranging from 0 to 24 [16,17]. Higher scores indicate more severe organ dysfunction. Compared with the SOFA-1 score, SOFA-2 incorporates contemporary organ support modalities and revised thresholds reflecting current clinical practice [16,17,21,25].
The following data were extracted from the registries: age, sex, infection focus, comorbidities, laboratory and clinical variables required for SOFA score calculation, SOFA scores, lactate levels, and survival outcomes.
Outcome measures
The primary outcome was in-hospital mortality. The secondary outcomes were 28-day mortality and 90-day mortality.
Primary data analysis
Continuous variables were expressed as means±standard deviations or medians [interquartile ranges], as appropriate. Categorical variables were expressed as numbers (percentages). Distributions of total SOFA scores and SOFA scores by organ system were evaluated using stacked bar graphs and heatmaps. Mortality distributions according to SOFA scores were evaluated using bar graphs and heatmaps. Discordance between SOFA-1 and SOFA-2 scores was evaluated using the Bowker test of symmetry.
Receiver operating characteristic (ROC) curve analyses were performed. Area under the ROC curves (AUROC) was used to evaluate the performance of the SOFA-1 and SOFA-2 scores for outcomes. AUROCs were compared using the DeLong test. Calibration for in-hospital mortality was assessed using calibration curves and Brier scores. Reclassification statistics, including the continuous Net Reclassification Index (NRI) and Integrated Discrimination Improvement (IDI), were also evaluated to assess the incremental predictive performance of SOFA-2 compared with SOFA-1. Confidence intervals were estimated using 1,000 bootstrap iterations.
To evaluate the independent association between SOFA scores and outcomes, multivariable logistic regression analyses were performed after adjustment for age, sex, infection focus, comorbidities, lactate levels, and ICU admission. ICU admission was included as an additional adjustment variable to account for differences in subsequent treatment intensity and resource utilization.
Subgroup analyses were performed according to enrollment period for each cohort (2016–2020 vs. 2021–2024) and ICU admission following ED presentation (patients who were admitted to ICU vs. those who were not).
Sensitivity analyses were performed after imputation of normal values (no organ dysfunction) of SOFA score components for cases with missing variables for calculating SOFA-1 and SOFA-2 scores (sensitivity analysis 1). An additional sensitivity analysis was performed after only including patients with GCS (excluding those with imputation with the AVPU scale; sensitivity analysis 2). Sensitivity analyses were also performed after excluding patients with metastatic cancer (sensitivity analysis 3) and after excluding patients who required vasopressors only after the first 24 h in the ED (sensitivity analysis 4).
Additionally, because Korea University Ansan Hospital contributed patients to both registries, patient-level matching using unique patient identifiers was performed and duplicate patients were excluded before pooled analyses. Subsequently, pooled analyses combining all cohorts were conducted using the same analytical framework.
A p value<0.05 was considered statistically significant. Correction for multiple comparisons was not performed. All analyses were conducted using R version 4.5.0 (R Foundation for Statistical Computing, Vienna, Austria).
Results
In Cohort A, 2,685 ED patients with septic shock were screened. Among them, 16 had missing variables required to calculate the SOFA score, and none were excluded due to unknown outcomes. Finally, 2,669 complete cases were included in the final analysis (Samsung Medical Center, n = 1,740; Asan Medical Center, n = 807; Korea University Ansan Hospital, n = 76; CHA Bundang Medical Center, n = 46; Fig. 1).
Fig. 1.
Flow chart of study population
In Cohort B, from 3110 patients with septic shock, 1525 ED patients with septic shock were screened. Of these, 82 had missing variables for the SOFA score calculation, and none were excluded due to unknown outcomes. The final analytic cohort comprised 1,443 complete cases.
Baseline characteristics of the study population are shown in Table 1. Mean SOFA-1 scores were 9.4 in Cohort A and 9.5 in Cohort B, while mean SOFA-2 scores were 8.2 and 8.7, respectively. Median ED length of stay was 11.5 and 10.5 h, and ICU admission rates were 58.6% and 62.2% in Cohorts A and B, respectively. In-hospital, 28-day, and 90-day mortality rates were 31.0%, 30.1%, and 43.3% in Cohort A and 32.7%, 37.5%, and 47.3% in Cohort B.
Table 1.
Baseline characteristics of study population
| Variables | Cohort A N=2669 |
Cohort B N=1443 |
|---|---|---|
| Age (years) | 68 [60–76] | 74 [62–82] |
| Sex | ||
| Men | 1717 (64.3%) | 830 (57.5%) |
| Women | 952 (35.7%) | 613 (42.5%) |
| Infection focus | ||
| Pneumonia | 555 (20.8%) | 657 (45.5%) |
| Abdomen | 1066 (39.9%) | 351 (24.3%) |
| Urinary tract infection | 317 (11.9%) | 328 (22.7%) |
| Others | 731 (27.4%) | 107 (7.4%) |
| Comorbidities | ||
| Hypertension | 1085 (40.7%) | 784 (54.3%) |
| Diabetes | 954 (35.7%) | 593 (41.1%) |
| Chronic heart disease | 390 (14.6%) | 179 (12.4%) |
| Chronic lung disease | 214 (8.0%) | 69 (4.8%) |
| Stroke | 262 (9.8%) | 239 (16.6%) |
| Chronic kidney disease | 259 (9.7%) | 197 (13.7%) |
| Chronic kidney disease with dialysis | 91 (3.4%) | 79 (5.5%) |
| Chronic liver disease | 433 (16.2%) | 74 (5.1%) |
| Metastatic cancer | 1081 (40.5%) | 32 (2.2%) |
| Hematologic malignancy | 353 (13.2%) | 23 (1.6%) |
| Clinical data | ||
| Brain | ||
| GCS | 15 [13–15] | 13 [8–15] |
| GCS imputation using AVPU scale | 2135 (80.0%) | 579 (40.1%) |
| AVPU scale | ||
| Alert | 1841 (70.7%) | 697 (48.3%) |
| Verbal | 219 (8.4%) | 343 (23.8%) |
| Pain | 188 (7.2%) | 189 (13.1%) |
| Unresponsive | 356 (13.7%) | 214 (14.8%) |
| Medication for delirium | 34 (1.5%) | 5 (0.3%) |
| Respiratory | ||
| PaO2/FiO2 (mmHg) | 287.6 [160.0–374.0] | 197.0 [89.0–328.1] |
| Advanced ventilatory support | 799 (29.9%) | 556 (38.5%) |
| Non-invasive ventilation (high-flow nasal cannula, CPAP, BiPAP, NIV, and other noninvasive respiratory support modalities) | 404 (15.1%) | 310 (21.5%) |
| Invasive mechanical ventilation | 555 (20.8%) | 341 (23.6%) |
| Home ventilation | 10 (0.4%) | 0 (0%) |
| ECMO (veno-venous) | 4 (0.2%) | 0 (0%) |
| Cardiovascular | ||
| Mean arterial pressure (mmHg) | 49.8 ± 10.4 | 56.3 ± 13.3 |
| Norepinephrine use* | 2506 (93.9%) | 1312 (90.9%) |
| Norepinephrine dose (μg/kg/min) | 0.200 [0.085–0.300] (N=2506) | 0.238 [0.164–0.434] (N=1312) |
| Epinephrine use | 215 (8.1%) | 87 (6.0%) |
| Epinephrine dose (μg/kg/min) | 0.100 [0.050–0.300] (N=215) | 0.089 [0.056–0.130] (N=87) |
| Vasopressin use | 1056 (39.6%) | 239 (16.6%) |
| Vasopressin dose (IU/min) | 0.030 [0.030–0.040] (N=1056) | 0.033 [0.027–0.040] (N=239) |
| Dopamine use | 29 (1.1%) | 46 (3.2%) |
| Dopamine dose (μg/kg/min) | 10.000 [5.000–15.000] (N=29) | 2.525 [0.991–3.556] (N=46) |
| Dobutamine use | 31 (1.2%) | 23 (1.6%) |
| Dobutamine dose (μg/kg/min) | 5.000 [3.000–5.000] (N=31) | 3.098 [2.563–5.119] (N=23) |
| Any type of mechanical cardiovascular support including ECMO (veno-arterial) | 0 (0%) | 0 (0%) |
| Liver | ||
| Total bilirubin (mg/dL) | 1.2 [0.7–2.6] | 0.8 [0.5–1.5] |
| Kidney | ||
| Creatinine (mg/dL) | 1.5 [1.1–2.4] | 1.6 [1.0–2.6] |
| RRT at emergency department | 6 (0.2%) | 52 (3.6%) |
| RRT indicated (machine unavailability, ceiling of treatment, RRT after ICU admission, or other constraints) | 356 (13.3%) | 118 (8.2%) |
| Hemostasis | ||
| Platelets (×103/μL) | 107.0 [49.0–187.0] | 159.0 [94.0–229.5] |
| Lactate (mmol/L) |
4.4 [3.0–6.8] 5.3 ± 3.3 |
4.7 [3.1–7.9] 6.1 ± 4.4 |
| Length of stay in ED (hours) | 11.5 [6.8–21.4] | 10.5 [6.9–19.7] |
| ICU admission | 1563 (58.6%) | 898 (62.2%) |
| SOFA score | ||
| Total SOFA-1 score |
9 [7–11] 9.4 ± 3.2 |
9 [7–12] 9.5 ± 3.2 |
| Total SOFA-2 score |
8 [5–10] 8.2 ± 3.7 |
8 [6–11] 8.7 ± 3.7 |
| Outcomes | ||
| In-hospital mortality | 828 (31.0%) | 472 (32.7%) |
| 28-day mortality | 804 (30.1%) | 541 (37.5%) |
| 90-day mortality | 1156 (43.3%) | 682 (47.3%) |
Data are presented as mean ± standard deviation, median [interquartile ranges], or number (percentages), as appropriate
GCS Glasgow Coma Scale, AVPU Alert Verbal Pain Unresponsive, PaO2/FiO2 ratio of partial pressure of oxygen to fraction of inspired oxygen, CPAP continuous positive airway pressure, BiPAP bilevel positive airway pressure, NIV noninvasive ventilation, ECMO extracorporeal membrane oxygenation, ED emergency department, RRT renal replacement therapy, ICU intensive care unit, SOFA Sequential Organ Failure Assessment
*Norepinephrine was administered after 24 h in 163 (6.1%) patients in Cohort A and 131 (9.1%) patients in Cohort B who were diagnosed with delayed septic shock
Distribution of SOFA-1 and SOFA-2 scores
Distributions of SOFA-1 and SOFA-2 scores were similar between the two cohorts (Fig. 2 and 3). Total SOFA-1 scores were significantly higher than the SOFA-2 scores in both cohorts (p < 0.05). There were significant differences in SOFA-1 and SOFA-2 scores in all organ systems in both cohorts (p < 0.05), with the largest changes in the respiratory, cardiovascular, and kidney systems.
Fig. 2.
Distributions of SOFA-1 and SOFA-2 scores. A Overall distributions and organ system-based distributions of SOFA-1 and SOFA-2 scores in Cohort A. B Overall distributions and organ system-based distributions of SOFA-1 and SOFA-2 scores in Cohort B
Fig. 3.
Heatmaps of distributions of SOFA-1 and SOFA-2 scores. A Heatmap of overall and organ system-based distributions of SOFA-1 and SOFA-2 scores in Cohort A. B Heatmap of overall and organ system-based distributions of SOFA-1 and SOFA-2 scores in Cohort B
In the respiratory system, the proportion of patients with 2 points decreased, whereas the proportion with 3 points increased in SOFA-2. In the cardiovascular system, the proportion of patients with 4 points decreased, whereas the proportion of patients with 2 and 3 points increased. In the kidney system, the proportion of patients with 4 points increased in SOFA-2.
Mortality distribution of SOFA-1 and SOFA-2 scores
Mortality distributions of SOFA-1 and SOFA-2 scores were similar between the two cohorts (Fig. 4). Mortality generally increased with higher SOFA-1 and SOFA-2 scores. Relatively higher mortality was observed in the SOFA-2 score 0–1 group and the SOFA-1 score 2–3 group, particularly in Cohort B. The number of patients in these score groups was small.
Fig. 4.
Mortality distributions and heatmaps of SOFA-1 and SOFA-2 scores. A Mortality distributions and heatmaps of SOFA-1 and SOFA-2 scores in Cohort A. The numbers (n) above each bar indicate the total number of patients in the corresponding SOFA score group. B Mortality distributions and heatmaps of SOFA-1 and SOFA-2 scores in Cohort B
Performance of SOFA-1 and SOFA-2 scores for mortality
The AUROCs of the SOFA-2 score for mortality outcomes were significantly higher than those of the SOFA-1 score in both cohorts (Fig. 5 and Supplementary Table 2). In Cohort A, AUROCs were 0.734 (95% confidence interval [CI] 0.714–0.755) versus 0.702 (95% CI 0.680–0.724) for in-hospital mortality (p < 0.001), 0.719 (95% CI 0.698–0.740) versus 0.694 (95% CI 0.672–0.716) for 28-day mortality (p < 0.001), and 0.680 (95% CI 0.659–0.700) versus 0.658 (95% CI 0.637–0.679) for 90-day mortality (p < 0.001).
Fig. 5.
Receiver operating characteristic curves of SOFA-1 and SOFA-2 scores for outcomes. A Cohort A. B Cohort B
In Cohort B, AUROCs were 0.719 (95% CI 0.691–0.748) versus 0.671 (95% CI 0.640–0.701) for in-hospital mortality (p < 0.001), 0.737 (95% CI 0.710–0.763) versus 0.689 (95% CI 0.660–0.717) for 28-day mortality (p < 0.001), and 0.711 (95% CI 0.685–0.738) versus 0.666 (95% CI 0.638–0.693) for 90-day mortality (p < 0.001).
In organ-specific analyses, SOFA-2 demonstrated significantly higher AUROCs in the respiratory, cardiovascular, liver, and kidney systems in Cohort A and in the respiratory, cardiovascular, and kidney systems in Cohort B, whereas the hemostasis system showed lower AUROCs in Cohort A (Supplementary Table 3).
For in-hospital mortality, the calibration curves for the total SOFA-2 score showed generally closer agreement between predicted and observed mortality than those for SOFA-1 (Supplementary Figs. 1–2). SOFA-2 also had lower Brier scores in both cohorts, indicating lower overall prediction error (Supplementary Table 4). For the organ-specific scores, the calibration curves and Brier scores generally favored SOFA-2 for the respiratory, cardiovascular, liver, and kidney systems. In contrast, the hemostasis system had higher Brier scores with SOFA-2.
Compared with the SOFA-1 score, continuous NRI and IDI favored the SOFA-2 score for in-hospital, 28-day, and 90-day mortality (Supplementary Table 5). Continuous NRI was 0.509 (95% CI, 0.275–0.647, p < 0.001) for in-hospital mortality, 0.349 (95% CI, 0.192–0.562, p < 0.001) for 28-day mortality, and 0.276 (95% CI, 0.163–0.414, p < 0.001) for 90-day mortality in Cohort A and 0.633 (95% CI, 0.487–0.740, p < 0.001) for in-hospital mortality, 0.635 (95% CI, 0.474–0.758, p < 0.001) for 28-day mortality, and 0.624 (95% CI, 0.500–0.731, p < 0.001) for 90-day mortality in Cohort B, respectively. Corresponding IDI values were 0.042 (95% CI, 0.030–0.054, p < 0.001), 0.033 (95% CI, 0.021–0.044, p < 0.001), and 0.024 (95% CI, 0.015–0.033, p < 0.001) in Cohort A and 0.057 (95% CI, 0.043–0.074, p < 0.001), 0.062 (95% CI, 0.046–0.079, p < 0.001), and 0.055 (95% CI, 0.042–0.072, p < 0.001) in Cohort B for in-hospital, 28-day, and 90-day mortality, respectively.
Multivariable logistic regression analysis
After adjustment for age, sex, infection focus, comorbidities, lactate levels, and ICU admission, each increase in the SOFA-2 score was independently associated with increased in-hospital mortality, 28-day mortality, and 90-day mortality in both cohorts (Supplementary Table 6). In Cohort A, the adjusted odds ratios (aORs) per one-point increase in the SOFA-2 score were 1.265 (95% CI, 1.228–1.304), 1.250 (95% CI 1.214–1.288), and 1.203 (95% CI, 1.171–1.237) for in-hospital, 28-day, and 90-day mortality (all p < 0.001). Corresponding aORs in Cohort B were 1.193 (95% CI, 1.148–1.241), 1.215 (95% CI, 1.169–1.265), and 1.177 (95% CI, 1.134–1.224), respectively (all p < 0.001).
Subgroup analysis
Findings were consistent across temporal (2016–2020 and 2021–2024) and ICU admission subgroups. The total SOFA-2 score showed significantly higher AUROCs than the total SOFA-1 score in both cohorts (Supplementary Table 7).
Sensitivity analysis
Findings were robust across all sensitivity analyses. After imputing missing SOFA components as normal (N=2,685 in Cohort A and N=1525 in Cohort B), restricting the analysis to patients with an available GCS (N=534 in Cohort A and N=864 in Cohort B), excluding patients with metastatic cancer (N=1,588 in Cohort A and N=1411 in Cohort B), and excluding patients who required vasopressors only after the first 24 h (N=2511 in Cohort A and N=1312 in Cohort B), predictive performance of SOFA-1 and SOFA-2 remained consistent with the primary analysis (Supplementary Table 7).
Pooled analysis
Forty-one duplicate patients were identified and excluded from the pooled analyses. In the pooled analysis (N=4071), findings were consistent with the cohort-specific analyses, including score distributions, mortality distributions, discrimination, calibration, reclassification, and multivariable associations of total SOFA-1 and SOFA-2 scores (Supplementary Figs. 3–5; Supplementary Tables 2–6). In organ-specific analyses, the respiratory, cardiovascular, liver, and kidney systems of SOFA-2 showed significantly higher AUROCs for in-hospital mortality than the corresponding SOFA-1 systems, whereas the hemostasis system showed lower AUROCs (Supplementary Table 3). Results of subgroup and sensitivity analyses are summarized in Supplementary Table 9.
Discussion
In this study, we validated the updated SOFA-2 score in ED patients with septic shock. We found significant differences in the distribution of SOFA-1 and SOFA-2 scores, with marked changes in the respiratory, cardiovascular, and kidney systems reflecting the incorporation of contemporary organ support modalities. Despite these distributional differences, mortality patterns across score ranges were similar. Importantly, SOFA-2 demonstrated better discrimination than SOFA-1 for in-hospital, 28-day, and 90-day mortality. As expected, the SOFA-2 score was independently associated with outcomes. These findings were consistent across both complementary validation cohorts. These findings support the external validity of SOFA-2 for prognostic assessment in ED patients with septic shock.
The distributional shifts between SOFA-1 and SOFA-2 scores observed in these ED cohorts were generally similar to those reported in the original SOFA-2 development study [16,17], except for the cardiovascular system. This difference may be explained primarily by differences in patient population rather than by differences between ED and ICU settings. Previous SOFA-2 studies included heterogeneous ICU populations [16,17], whereas the present study focused specifically on patients with septic shock. Because septic shock is characterized by vasopressor requirement, patients clustered within the higher cardiovascular SOFA-1 categories were redistributed across a broader range of SOFA-2 cardiovascular scores, resulting in a more pronounced shift in score distribution. However, ICU-based studies focusing specifically on septic shock remain limited. Further studies are needed to determine whether these distributional shifts in ED septic shock are similar to or distinct from those observed in ICU septic shock populations.
The cardiovascular, respiratory, and kidney systems were the main contributors to the improved performance of SOFA-2. Compared with SOFA-1, SOFA-2 incorporates updated thresholds and contemporary organ support modalities, including revised vasopressor categories, noninvasive respiratory support, and renal replacement therapy criteria [16,17]. These modifications better reflect current clinical practice and provided a broader severity spectrum in ED patients with septic shock, which may contribute to the improved prognostic performance of SOFA-2 in the ED.
No substantial changes were observed in the distribution or performance of the neurologic component between SOFA-1 and SOFA-2. SOFA-2 revised the GCS thresholds and added the requirement for delirium medication [16,17]. The limited impact of these modifications may reflect the rarity of delirium medication use in the ED and the use of AVPU-based imputation for missing GCS values. This is an important limitation because AVPU-based imputation may not fully capture the granularity of neurologic dysfunction intended by the revised SOFA-2 criteria. Further ED-based validation studies with GCS data and higher rates of delirium treatment are warranted.
For the liver system, SOFA-2 showed a slight improvement in predictive performance in Cohort A, whereas no significant difference was observed in Cohort B. Given that the revision was limited to bilirubin thresholds, its impact on prognostic performance appears modest in ED septic shock.
Notably, the SOFA-2 hemostasis system showed slightly lower performance in Cohort A and no significant difference in Cohort B. This finding may be related to the non-linear association between platelet count and mortality observed during SOFA-2 development, in contrast to the more monotonic relationships observed for other organ-specific variables [16,17]. This non-linear relationship may explain the non-significant or slightly lower performance of the SOFA-2 hemostasis system compared with the SOFA-1 hemostasis system.
SOFA-2 demonstrated significantly better predictive performance than SOFA-1 in both cohorts; however, the absolute differences in AUROC were modest. Similar findings have been reported in recent ICU-based validation studies [27–29]. Nevertheless, in this study, SOFA-2 also showed generally closer agreement between predicted and observed in-hospital mortality on the calibration curves. It also showed lower overall prediction error as reflected by the Brier scores. In addition, it demonstrated improved reclassification as assessed using continuous NRI and IDI. These metrics suggest incremental gains in risk prediction but should be interpreted as complementary measures rather than definitive evidence of clinical utility. In addition, SOFA-2 incorporates contemporary organ support modalities and updated clinical practices, which may enhance its relevance in current critical care settings.
Mortality generally increased with higher SOFA-1 and SOFA-2 scores. This pattern parallels that reported in the original SOFA-2 development study [16,17], supporting the validity of both scores in the ED. However, mortality was relatively higher in the SOFA-2 0–1 and SOFA-1 0–3 score ranges than in other low-score groups. These low-score strata contained relatively few patients, making mortality estimates more susceptible to the influence of a small number of deaths. In addition, some patients may have developed worsening organ dysfunction or vasopressor requirements after the initial 24-h assessment period, particularly in Cohort B, which included a broader spectrum of ED septic shock presentations. Therefore, increased mortality in low-score groups may partly reflect delayed recognition, deterioration, or resuscitation in the ED.
The predictive performance observed in this study was lower than that reported in previous ICU-based SOFA-2 studies [16,17,27–30], which may reflect differences in study populations and timing of assessment. Previous studies evaluated heterogeneous ICU populations, whereas the present study focused exclusively on ED patients with septic shock. ED patients are more likely to undergo early resuscitation and achieve transient stabilization before ICU admission, potentially affecting physiologic parameters during this interval. Furthermore, because septic shock is defined by vasopressor requirement and current Surviving Sepsis Campaign guidelines recommend norepinephrine as the first-line vasopressor, most patients were categorized into relatively high cardiovascular SOFA strata (SOFA-1: 3–4 points; SOFA-2: 2–4 points). This clustering may have reduced discriminatory performance. Given that serial SOFA-2 assessment has shown improved performance in ICU populations [16,17], serial SOFA-2 measurement in the ED may further enhance prognostic performance and help identify high-risk patients with initially low scores.
The SOFA score was not originally developed as a mortality prediction tool but as a serial measure of organ dysfunction severity in critically ill patients [10–12]. Both SOFA-1 and SOFA-2 scores were developed and validated using ICU mortality as a measure of severity. In this study, we extended validation to longer-term outcomes, including 28-day and 90-day mortality, and found that SOFA-2 provided better prognostic discrimination than SOFA-1. Nevertheless, the SOFA-2 score should not be interpreted as a standalone mortality prediction tool but rather as a measure of organ dysfunction.
In this study, SOFA scores were calculated using the worst values recorded during the first 24 h in the ED. Accordingly, the scores may reflect both baseline severity and the effects of treatment and organ support during the early ED course rather than severity at presentation alone. In addition, some organ support modalities incorporated in the SOFA-2 score may be limited in the prehospital setting or during initial ED triage. Therefore, further validation and potential modification of SOFA-2 are required for the assessment of initial severity in ED and prehospital settings.
This study has several strengths. First, to our knowledge, this is the first study to validate the SOFA-2 score specifically in the ED. We used two prospective cohorts with complementary characteristics, including early hypoperfusion septic shock and broader ED septic shock populations, and found consistent findings. These findings support the generalizability of the prognostic performance of SOFA-2 across different ED septic shock populations. Second, we evaluated multiple mortality outcomes, including longer-term outcomes compared with the previous development study [16,17]. Third, we performed multivariable analyses after adjustments for clinically relevant confounders and found an independent association between the SOFA-2 score and outcomes. However, these analyses were not intended to compare the predictive performance of SOFA-2 and SOFA-1, but rather to confirm the independent prognostic association of SOFA-2. Fourth, multiple sensitivity analyses yielded consistent results, supporting the statistical robustness of our findings.
There are some limitations. First, this study was conducted in a single country and predominantly included Asian patients. This may limit generalizability to other populations. Second, sepsis was defined using the Sepsis-3 definition based on SOFA-1. Consequently, patient selection was not independent of the original scoring system, representing a conceptual limitation in the evaluation of SOFA-2. Furthermore, the parallel use of SOFA-1 for sepsis diagnosis and SOFA-2 for prognostic assessment may limit the direct clinical implementation of SOFA-2 under the current sepsis definitions. Third, we analyzed patients with septic shock in the ED. The SOFA-2 score should be validated in other critically ill conditions in the ED, which requires further studies. In addition, as this study focused on septic shock, further studies are warranted for validation of the ED-based SOFA-2 score in patients with infection and sepsis. Fourth, GCS was imputed using the AVPU scale for some patients. The primary analysis depends substantially on AVPU-based GCS imputation, especially in Cohort A. However, conversion of the AVPU scale to GCS is well validated [22,24,26], and only minor adjustments were made to the GCS in the brain system of the SOFA-2 score [16,17]. These factors may partially mitigate the impact of this limitation. In addition, consistent findings in sensitivity analyses restricted to patients with available GCS further support the robustness of our results. Fifth, multiple organ-specific, subgroup, sensitivity, and pooled analyses were performed without correction for multiple comparisons. These analyses should be interpreted as exploratory. Despite these limitations, the large sample size, multicenter design, and consistency of the results across two complementary cohorts strengthen the validity of the study’s conclusions.
Conclusions
In two prospective cohorts of ED patients with septic shock, the updated SOFA-2 score showed modestly improved discrimination and more favorable reclassification metrics for in-hospital, 28-day, and 90-day mortality, as well as closer agreement between predicted and observed in-hospital mortality with lower overall prediction error, compared with the SOFA-1 score. These findings support the external validation of SOFA-2 in ED patients with septic shock. Further studies are needed to determine its role in clinical practice and broader implementation in ED settings.
Supplementary Information
Acknowledgements
We would like to express our sincere gratitude to the Korean Shock Society for their valuable support in research collaboration.
Author contributions
Conceptualization: SA, SMK, TGS, KK; Methodology: all authors; Validation: all authors; Formal Analysis: SA, SMK, TGS, KK; Investigation: all authors; Data Curation: all authors; Resources: JHP, WYK, TGS, KK; Writing: SA, SMK, TGS, KK; Writing – Review and Editing: all authors; Visualization: SA; Supervision: JHP, WYK, TGS, KK; Funding acquisition: KK, TGS; All authors have made substantial contributions to drafting the article. All authors read and approved the final article.
Funding
This research was supported by a grant of Patient-Centered Clinical Research Coordinating Center (PACEN) funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2024-00398566). This research was also supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (RS-2024-00335935). These funding sources had no role in the design of this study and will not have any role during its execution, analyses, interpretation of the data, or decision to submit results.
Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
This study was approved by the Institutional Review Board of the Korea University Ansan Hospital (2026AS0001), which waived the requirement for informed consent due to the nature of the observational study.
Consent for publication
Not applicable.
Competing interests:
The authors declare that they have no competing interests.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Sejoong Ahn and Sang-Min Kim have contributed equally to this work.
Contributor Information
Tae Gun Shin, Email: drshin88@gmail.com.
Kyuseok Kim, Email: dreinstein70@gmail.com.
References
- 1.Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. 2016;315(8):801–10. 10.1001/jama.2016.0287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47(11):1181–247. 10.1007/s00134-021-06506-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Crit Care Med. 2021;49(11):e1063–143. 10.1097/CCM.0000000000005337. [DOI] [PubMed] [Google Scholar]
- 4.GBD 2021 Global Sepsis Collaborators. Global, regional, and national sepsis incidence and mortality, 1990-2021: a systematic analysis. Lancet Glob Health. 2025;13(12):e2013–26. 10.1016/S2214-109X(25)00356-0. [DOI] [PubMed] [Google Scholar]
- 5.Shin TG, Ko E, Han SH, Kim T, Choi DH. Epidemiology of sepsis in emergency departments: insights from the National Emergency Department Information System (NEDIS) database in Korea, 2018-2022. Clin Exp Emerg Med. 2025;12(3):185–7. 10.15441/ceem.25.174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wang HE, Jones AR, Donnelly JP. Revised national estimates of emergency department visits for sepsis in the United States. Crit Care Med. 2017;45(9):1443–9. 10.1097/CCM.0000000000002538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Strehlow MC, Emond SD, Shapiro NI, Pelletier AJ, Camargo CA Jr. National study of emergency department visits for sepsis, 1992 to 2001. Ann Emerg Med. 2006;48(3):326-331.e3313. 10.1016/j.annemergmed.2006.05.003. [DOI] [PubMed] [Google Scholar]
- 8.Julián-Jiménez A, Supino M, López Tapia JD, et al. Sepsis in the emergency department: key points, controversies, and proposals for improvements in Latin America Puntos clave y controversias sobre la sepsis en los servicios de urgencias: propuestas de mejora para Latinoamérica. Emergencias. 2019;31(2):123–35. [PubMed] [Google Scholar]
- 9.Knack SKS, Scott N, Driver BE, et al. Early physician gestalt versus usual screening tools for the prediction of sepsis in critically ill emergency patients. Ann Emerg Med. 2024;84(3):246–58. 10.1016/j.annemergmed.2024.02.009. [DOI] [PubMed] [Google Scholar]
- 10.Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related organ failure assessment) score to describe organ dysfunction/failure. On behalf of the working group on sepsis-related problems of the European society of intensive care medicine. Intensive Care Med. 1996;22(7):707–10. 10.1007/BF01709751. [DOI] [PubMed] [Google Scholar]
- 11.Vincent JL, de Mendonça A, Cantraine F, et al. Use of the SOFA score to assess the incidence of organ dysfunction/failure in intensive care units: results of a multicenter, prospective study Working group on “sepsis-related problems” of the European society of intensive care medicine. Crit Care Med. 1998;26(11):1793–800. 10.1097/00003246-199811000-00016. [DOI] [PubMed] [Google Scholar]
- 12.Moreno R, Vincent JL, Matos R, et al. The use of maximum SOFA score to quantify organ dysfunction/failure in intensive care results of a prospective, multicentre study working group on sepsis related problems of the ESICM. Intensive Care Med. 1999;25(7):686–96. 10.1007/s001340050931. [DOI] [PubMed] [Google Scholar]
- 13.de Grooth HJ, Geenen IL, Girbes AR, Vincent JL, Parienti JJ, Oudemans-van Straaten HM. SOFA and mortality endpoints in randomized controlled trials: a systematic review and meta-regression analysis. Crit Care. 2017;21(1):38. 10.1186/s13054-017-1609-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Kim TH, Jeong D, Park JE, et al. Prognostic accuracy of initial and 24-h maximum SOFA scores of septic shock patients in the emergency department. Heliyon. 2023;9(9):e19480. 10.1016/j.heliyon.2023.e19480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Pölkki A, Pekkarinen PT, Takala J, Selander T, Reinikainen M. Association of Sequential Organ Failure Assessment (SOFA) components with mortality. Acta Anaesthesiol Scand. 2022;66(6):731–41. 10.1111/aas.14067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ranzani OT, Singer M, Salluh JIF, et al. Development and validation of the Sequential Organ Failure Assessment (SOFA)-2 score. JAMA. 2025;334(23):2090–103. 10.1001/jama.2025.20516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Moreno R, Rhodes A, Ranzani O, et al. Rationale and methodological approach underlying the development of the Sequential Organ Failure Assessment (SOFA)-2 score: a consensus statement. JAMA Netw Open. 2025;8(10):e2545040. 10.1001/jamanetworkopen.2025.45040. [DOI] [PubMed] [Google Scholar]
- 18.Moreno R, Rhodes A, Piquilloud L, et al. The Sequential Organ Failure Assessment (SOFA) score: has the time come for an update? Crit Care. 2023;27(1):15. 10.1186/s13054-022-04290-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Moreno R, Singer M, Rhodes A. Por que o escore Sequential Organ Failure Assessment precisa ser atualizado? Critical Care Sci. 2024;36:e20240296en. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Shin TG. Assessment of organ failure in sepsis patients in the emergency department: clinical evaluation, Sequential Organ Failure Assessment (SOFA) score, and future perspectives. Clin Exp Emerg Med. 2024;11(4):327–30. 10.15441/ceem.24.330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Seymour CW. A revision to organ failure assessment in critically ill patients. JAMA. 2025;334(23):2075–7. 10.1001/jama.2025.20255. [DOI] [PubMed] [Google Scholar]
- 22.Lee HJ, Ko BS, Ryoo SM, et al. Modified cardiovascular SOFA score in sepsis: development and internal and external validation. BMC Med. 2022;20(1):263. 10.1186/s12916-022-02461-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Jin BY, Lee S, Kim W, et al. Association of metformin administration after septic shock with short-term and long-term survival in septic shock patients with diabetes. Ann Intensive Care. 2025;15(1):68. 10.1186/s13613-025-01490-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kelly CA, Upex A, Bateman DN. Comparison of consciousness level assessment in the poisoned patient using the alert/verbal/painful/unresponsive scale and the Glasgow Coma Scale. Ann Emerg Med. 2004;44(2):108–13. [DOI] [PubMed] [Google Scholar]
- 25.Ryoo SM. Updated Sequential Organ Failure Assessment (SOFA)-2 score: major changes and emergency department perspectives. Clin Exp Emerg Med. 2025;12(4):303–5. 10.15441/ceem.25.285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Ramgopal S, Cash RE, Okubo M, Martin-Gill C. Mapping Glasgow Coma scale to AVPU scores among adults in the prehospital setting. Am J Emerg Med. 2024;86:1–4. 10.1016/j.ajem.2024.09.043. [DOI] [PubMed] [Google Scholar]
- 27.Iwasaki Y, Kinoshita T, Yoshimura J, et al. Admission-day SOFA-2 versus SOFA for mortality discrimination: external validation in a large multicenter ICU database. Crit Care. 2026. 10.1186/s13054-026-06020-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Zhu H, Li P, Wang B, et al. Impact of the updated SOFA-2 score on sepsis diagnosis and prognosis: a retrospective multicenter cohort study. Crit Care. 2026. 10.1186/s13054-026-06070-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wei S, Wang T, Li C, et al. SOFA-2 score predicts mortality in pneumonia-associated sepsis: a retrospective cohort study. Crit Care. 2026. 10.1186/s13054-026-06027-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Liufu R, Fu XY, Shi XQ, et al. Performance of sequential organ failure assessment 1 versus 2 for sepsis identification in patients with suspected infection: a multicenter retrospective cohort study. Crit Care. 2026. 10.1186/s13054-026-05981-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.