Outcome of Children With Systemic Rheumatologic and Autoinflammatory Diseases Admitted to a Pediatric Intensive Care Unit: A Retrospective Cohort Study With Development of a Prognostic Risk-stratification Tool

Objective: Children with systemic rheumatologic diseases (SRD) and autoinflammatory diseases can experience life-threatening deterioration requiring intensive care. We conducted a retrospective cohort study with 2 complementary components: (1) a descriptive analysis of clinical presentations and outcomes; and (2) an exploratory derivation of a bedside risk-stratification tool, intended as a candidate severity overlay on established pediatric intensive care unit (PICU) scoring systems rather than as a fully validated general-purpose prognostic model. Design: Retrospective observational cohort study with exploratory prognostic model development. Propensity-score matching was used as a secondary, descriptive analysis to contextualize outcomes against comparable populations and is not intended to support causal inference. Setting: Single academic pediatric intensive care unit (January 2005 to December 2015). Patients: Forty children with confirmed SRD requiring PICU admission (first admission per patient) and propensity-matched controls (2:1 ratio). Measurements and Main Results: Nearly two-thirds (62.2%) of patients experienced their first disease manifestation as a life-threatening crisis. Hemophagocytic lymphohistiocytosis/macrophage activation syndrome (HLH/MAS) was associated with the majority of deaths, accounting for 83.3% of fatalities despite representing only 20% of admissions (disease-specific mortality 62.5% vs. 6.3% for other diagnoses combined, p <0.001). Backward stepwise logistic regression identified 3 predictors that were combined, with equal weighting, into the exploratory PHV Score (PRISM-HLH-Vasoactive): Pediatric Risk of Mortality III (PRISM-III) at 24 hours >20, HLH/MAS diagnosis, and requirement for ≥2 vasoactive agents. In this derivation sample the score showed encouraging discrimination [apparent area under the curve (AUC) 0.91, 95% CI 0.83-0.99; optimism-corrected AUC 0.88]; however, given the very limited event count (n=6 deaths), wide CIs, and the fact that the PHV Score partially incorporates PRISM-III, head-to-head AUC comparisons with PRISM-III (AUC 0.72), PELOD-2 (Pediatric Logistic Organ Dysfunction-2; AUC 0.68), and pediatric SOFA (Sequential Organ Failure Assessment; AUC 0.64) should be interpreted with caution and regarded as hypothesis-generating. At the optimal cutoff of ≥2 (Youden index), sensitivity was 83.3% and specificity 91.2%; predictive values are prevalence-dependent and may not generalize to centers with different case mixes. Conclusions: In this retrospective single-center cohort, HLH/MAS was associated with most PICU deaths among children with systemic rheumatologic and autoinflammatory diseases. The PHV Score is an exploratory, candidate bedside tool that appears to refine risk stratification in our cohort but requires prospective multicenter external validation.