Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Doctor of Philosophy

Program

Computer Science

Supervisor

Kamran Sedig

Abstract

The increasing use of electronic platforms in healthcare has resulted in the generation of unprecedented amounts of data in recent years. The amount of data available to clinical researchers, physicians, and healthcare administrators continues to grow, which creates an untapped resource with the ability to improve the healthcare system drastically. Despite the enthusiasm for adopting electronic health records (EHRs), some recent studies have shown that EHR-based systems hardly improve the ability of healthcare providers to make better decisions. One reason for this inefficacy is that these systems do not allow for human-data interaction in a manner that fits and supports the needs of healthcare providers. Another reason is the information overload, which makes healthcare providers often misunderstand, misinterpret, ignore, or overlook vital data. The emergence of a type of computational system known as visual analytics (VA), has the potential to reduce the complexity of EHR data by combining advanced analytics techniques with interactive visualizations to analyze, synthesize, and facilitate high-level activities while allowing users to get more involved in a discourse with the data. The purpose of this research is to demonstrate the use of sophisticated visual analytics systems to solve various EHR-related research problems. This dissertation includes a framework by which we identify gaps in existing EHR-based systems and conceptualize the data-driven activities and tasks of our proposed systems. Two novel VA systems (VISA_M3R3 and VALENCIA) and two studies are designed to bridge the gaps. VISA_M3R3 incorporates multiple regression, frequent itemset mining, and interactive visualization to assist users in the identification of nephrotoxic medications. Another proposed system, VALENCIA, brings a wide range of dimension reduction and cluster analysis techniques to analyze high-dimensional EHRs, integrate them seamlessly, and make them accessible through interactive visualizations. The studies are conducted to develop prediction models to classify patients who are at risk of developing acute kidney injury (AKI) and identify AKI-associated medication and medication combinations using EHRs. Through healthcare administrative datasets stored at the ICES-KDT (Kidney Dialysis and Transplantation program), London, Ontario, we have demonstrated how our proposed systems and prediction models can be used to solve real-world problems.

Summary for Lay Audience

Advances in healthcare technology have resulted in the generation of large amounts of electronic data in the form of electronic health records (EHRs). Adoption of EHR makes it easy to organize, access, and store medical records through computerized data management tools. Despite the potential benefits, healthcare professionals continue to report difficulty in adopting EHR-based systems. One of the main reasons for this problem is the complicated and improperly designed user interfaces in these systems, which often makes healthcare providers overlook vital information. The purpose of this research is to prove the use of visual analytics (VA) to solve various EHR-related problems. VA combines automated analysis with interactive visualizations for effective reasoning, understanding and decision making based on complex data. Through a literature survey and proposed framework, we first analyze the existing EHR-based systems and understand why they fail to fulfill the computational demand of EHRs. Two novel VA systems (VISA_M3R3 and VALENCIA) and two studies are designed to demonstrate how the VA approach can be used to overcome the challenges of EHRs. VISA_M3R3 is designed to assist healthcare providers in the identification of medications that may associate with a higher risk of developing acute kidney injury (AKI). VALENCIA provides users with the ability to explore high-dimensional EHRs using a number of dimension reduction and cluster analysis algorithms. The studies are conducted to identify AKI-associated medication and medication combinations and predict the risk of developing AKI using EHRs. Through healthcare administrative datasets stored at the ICES-KDT (Kidney Dialysis and Transplantation program), we have shown how our proposed approach can be used to solve real-world problems.

Download

Share

COinS