External validation of the Phoenix Sepsis Score in Singapore: a retrospective cohort study.

Abstract

The Phoenix Sepsis Score (PSS) was derived and validated in 10 healthcare systems internationally. We sought to perform an external validation of the PSS among children with suspected infection in a high-resource country in Asia. Retrospective cohort study. A tertiary paediatric hospital in Singapore, between 1 January 2020 and 31 December 2024. Children <18 years old hospitalised with suspected infection who received broad-spectrum intravenous antibiotics. Application of the PSS. We defined sepsis as PSS ≥2, as per the Phoenix criteria. The main outcome was in-hospital mortality. We described the performance of PSS using area under the receiver operating characteristic curve (AUROC) and area under the precision recall curve (AUPRC). Among 25 202 children (median age 3.3 years (IQR, 0.7-8.4)) with suspected infection, 13 484 were males (53.5%). 82 (0.3%) children died. There were 660 children (2.6%) with PSS ≥2 and 411 (1.6%) met criteria for septic shock. A PSS ≥2 was associated with a mortality risk of 8.5% (56/660) and predicted for mortality with a sensitivity of 68.3% (95% CI 57.5% to 78.4%) and specificity 97.6% (95% CI 97.4% to 97.8%). PSS had an AUROC of 0.90 (95% CI 0.86 to 0.94) and AUPRC of 0.36 (95% CI 0.24 to 0.47) in our study population, outperforming that of the original cohort. A PSS ≥2 was highly specific when predicting mortality among children with suspected infection. Our study confirms that the Phoenix criteria can be used as a data-driven benchmark for sepsis research.