Background: There is increasing interest in conceptualizing severe COVID-19 (coronavirus disease 2019) as viral sepsis – a syndrome of acute multi-organ dysfunction consequent to a dysregulated host response to a SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) infection (1–3). The validity of clinical sepsis criteria in identifying patients at risk of adverse outcomes is, however, not well defined. The Surviving Sepsis Campaign’s COVID-19 guidelines do not endorse any criteria for operationally characterizing a patient as having severe COVID-19 or viral sepsis from SARS-CoV-2 (4). We sought to study whether conventional sepsis criteria and early warning scores are associated with adverse outcomes independently of baseline health status.

Methods: In a 600 bed academic medical center, we identified all encounters where a SARS-CoV-2 infection was diagnosed upon presentation to the emergency room (ER) in the first six months of our pandemic experience (n = 376; March-August 2020). From data recorded in the ER, we determined the most deranged values for the Systemic Inflammatory Response Syndrome (SIRS), Sequential Organ Failure Assessment (SOFA), quick SOFA (qSOFA) and National Early Warning Score (NEWS). We reviewed all charts to ascertain baseline health markers like physical and cognitive function, comorbidities and baseline SOFA score. As outlined in Figure 1, we measured inter-rater reliability of chart reviews with Krippendorff’s alpha. We assessed predictive validity using measures of discrimination for adverse outcomes (ICU transfer or in-hospital mortality – 21%) such as fold change in rates of outcome and area under receiver operating characteristic curve (AUROC).

Results: We found that markers of baseline health like age, mobility, obesity, Charlson Comorbidity Index (CCI) and baseline SOFA, had strong discrimination for adverse outcomes typical of severe COVID-19 (AUROC 0.82). Among the sepsis criteria (SIRS ≥ 2, SOFA ≥ 2 & qSOFA ≥ 2), only SOFA improved model performance to a statistically significant degree (AUROC: 0.86 for SOFA vs 0.82 at baseline, p = 0.01; fold-rise in rate of outcome: 7.2, 95% CI: 4.3-14.5). The NEWS score (sepsis defined as NEWS ≥ 6) performed as well as SOFA and outperformed SIRS or qSOFA (AUROC: 0.87 for NEWS v/s 0.82 at baseline, p < 0.01; fold-rise in rate of outcome: 6.8, 95% CI: 4.6-12). Figure 2 displays these results.

Conclusions: Our study confirms that baseline health plays a substantial role in determining outcomes. It also suggests that, similar to their use in general sepsis, the SOFA and NEWS scores may help identify patients at higher risk for poor outcomes typical of severe COVID-19. Given the simplicity of implementation, we envision the NEWS score being useful for surveillance of infected patients in non-ICU settings (ER, acute care, nursing homes). And, given the well-established utility of the SOFA score in clinical trials, we envision it being useful as a patient characterization tool and/or endpoint for COVID-19 trials.

IMAGE 1: Figure 1: Data collection and cohort characteristics.

IMAGE 2: Figure 2: Modelling severity of COVID-19 with clinical sepsis criteria.