Author ORCID Identifier

https://orcid.org/0009-0007-9955-1353

Semester

Summer

Date of Graduation

2023

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Industrial and Managements Systems Engineering

Committee Chair

Imtiaz Ahmed

Committee Member

Abdullah Al-Mamun

Committee Member

Avishek Choudhury

Abstract

Transthyretin Amyloid Cardiomyopathy (ATTR-CM) is a rare, progressive, and fatal disease. Prevalence of ATTR-CM ranges from 4 to 17 per 100000 cases where the mean survival time is less than 4 years. It has a history of being underdiagnosed and misdiagnosed. The diagnosis delay has a weighted mean of 6.1 years for wild-type ATTR-CM. Low awareness, the necessity of invasive procedures, and lack of treatment are the key reasons for delayed diagnosis. But, with the introduction of non-invasive tests like nuclear scintigraphy with 99mTC-PYP and the disease modifying drug Tafamidis, the diagnosis delay signifies a missed opportunity to increase life expectancy by early treatment. Studies show that mean life expectancy can be increased by 5.46 years by early treatment if the 6.1 years of diagnosis delay can be eliminated, whereas the current mean survival time is less than 4 years. Though there is no definitive symptom for it, studies have found out some key prognostic flags: symptoms and comorbidities that are co-existent with ATTR-CM. A prediction model can be developed using the electronic health records (EHR) information in hand to diagnose it early and aid to increase the mean life expectancy. This study aims to identify the top phenotypes that can be used for early diagnosis of ATTR-CM and to predict ATTR-CM using machine learning models among heart failure patients. Patient records from North American healthcare organizations were derived from an EHR system ‘TrixNetX’ for this study. Several statistical analyses (e.g., logistic regression, forward and backward elimination, LASSO, and Survival analysis) were utilized to find out the top diagnostic procedures and comorbidities related with the diagnosis of wild-type ATTR-CM. These key factors were used as features to train machine learning models (e.g., XGBoost, Random Forest) and predict ATTR-CM early among heart failure patients. The study results found the key factors related to diagnosis delay and predicting early cases to improve life expectancy and quality of life.

Download

Share

COinS