Utilizing Theory of Machine Learning and Signal Processing to Incorporate Physiological Signals, Assessment of Sympathovagal Imbalance, Conventional and Novel Inflammation and Immune Biomarkers to Develop and Validate the Clinical Prediction Models F

Sepsis is a common disorder with high mortality worldwide. Up to date, there have been more than among 178 biomarkers for infection determination and 34 biomarkers for sepsis diagnosis and severity stratification developed, however, none has been proved to own appropriate performance characteristics to be recommenced in daily practice. Therefore, there have been many researchers advocating the necessity to form a “panel of biomarkers”, or the “signature” of sepsis to be utilized for frontline health care providers. With the advance of biomedical engineering, more physiological indicators are now available with continuous and non-invasive methods to monitor the hemodynamic status of critically-ill patients. As indicated by many previous researchers, the monitoring of sympathovagal status such as heart rate variability would detect the deterioration of patients with infectious disease earlier than the conventional vital signs monitoring. However, most of the observational studies for sepsis focus only on either several physiological indicator or biomarkers to early prognosticate the outcome. With the broadly applied theory of machine learning and signal processing, it is time for the large observational study. We will further observe the process of the sepsis and attempt to find the indicators of treatment success, other than the survival rate, septic shock and complications. In this study, we plan to start a four-part serial study to investigate the biosignature of sepsis: 1) several systemic reviews and meta-analyses in order to evaluate the performance of biomarkers; 2) a 12-year retrospective sepsis cohort study to compare the effectiveness the current clinical practice and biomarkers; 3) a prospective nested case-control study of 240 patients to evaluate the potential combination of biomarkers and physiological indicators in different PIRO stage to early diagnose sepsis; and 4) a complete Biobank of specimens with linkage to the comprehensive prospective sepsis cohort.

Project ID:PC10607-0357
External Project ID:MOST106-2314-B182-028