Research Article
Clinical Decision Making and Outcome Prediction for COVID-19 Patients Using Machine Learning
@INPROCEEDINGS{10.1007/978-3-030-99194-4_1, author={Adamopoulou Maria and Velissaris Dimitrios and Michou Ioanna and Matzaroglou Charalampos and Messaris Gerasimos and Koutsojannis Constantinos}, title={Clinical Decision Making and Outcome Prediction for COVID-19 Patients Using Machine Learning}, proceedings={Pervasive Computing Technologies for Healthcare. 15th EAI International Conference, Pervasive Health 2021, Virtual Event, December 6-8, 2021, Proceedings}, proceedings_a={PERVASIVEHEALTH}, year={2022}, month={3}, keywords={COVID-19 Clinical decision making Machine learning Patient management}, doi={10.1007/978-3-030-99194-4_1} }- Adamopoulou Maria
Velissaris Dimitrios
Michou Ioanna
Matzaroglou Charalampos
Messaris Gerasimos
Koutsojannis Constantinos
Year: 2022
Clinical Decision Making and Outcome Prediction for COVID-19 Patients Using Machine Learning
PERVASIVEHEALTH
Springer
DOI: 10.1007/978-3-030-99194-4_1
Abstract
In this paper, we present the application of a Machine Learning (ML) approach that generates predictions to support healthcare professionals to identify the outcome of patients through optimization of treatment strategies. Based on Decision Tree algorithms, our approach has been trained and tested by analyzing the severity and the outcomes of 346 COVID-19 patients, treated through the first two pandemics “waves” in a tertiary center in Western Greece. Its’ performance was achieved, analyzing entry features, as demographic characteristics, comorbidity details, imaging analysis, blood values, and essential hospitalization details, like patient transfers to Intensive Care Unit (ICU), medications, and manifestation responses at each treatment stage. Furthermore, it has provided a total high prediction performance (97%) and translated the ML analysis to clinical managing decisions and suggestions for healthcare institution performance and other epidemiological or postmortem approaches. Consequently, healthcare decisions could be more accurately figured and predicted, towards better management of the fast-growing patient subpopulations, giving more time for the effective pharmaceutical or vaccine armamentarium that the medical, scientific community will produce.
