Discovering novel pneumonia phenotypes in Vietnamese patients using multimodal data and artificial intelligence

Community-acquired pneumonia (CAP) and ventilator-associated pneumonia (VAP) are respectively the most common indications for hospital and intensive care unit (ICU) treatment in low- and middle-income countries (LMICs). Yet most CAP and VAP research comes from high-income countries. Predicting severe CAP requiring ICU respiratory support (high-flow nasal oxygen or mechanical ventilation), and determining who with suspected VAP requires antibiotics, are critical decisions in LMICs. However, decision-making in LMICs is hindered by a lack of validated risk scores and limited knowledge of the underlying disease mechanisms. Over the last 8 years, our team have created artificial intelligence (AI) systems to analyse complex clinical, imaging and continuous physiological data acquired via custom-made platforms developed in Vietnam. In patients with life-threatening infections, we have discovered underlying clinical and inflammatory phenotypes linked to outcome, and pioneered advanced pathogen diagnostics. Our proposal builds on these foundations to create novel AI tools that will identify new phenotypes of CAP and VAP associated with the need for ICU respiratory support and antibiotic therapy respectively in LMICs. Combining multimodal clinical data and advanced pathogen identification with host response immune and inflammatory phenotyping we will also advance mechanistic understanding of these phenotypes that will inform new diagnostic and therapeutic strategies.