In IEEE journal of biomedical and health informatics
Each year there are nearly 57 million deaths worldwide, with over 2.7 million in the United States. Timely, accurate and complete death reporting is critical in public health, especially during the COVID-19 pandemic, as institutions and government agencies rely on death reports to analyze vital statistics to formulate responses to communicable diseases. Unfortunately, determining the true causes of death is challenging even for experienced physicians. The novel coronavirus and its variants may further complicate the task, as physicians and researchers are still investigating COVID-related complications. To facilitate physicians in accurately reporting causes of death, an advanced artificial intelligence (AI) approach is presented to determine a chronically ordered sequence of clinical conditions that lead to death (named as the causal sequence of death), based on decedents last hospital discharge record. The key technical issue of this problem is to learn the causal relationship between clinical codes and to identify death-related conditions. Specifically, three challenges in determining the causal sequence of death are identified: multiple clinical coding system versions, medical domain knowledge constraint, and data interoperability. To overcome the first challenge, the advanced neural machine translation models with various attention mechanisms are applied to generate target sequences. The BLEU (BiLingual Evaluation Understudy) score is used along with three accuracy metrics to evaluate the quality of generated sequences. We achieve state-of-art results. To address the second challenge, expert-verified medical domain knowledge is incorporated as constraints during cause of death sequence generation. Lastly, a Fast Healthcare Interoperability Resources (FHIR) interface demonstrates the usability of this work in clinical practice. During this ongoing pandemic, this work can potentially benefit physicians in understanding comorbidities contributing to coronavirus morbidity and mortality.
Zhu Yuanda, Sha Ying, Wu Hang, Li Mai, Hoffman Ryan, Wang May Dongmei