A clinically actionable and explainable real-time risk assessment framework for stroke-associated pneumonia
Lutao Daia, Xin Yangb c, Hao Lid, Xingquan Zhaoe, Lin Linf, Yong Jiangd, Yongjun Wangb c e g h✉, Zixiao Lib c e i j✉, Haipeng Shena k✉
- aFaculty of Business and Economics, The University of Hong Kong, Hong Kong
- bChina National Clinical Research Center for Neurological Diseases, Center for Healthcare Quality and Research, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, PR China
- cNational Center for Healthcare Quality Management in Neurological Diseases, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, PR China
- dChina National Clinical Research Center for Neurological Diseases, Center for Big Data Analytics and Artificial Intelligence, Beijing 100070, PR China
- eVascular Neurology, Department of Neurology, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, PR China
- fInformation Management and Data Center, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, PR China
- gBeijing Key Laboratory of Translational Medicine for Cerebrovascular Disease, Beijing 100070, PR China. Electronic address: yongjunwang111@aliyun.com
- hBeijing Key Laboratory of Translational Medicine for Cerebrovascular Disease, Beijing 100070, PR China
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
The current medical practice is more responsive rather than proactive, despite the widely recognized value of early disease detection, including improving the quality of care and reducing medical costs. One of the cornerstones of early disease detection is clinically actionable predictions, where predictions are expected to be accurate, stable, real-time and interpretable. As an example, we used stroke-associated pneumonia (SAP), setting up a transformer-encoder-based model that analyzes highly heterogeneous electronic health records in real-time. The model was proven accurate and stable on an independent test set. In addition, it issued at least one warning for 98.6 % of SAP patients, and on average, its alerts were ahead of physician diagnoses by 2.71 days. We applied Integrated Gradient to glean the model's reasoning process. Supplementing the risk scores, the model highlighted critical historical events on patients' trajectories, which were shown to have high clinical relevance.
TL;DRA transformer-encoder-based model that analyzes highly heterogeneous electronic health records in real-time and highlighted critical historical events on patients' trajectories was shown to have high clinical relevance and was proven accurate and stable on an independent test set. (Semantic Scholar)