Validation of the new Sequential Organ Failure Assessment 2 score in community-acquired pneumonia

Extract

Community-acquired pneumonia (CAP) is an acute infectious disease that may be accompanied by organ dysfunction. During the acute episode, the respiratory, renal, cardiovascular/haemodynamic and neurological systems are most frequently affected. The extent of organ dysfunction correlates closely with prognosis in CAP [1]. Early and accurate assessment of the degree of organ involvement is therefore essential to predict outcomes and identify patients at risk of complications. The Sequential Organ Failure Assessment (SOFA) score, developed in 1996, was originally intended to support the management of critically ill patients in the intensive care unit (ICU) [2].

Shareable abstract

The Sequential Organ Failure Assessment (SOFA) score was updated to the SOFA-2 version. Its use is validated for initial severity assessment in community-acquired pneumonia outside the intensive care unit. https://bit.ly/3QsAQwT

To the Editor:

Community-acquired pneumonia (CAP) is an acute infectious disease that may be accompanied by organ dysfunction. During the acute episode, the respiratory, renal, cardiovascular/haemodynamic and neurological systems are most frequently affected. The extent of organ dysfunction correlates closely with prognosis in CAP [1]. Early and accurate assessment of the degree of organ involvement is therefore essential to predict outcomes and identify patients at risk of complications. The Sequential Organ Failure Assessment (SOFA) score, developed in 1996, was originally intended to support the management of critically ill patients in the intensive care unit (ICU) [2]. Since then, it has been widely adopted. In 2016, the Sepsis-3 consensus defined sepsis as life-threatening organ dysfunction resulting from a dysregulated host response to infection [3] and incorporated the SOFA score into its operational criteria. These criteria have subsequently been validated in patients with CAP [4, 5]. More recently, the SOFA score has been updated to SOFA-2, which incorporates contemporary organ-support modalities and revised scoring thresholds [6]. However, as acknowledged by its developers, SOFA-2 was derived and validated exclusively in ICU cohorts, and its applicability to patients managed outside the ICU remains uncertain. The aim of the present study, therefore, is to evaluate the performance of the SOFA-2 score in patients admitted to hospital with CAP.

We conducted a post hoc analysis of a prospective cohort from La Fe University and Polytechnic Hospital (Valencia, Spain) [7, 8] comprising immunocompetent adults (≥18 years of age) hospitalised with CAP. The primary outcome was in-hospital mortality. The study was approved by the Biomedical Research Ethics Committee of Hospital La Fe (2013/0204) and written informed consent was obtained from all participants. Variables used in calculations were derived from laboratory and clinical data recorded at emergency department presentation. The arterial oxygen tension/inspiratory oxygen fraction (F_IO2) ratio was calculated using arterial blood gas values and the corresponding F_IO2; when arterial blood gas analysis was not available, the peripheral oxygen saturation/F_IO2 ratio was used as a surrogate, as per the score definition. The Glasgow Coma Scale was systematically assessed in all patients at the initial evaluation. All variables were collected at presentation, with no additional time window applied for score calculation.

Continuous variables were summarised as median (interquartile range) and categorical variables as n (%). The proportion of in-hospital deaths for each SOFA-1 and SOFA-2 score was illustrated. Predictive validity for the primary outcome was assessed using the areas under the receiver operating characteristic curves (AUC) for SOFA-1, SOFA-2, CURB-65 (confusion, urea >7 mmol·L⁻¹, respiratory rate ≥30 breaths per min, blood pressure <90 mmHg (systolic) or ≤60 mmHg (diastolic), age ≥65 years) and the Pneumonia Severity Index (PSI), based on values obtained in the emergency department. Calibration was examined with the Hosmer–Lemeshow test and Estimated Calibration Index (ECI), and overall model performance was evaluated using the scaled Brier score and Nagelkerke's R². A sensitivity analysis assessed the consistency of findings between normal-value imputation and complete-case results. Statistical analyses were performed in R (version 4.5.2) using the pROC (version 1.19.0.1) and DescTools (version 0.99.60) packages.

A total of 1842 patients were enrolled. Only 95 patients had missing SOFA-1 and SOFA-2 scores due to absent bilirubin values (hepatic domain). The median age was 71 (58–80) years; 1126 (61.1%) were men. Chronic respiratory disease was present in 33.3% and chronic heart disease in 30.9%. The median CURB-65 score and PSI were 1 (1–2) and 82 (63–101), respectively. Median SOFA-1 and SOFA-2 scores were 2 (1–3) and 1 (0–2) points. 92 (5.0%) patients required direct ICU admission and 122 (6.6%) required ICU admission at any point during hospitalisation. Overall, 53 (2.9%) patients died during hospitalisation.

In-hospital mortality according to SOFA-1 and SOFA-2 scores is shown in figure 1a. As expected, higher scores on both scales were associated with increased mortality. Discrimination for in-hospital mortality is presented in figure 1b. No significant calibration issues were identified using the Hosmer–Lemeshow test (p=0.75, p=0.37, p=0.87, p=0.56 and p=0.99 for SOFA-1, SOFA-2, CURB-65, PSI and the Infectious Diseases Society of America/American Thoracic Society (IDSA/ATS) minor criteria, respectively) or the ECI (IDSA/ATS minor criteria, 0.0001; SOFA-1, 0.030; PSI, 0.041; SOFA-2, 0.043; and CURB-65, 0.048), indicating overall good calibration across scores. Overall performance was highest for the IDSA/ATS minor criteria (scaled Brier score 0.116; R²=0.255). In comparison, SOFA-1, SOFA-2, CURB-65 and PSI had scaled Brier scores of 0.058, 0.067, 0.050 and 0.098, and R²=0.142, R²=0.161, R²=0.143 and R²=0.240, respectively. Sensitivity analyses using multiple imputation yielded findings consistent with the complete-case results (data not shown).

FIGURE 1.

In-hospital mortality and score performance. a) In-hospital mortality according to Sequential Organ Failure Assessment (SOFA)-1 and SOFA-2 scores. Points represent medians and whiskers indicate the 95% confidence interval. b) Discriminatory performance for in-hospital mortality was assessed using the areas under the receiver operating characteristic curves (AUC) and 95% confidence intervals. CURB-65: confusion, urea >7 mmol·L⁻¹, respiratory rate ≥30 breaths per min, blood pressure <90 mmHg (systolic) or ≤60 mmHg (diastolic), age ≥65 years; PSI: Pneumonia Severity Index; IDSA/ATS: Infectious Diseases Society of America/American Thoracic Society.

The main findings of our study were that the SOFA-2 score proved to be a reliable tool for the initial severity assessment of hospitalised CAP patients and accurately predicted in-hospital mortality. Furthermore, SOFA-2 demonstrated superior predictive performance compared with SOFA-1 and CURB-65, although it did not outperform the PSI or the IDSA/ATS minor criteria.

The predictive performance of SOFA-2 in this CAP cohort (AUC 0.79, 95% CI 0.71–0.84) was virtually identical to that reported in ICU patients by Ranzani et al. [6] (AUC 0.79, 95% CI 0.76–0.81); despite being applied to a cohort of CAP patients using values obtained in the emergency department. This supports the external validity of SOFA-2 and indicates that it is reliable for patients with CAP outside the ICU. As the Sepsis-3 definition currently incorporates the original SOFA score for clinical implementation [3], adapting SOFA-2 to sepsis criteria may support its use in CAP.

While PSI outperformed SOFA-2 and CURB-65 showed similar discrimination, our results underscore the need to advance CAP-specific prognostic models. Future efforts should aim to refine these tools, incorporating modern organ dysfunction measures, to address the heterogeneity of CAP and evolving sepsis definitions.

This study has limitations. It was conducted at a single centre in Spain, which may limit generalisability. Nevertheless, the use of a prospective CAP cohort ensured high-quality data. Outcomes were objective, with minimal loss to follow-up, reducing the likelihood of bias. Another limitation is the relatively low number of deaths in our cohort, which may affect the precision of discrimination and calibration estimates. Therefore, external validation and potential recalibration may be required in populations with higher baseline risk. In addition, treatment limitations were not considered for the study. Finally, we were unable to distinguish between acute and chronic organ dysfunction. Pre-existing organ impairment may increase baseline SOFA values independent of the acute illness, potentially inflating scores and contributing to differences compared with ICU-derived cohorts.

In conclusion, SOFA-2 can support initial risk assessment of CAP patients in the emergency department by reflecting acute organ dysfunction. Nevertheless, CAP-specific scores such as PSI, which integrate age and comorbidity burden, remain superior for mortality prediction. Our findings reaffirm the central role of disease-specific prognostic models, with SOFA-2 serving as a complementary tool that adds value to early patient evaluation outside the ICU.