Cornell University
Machine Learning Approaches to Predicting Chronic Disease Trajectories from Electronic Health Records
Abstract
dc:description.abstractPopulation aging and increased survivability of many diseases are rapidly escalating the incidence and prevalence of chronic conditions, which now affect approximately 133 million Americans. While treatment options for these diseases continue to expand, the complex, multifactorial nature of chronic illnesses greatly complicates their effective management and treatment. In this thesis, I explore the application of machine learning models to study the progression of chronic diseases using longitudinal electronic health record (EHR) data in two clinical contexts: predicting healing trajectories of chronic wounds and modeling myocardial recovery in heart failure patients. Leveraging real-world data of 261,398 wound care and 14,765 heart failure patients, machine learning models were able to determine wound healing status with 84% accuracy and an AUC of 0.92, as well as forecast myocardial recovery with 70% accuracy and an AUC of 0.83. If integrated in real-world care, insights generated by these models could be able to guide more effective and efficient treatment decisions for chronic disease management.
Degree
thesis:*- Name thesis:degree_name
- M.S., Computer Science
- Level thesis:degree_level
- Master of Science
- Discipline thesis:degree_discipline
- Computer Science
- Grantor
- Cornell University
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Dallmann, Anissa
- Chair dc:contributor.chair
-
- Weinberger, Kilian
- Committee member dc:contributor.committeemember
-
- Kwak, Hojoong
Subjects
dc:subject × 4Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Bibid
- 16575466
- Dc Identifier Other
-
ProQuest Submission ID: 12092
ProQuest Publication ID: 31242959