Electronic medical record-based causal network modeling for acute myocardial infarction diagnosis in the emergency department.

For the acute myocardial infarction (AMI) diagnosis in the emergency department, the atypical manifestations and limited information lead to clinical challenge. Data-driven methods often fail in the generalizability against the atypical and limited information. In this work, the causality of AMI is studied based on electronic medical record (EMR), and a framework to construct causal network for AMI diagnosis is proposed. The EMRs with seven categories and 6,001 samples are included. Score-based algorithm, structural equation model, and network coarse-graining are adopted to build causal network with medical knowledge. A model validation procedure is proposed to test the model performance when only part of variable information is obtained. Compared with data-driven methods, causal network achieves best comprehensive performance. Further, the causal effects between variables and AMI can be quantified, which are verified by the sensitivity analysis on unobserved confounders. Such results can support the disease diagnosis, treatment, and healthcare in clinic.