CMStatistics 2020: Start Registration
View Submission - CMStatistics
B1074
Title: Classification of disease progression via recurrent biomarkers using EM algorithm Authors:  Huijun Jiang - University of North Carolina at Chapel Hill (United States) [presenting]
Quefeng Li - University of North Carolina - Chapel Hill (United States)
Jessica Lin - University of North Carolina at Chapel Hill (United States)
Feng-Chang Lin - University of North Carolina - Chapel Hill (United States)
Abstract: Many infectious diseases have more than one potential cause. The classification of infections from more than one possible cause is critical in effective disease control. Multistate model based on Markov processes is a typical approach to estimating the transition rate between the status of the disease. However, it can perform poorly when the problem of interest is the classification of unknown disease status. We aim to demonstrate that the transition likelihoods of disease biomarkers can be utilized to distinguish relapse from reinfection for malaria infection with high accuracy. A more general model for disease progression can be constructed to allow for additional disease states. We start from a multinomial logit model to estimate the disease transition probabilities and then utilize the transition information of disease biomarkers to provide a more accurate classification result. We apply the Expectation-Maximization (EM) algorithm for the estimation of unknown parameters, including the marginal probabilities of disease status. A comparison to the existing two-stage method shows that our classifier is consistent and has better accuracy, especially when the sample size is small. An application to data from 78 Cambodian P. vivax malaria patients is presented to demonstrate the practical use of our proposed method.