Community-acquired pneumonia (CAP) remains a high-volume driver of ED visits, and yet, risk stratification tools tailored to well-resourced emergency departments are limited. A new study from the Pediatric Emergency Research Network (PERN) aimed to change that.

Published in The Lancet Child & Adolescent Health, this prospective cohort study of over 2200 children across 73 Emergency Departments in 14 countries developed and validated clinical prediction models to help clinicians distinguish mild CAP from moderate or severe cases, and ultimately guide decisions about hospital admission.

What Did They Do?

  • Population: Children 3 months to <14 years with clinically diagnosed CAP (radiographic confirmation not required).
  • Setting: 73 EDs across 14 countries (primarily high-income nations).
  • Exclusions: Children with complex chronic conditions or severe disease evident in the first 4 hours of ED care.
  • Primary Outcome: A three-tiered pneumonia severity classification (mild, moderate, severe) based on clinical interventions and outcomes within 7 days.
  • Analytic Approach: Logistic regression with bootstrap validation, converted into a user-friendly point-based risk score.

Key Findings

Out of 1901 children with complete data:

  • 53.2% had mild CAP
  • 40.6% had moderate CAP
  • 6.2% had severe CAP

Eight clinical variables independently predicted moderate or severe CAP:

Predictor Adjusted Odds Ratio Point Value
Chest retractions 2.86 +3
Oxygen saturation <90% 13.39 +6
Oxygen saturation 90–92% 3.24 +3
Respiratory rate >95th %ile 1.63 +1
Heart rate >95th %ile 1.64 +1
Refusal to drink 1.57 +1
Abdominal pain 1.52 +1
Antibiotics before ED visit 1.64 +1
Rhinorrhea/congestion 0.59 −1

Model performance was excellent:

  • c-statistic = 0.82, both in the general cohort and in those with radiographic CAP.
  • Point thresholds performed well for both sensitivity (92.6% at score ≥1) and specificity (92.0% at score ≥6).

Why Does This Matter?

This study’s risk models:

  • Were built using objective and widely available data (no labs required).
  • Distinguish mild from moderate/severe disease to inform ED disposition decisions.
  • Are pragmatic enough to integrate into clinical decision support tools (apps, EHRs).
  • May outperform clinician gestalt (model c-statistic 0.82 vs clinician 0.75 in previous studies).

Importantly, these models also showed consistent performance in children with radiographic pneumonia, with additional predictors such as multifocal opacities and decreased breath sounds.

Bottom Line

This PERN study delivers a robust, externally relevant tool for risk stratifying pediatric CAP in the ED. The scoring system can help avoid unnecessary hospitalizations and ensure that higher-risk children get the care they need sooner. While external validation is still needed, the groundwork is set for incorporating this evidence into practice.

Reference

Florin TA, Tancredi DJ, Ambroggio L, et al. Predicting paediatric pneumonia severity in the emergency department: a multinational prospective cohort study of the Pediatric Emergency Research Network. Lancet Child Adolesc Health. 2025;9(6):383–92. DOI: 10.1016/S2352-4642(25)00094-X