Using Markov Model to Assess Transition Probabilities of Relapse and Remission for Epilepsy Patients

APHA Scientific Session and Event Listing

Jianxin Lin, Biostatistics, University of Medicine & Dentistry of New Jersey, 1634 Raspberry Ct., Edison, NJ 08817, 848-248-6562, linji@umdnj.edu and Anne Berg, Biology Department, Northern Illinois University, Biology Department, Dekalb, IL 60115.

Abstract In some clinical studies of chronic diseases, a patient's prognosis depends on the occurrence of intermediate states. Traditional statistical methods for modeling data from these studies have involved estimation of transition intensities. However, clinicians are not particularly interested in the estimates of these transition intensities. Rather they need estimates of the predictive probabilities of a patient's future progression and their ultimate outcome given their current history. The Markov chain model has proved to be useful in this situation. In this project we develop a nonhomogeneous Markov model for analyzing seizure data. Over many years, repeated remissions and relapses may occur. These are difficult to quantify with standard survival techniques used in analysis of remission and relapse. The Markov model allows one to track a patient's state over time, provides a suitable approach for studying repeated remissions and relapses, which is to allow the calculation of net effects by combining transition intensities into transition probabilities, and predicting the final outcome based on patient's current status. Keyword: Markov model, transition probability, epilepsy.

Learning Objectives:

Using Markov Model to Assess Transition Probabilities of

Relapse and Remission for Epilepsy Patients

Jianxin Lin*, MS and Anne Berg**, PhD

Abstract

In some clinical studies of chronic diseases, a patient’s prognosis depends on the occurrence of intermediate states. Traditional statistical methods for modeling data from these studies have involved estimation of transition intensities. However, clinicians are not particularly interested in the estimates of these transition intensities. Rather they need estimates of the predictive probabilities of a patient’s future progression and their ultimate outcome given their current history. The Markov chain model has proved to be useful in this situation. In this project we develop a nonhomogeneous Markov model for analyzing seizure data. Over many years, repeated remissions and relapses may occur. These are difficult to quantify with standard survival techniques used in analysis of remission and relapse. The Markov model allows one to track a patient’s state over time, provides a suitable approach for studying repeated remissions and relapses, which is to allow the calculation of net effects by combining transition intensities into transition probabilities, and predicting the final outcome based on patient’s current status.

Keyword

* Biostatistics Department, University of Medicine & Dentistry of New Jersey
Piscataway, NJ, 08854
** Biology Department, Northern Illinois University
DeKalb, IL, 60115.

Presenting author's disclosure statement:

Any relevant financial relationships? No

Statistical Methodology

The 134th Annual Meeting & Exposition (November 4-8, 2006) of APHA