A Data-Driven Framework for Improving Clinical Managements of Severe Paralytic Ileus in ICU: From Path Discovery, Model Generation to Validation
摘要
Paralytic ileus (PI) is a severe health condition associated with poor clinical outcomes and longer hospital stays. Due to the high variability in clinical pathways, identifying risk factors on high-frequency pathways may facilitate the efficient optimization of clinical processes. This paper illustrated a data-driven framework that combines local process optimization and conceptual model validation. Frequent clinic pathways and contributing factors were discovered by leveraging local process modelling (LPM) and Partial Least Squares-based Structural Equation Modeling (PLS-SEM). Principle component analysis (PCA) was used to identify latent factors. LPM was used to identify structural relationships in the high-frequent process pathways. PLS-SEM was adopted to evaluate the magnitude of relations. Through this framework, the study identified one frequent clinic pathway and six contributing factors for severe PI patients.