To the on-site nursing staff or field management in prehospital emergency care, it seems baffling to conduct more targeted checklist tests for a specific disease. To address this problem, we proposed a decision support method for prehospital emergency care based on ranking the importance of physiological variables. We used multiple logistic regression models to explore the effects of various physiological variables on diseases based on the area under the curve (AUC) value. We implemented the method on the intensive care database (i.e., the Medical Information Mart for Intensive Care (MIMIC-III) database) and explored the importance of 17 physiological variables for 24 diseases, both chronic and acute. We included 33,798 adult patients, using the full physiological dataset as experiment data. We ranked the importance of the physiological variables related to the diseases according to the experiments' AUC value. We discussed which physiological variables should be considered more important in adult intensive care units (ICUs) for prehospital emergency care conditions. We also discussed the relationships among the diseases based on ranking the importance of physiological variables. We used large-scale ICU patient data to obtain a cohort of physiological variables related to specific diseases. Ranking a cohort of physiological variables is a cost-effective means of reducing morbidity and mortality under prehospital emergency care conditions.
Keywords: ICU; disease; machine learning; physiological variable.