Objective: to predict severe pneumonia outcomes in children.
Design: prospective observational study for the purpose of developing a prognostic model for clinical practice.
Setting: hospital-based study in three paediatric hospitals located in Memphis, Nashville and Salt Lake City (USA).
Study population: patients aged less than 18 years admitted to hospital with a diagnosis of community-acquired pneumonia between January 2010 and June 2012. The inclusion criteria were admission to hospital with signs or symptoms of acute infection, acute respiratory illness and radiographic evidence of pneumonia. Children with a recent hospitalisation, severe immunosuppression, cystic fibrosis, tracheostomy or a clear alternative diagnosis were excluded.
Risk factor assessment: based on the most severe outcome that occurred during their hospital stay, children were classified as having a severe (children that died, required invasive mechanical ventilation or developed shock requiring vasoactive medications), moderate (children admitted to the intensive care unit but who did not meet the severity criteria) or mild prognosis (all others).
Outcome measurement: the authors selected twenty predictor variables that included demographic, clinical, radiologic, laboratory and comorbidity variables for the purpose of developing three prognostic models. The first one included all 20 variables. For the second model, the authors selected 14 variables that were mainly clinical and assessed on a scale of 1 to 5, and those with a median score of 4 or higher and which were considered important or very important were ultimately selected, resulting in a reduced model with 10 variables. The third model included nine variables that are documented routinely at the time of admission (age, sex, ethnicity, vital signs and white blood cell count). Aetiologic assessments included blood culture, serology for eight respiratory viruses, pneumococcal and group A streptococcal polymerase chain reaction (PCR), and naso-oropharyngeal swabs for PCR for 13 respiratory viruses, Mycoplasma pneumoniae and Chlamydophila pneumoniae.
Main results: Twenty-one percent of the 2319 children included in the study had a severe or moderate prognosis. The three assessed models correctly predicted the risk of moderate or severe pneumonia. The concordance index was 0.81 (95% confidence interval [95 CI], 0.79 to 0.83) for the first model, 0.79 (95 CI, 0.77 to 0.81) for the second, and 0.78 (95 CI, 0.76 to 0.80) for the third. The variables associated most strongly with severe prognosis were age (the lower the age, the poorer the prognosis), altered mental status, breathing difficulty, abnormal vital signs (pulse oximetry, body temperature, systolic blood pressure) and presence of a radiologic pattern of multilobar or interstitial infiltration. The reduced models with ten and nine variables included the main predictors of severity. Cases of pneumonia by Streptococcus pneumoniae, Staphylococcus aureus and Streptoccoccus pyogenes were associated more frequently with moderate or severe pneumonia.
Conclusion: the authors present three prognostic models for the accurate estimation of risk of severe pneumonia in children, which would allow improving care and outcomes in this population.
Conflicts of interest: none noted.
Funding source: National Institutes of Health, Agency for Healthcare Research and Quality, National Center for Advancing Translational Sciences, National Center for Immunizations and Respiratory Diseases at the CDC.