Enrichment using the Phoenix Sepsis Score in children: heterogeneous treatment effects in a hydrocortisone target trial emulation

Prognostic and predictive enrichment strategies could be used to address the heterogeneity of pediatric sepsis and identify subgroups of patients most likely to benefit from targeted therapies. Our aim was to determine if the use of the Phoenix Sepsis Score (PSS) can identify critically ill children with community-acquired sepsis who are most likely to benefit from early hydrocortisone therapy. This was an observational cohort study and target trial emulation in 15 pediatric intensive care units (PICUs) in the United States, including a development cohort and replication in a validation cohort. Patients < 18 years of age admitted to a PICU with confirmed or suspected community-acquired infection between 2010 and 2019 were included. Patients who were treated with systemic stress dose hydrocortisone in the first 24 h of PICU admission were compared with untreated patients across risk groups. There were 48,345 patients included in the main analysis. A PSS ≥ 4 points in the first 12 h of PICU admission had the best discrimination of outcomes, with a 10-fold higher mortality than those with a PSS < 4 in both the development and the validation cohorts. In the target trial emulation analysis, after adjusting for covariates associated with hydrocortisone use (age, malignancy, severity of illness, refractory shock status, and hospital), patients who received early hydrocortisone and were in the high-risk phenotype (PSS ≥ 4 in first 12 h) had a lower risk of in-hospital mortality (adjusted risk ratio [aRR] 0.72 [95% confidence interval [CI] 0.65–0.80]), while low-risk patients had increased risk (aRR 1.90 [95%CI 1.66–2.16], interaction p-value < 0.001) in the development cohort. The results were replicated in the validation cohort (aRR 0.79 [95%CI 0.68, 0.92] vs. aRR 2.95 [95%CI 2.43, 3.59], interaction p-value < 0.001). A highly generalizable risk stratification strategy using the easy-to-calculate PSS can be used for both prognostic and predictive enrichment in critically ill children with community-acquired sepsis. Further external and prospective validation of our results is warranted. Not applicable.