A Prediction Model for Pediatric Radiographic Pneumonia

Area of Science:

• Pediatric Pulmonology
• Diagnostic Imaging
• Clinical Decision Support

Background:

• Chest radiographs (CXRs) are standard for diagnosing community-acquired pneumonia (CAP).
• Overutilization of CXRs in pediatric patients raises concerns regarding radiation exposure and healthcare costs.
• There is a need for clinical tools to optimize CXR use in suspected CAP cases.

Purpose of the Study:

• To develop and validate a predictive model for radiographic CAP using clinical features.
• To reduce the reliance on CXRs for CAP diagnosis in children.
• To improve the efficiency of CAP diagnosis in pediatric care.

Main Methods:

• A prospective, single-center study included 1142 pediatric patients (3 months to 18 years) with suspected CAP.
• Penalized multivariable logistic regression and bootstrapped backward selection were used to develop predictive models.
• Model performance was evaluated using receiver operating characteristic (ROC) curves and risk thresholds.

Main Results:

• Radiographic CAP was confirmed in 22.2% of patients.
• Key predictors for CAP included increasing age, prolonged fever, tachypnea, and focal decreased breath sounds.
• A reduced model incorporating age, fever duration, and decreased breath sounds achieved an area under the ROC curve of 0.80.
• The model demonstrated high sensitivity (94.9%) and specificity (90%) at specific risk thresholds, identifying low-risk children who were unlikely to have CAP.

Conclusions:

• A predictive model utilizing age, fever duration, and decreased breath sounds effectively identifies radiographic CAP in children.
• This model shows excellent discrimination and may aid in clinical decision-making regarding CXR and antibiotic use for CAP.
• External validation is recommended to confirm the model's utility in diverse clinical settings.