Advancing severity grading for risk-differentiated paediatric pneumonia care

Globally, community-acquired pneumonia (CAP) remains the leading cause of morbidity and mortality among young children, accounting for 20% of all deaths among children aged 1–59 months.¹ Nevertheless, many children with CAP might not require hospitalisation and could be safely managed at home.² Without objective risk stratification tools, clinicians often rely on their clinical impression to assess illness severity, which has demonstrated only fair accuracy.³ Yet, widely validated risk stratification tools that enable clinicians to accurately classify the initial severity of any given child with CAP, or help predict who might later develop severe disease or require hospitalisation, are lacking.⁴

Precise risk stratification for paediatric CAP at the time of presentation has the potential to improve unstructured clinical decision making, to facilitate the direction of resources to children most at risk for severe outcomes, to avoid unnecessary treatments among those deemed low risk for severe disease, and to provide reassurance to clinicians who discharge low-risk children. In this context, the study by Todd A Florin and colleagues in The Lancet Child & Adolescent Health⁵ is an important step toward a practical risk stratification tool that could help clinicians in high-income countries differentiate children with CAP in the emergency department (ED) who might require hospitalisation from those who could be safely managed at home.

Florin and colleagues⁵ conducted a prospective observational cohort study in 73 EDs in 14 countries to develop two risk stratification models: one model to identify children with mild CAP who, between 4 h and 7 days after the ED encounter, most likely would not require other care than that provided in the inpatient setting; the second model to identify children who, during this same timeframe, were at greater risk of developing severe disease and needing hospital-based care (eg, the presence of an effusion requiring drainage, respiratory failure, or intensive care unit admission). Although the investigators did not meet their target enrolment of 2500 participants due to COVID-19-related enrolment challenges, their population included 2222 children aged 3 months to <14 years with either clinically or radiographically diagnosed CAP. Their rigorous methodology allowed them to develop scores with good discriminatory value using no more than ten features easily evaluable by clinicians during ED encounters.

Several of the patient-level features identified in this study overlap with features (eg, tachypnoea, markers of respiratory distress, and hypoxaemia) identified in other similar studies conducted in single countries or at single sites in high-income countries.⁴ Moreover, these features have been predictive of CAP-related mortality among hospitalised children in multinational studies conducted in low-income and middle-income countries (LMICs).^6,7 This overlap in features suggests that these could be true biological phenomena that predict severe outcomes among children with CAP and not just a function of the datasets from which these scores were developed. Nevertheless, some of the features included in this study rely on clinician assessment, which introduces subjectivity as inter-rater reliability of features like auscultatory findings might be suboptimal.⁸

Moreover, the presence of some of the included features might be a proxy for the outcome. Specifically, hypoxaemia was the feature most predictive of the requirement for hospitalisation, and the presence of hypoxaemia would automatically trigger administration of supplemental oxygen and hospitalisation in many settings included in this study, thus negating any predictive ability of this variable to identify an event that occurs later. This is not to downplay the importance of hypoxaemia, which is not routinely assessed in many LMIC settings,⁹ as it is an objective measure that has been predictive of severe outcomes in multiple studies.^6,7 Future studies to enhance CAP prognostics could integrate other objective measures such as biomarkers, which have demonstrated excellent prognostic capacity for CAP in children.¹⁰ Additionally, future studies could include follow-up periods beyond 7 days as complications of CAP can present beyond the initial timeframe.

There are now at least five risk stratification models for moderate and severe CAP among children.⁴ The Pediatric Emergency Research Network (PERN) risk stratification models are the most inclusive given the heterogeneous settings from which they were derived. We applaud Florin and colleagues and encourage future assessment of these models (with and without the additional measurement of objective prognostic biomarkers) in acute care settings, particularly in LMICs, to move toward more precise, risk-differentiated care for children with CAP where morbidity is the highest. Additionally, assessing the accuracy of the PERN models for paediatric CAP in settings other than the ED could provide additional evidence of their generalisability. Moving forward, the PERN risk stratification models should be externally validated before their implementation, but Florin and colleagues’ work is an important step forward in the field of risk differentiation for the evaluation of paediatric CAP.