Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.
Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Acute Coronary Syndrome Prediction: A Data-Driven Machine Learning Modeling Approach In Emergency Care
Date of Award
Industrial and Manufacturing Engineering
Leslie F. Monplaisir
Celestine C. Aguwa
Healthcare facilities are faced with significant challenges all year round, with patients presenting to the emergency department (ED) with different health issues. Of these challenges, heart disease seems to be an outlier. With heart disease being the primary cause of mortality and morbidity in both developed and developing countries, clinical concerns for acute coronary syndrome (ACS) are one of emergency medicine’s most common patient encounters. Of the three sub-categories of ACS, non-ST-segment elevation myocardial infarction (NSTEMI) has a long-term impact on the well-being of patients if left untreated. Previous efforts in hospital management have applied machine learning algorithms in differentiating NSTEMI from unstable angina (UA), another sub-category of ACS.Nonetheless, currently, we are faced with people who present to the ED with other clinical concerns for ACS but do not necessarily have ACS. No research has been done to differentiate NSTEMI from UA and other non-ACS etiologies. This research work aims to develop models that aid in answering these research questions: (a) How can we effectively develop a decision support tool to classify NSTEMI patients with clinical concerns for ACS? (b) What effect do the clinical narratives have on classifying NSTEMI patients with clinical concerns for ACS? (c) How can we model to minimize the length of stay for NSTEMI patients in the ED subject to controllable parameters? We propose an ensemble learning-driven simulation-optimization framework to help in addressing these questions following three phases. Phase one – developing a unique multi-class classification algorithm to identify risk factors that allow physicians to rule out NSTEMI patients and effectively classify them with the proper evaluation metrics. Phase two – building a multi-class classification framework that incorporates clinical narratives from the physicians’ comments and the clinical data to see whether it enhances the classification power of our model. Phase three – developing a simulation-optimization framework to investigate how resource allocation affects the ED’s performance for NSTEMI patients. The expected outcomes of the study are the combination of risk factors to help physicians rule out NSTEMI patients, thereby enhancing the performance of the ED and studying the interactions between the different ED resources to improve the quality of service for NSTEMI patients. Our proposed framework will assist healthcare administrators, and physicians plan effectively to address issues with NSTEMI patients instead of just sticking to the status quo.
Emakhu, Joshua Oluwatobiloba, "Acute Coronary Syndrome Prediction: A Data-Driven Machine Learning Modeling Approach In Emergency Care" (2022). Wayne State University Dissertations. 3493.