RESUMEN
The acute respiratory distress syndrome (ARDS) is a heterogeneous syndrome characterized by distinct etiologies and complicated pathobiological mechanism. It is difficult to discriminate patients with unique biological features or individual response to specific therapy by traditional definition and subgrouping. Unfortunately, there are few clinical evidences supporting effective drug therapy to ARDS. The sub-phenotype or endotype of ARDS is related to potential mechanism of the syndrome, and is critical to personalized treatment of ARDS. An appropriate sub-phenotype of ARDS may be defined by data-driven assessment of the available data including clinical features, biological biomarkers and respiratory parameters of the patients. Latent class analysis or machine learning has potential to establish new sub-phenotype of ARDS stably, which is helpful to guide precision medicine approach.