Objective:

To systematically identify distinct clinical phenotypes of pediatric community-acquired pneumonia (CAP) and develop a prediction model for the Mixed-Infection phenotype using unsupervised machine learning, ultimately aiming to improve treatment outcomes.

Key Findings:

• Three clinical phenotypes identified: Mycoplasma-Dominant (37.7%), Mixed-Infection (28.2%), and High-Inflammation (34.1%). The Mixed-Infection phenotype had the highest proportion of prolonged hospitalization (31.4%), though this difference was not statistically significant (p = 0.117), indicating a trend that warrants further investigation.
• Prediction model based on white blood cell count, lactate dehydrogenase, and procalcitonin showed AUC = 0.917 and accuracy = 91.3%, suggesting strong predictive capability.

Interpretation:

The study reveals distinct pathogen-host interaction patterns in pediatric CAP and provides a tool for early identification of the Mixed-Infection phenotype, which may improve clinical management and patient outcomes.

Limitations:

• Findings are based on a bronchoscopy/BAL-selected cohort, limiting generalizability to all pediatric CAP patients; this selection may skew the understanding of the broader population.
• Further validation in larger prospective cohorts is needed to confirm the model's applicability and address the limitations of the current study.

Conclusion:

The study successfully identifies clinical phenotypes in pediatric CAP and develops a promising early identification tool for Mixed-Infection, warranting further research to establish broader applicability and enhance clinical practice.