Deep Interpretable Learning System for In Hospital Clinical Deterioration Detection
Author: Trong-Nghia Nguyen
Abstract: Early recognition of clinical deterioration is one of the key steps to reducing inpatient morbidity and mortality. In recent years, the early warning system (EWS) was considered the most effective method for this problem. It has been widely integrated and applied in hospital systems. Although the traditional early warning systems have been widely applied, they still contain many drawbacks, such as the high false warning rate and low sensitivity. This paper proposed a strategy that involves a deep learning approach based on a novel interpretable deep tabular data learning architecture, TabNet, for an early warning system that achieves better performance than other machine learning methods. This study has been processed and validated on a set of data collected from two hospitals of Chonnam National University Hospital (CNUH), Korea, in over ten years. Beside, we also experiment on Brazilian hospitals data of Rob Laura system for comparison. The learning metrics used for the experiment are the area under the receiver operating characteristic curve score (AUROC) and the area under the precision-recall curve score (AUPRC). The experiment on a large real-time dataset shows that our method improves compared to other machine learning-based approaches.
Keywords: Early Warning System, Predictive medicine, Machine Learning, Explainable AI, Healthcare, Vital Signs
Full Text, BibText (soon)
A publicly available sample of one of the biggest Cancer Hospitals in southern Brazil with 13,652 attendances previously approved by the Erasto Gaertner Hospital Research Ethics Committee - n 99706718.9.0000.0098