Advisory Committee Chair
Charity J Morgan
Advisory Committee Members
Stacey S Cofield
Gary R Cutter
Ellen F Eaton
Byron C Jaeger
John R Rinker
Document Type
Dissertation
Date of Award
2019
Degree Name by School
Doctor of Philosophy (PhD) School of Public Health
Abstract
Modeling disability in multiple sclerosis (MS) is challenging due to its complexity and non-linearity, with utilized methodology having many limitations. Lack of a suitable biomarker has led to relying on statistical models to understand disease progression. Markov methodology has been limitedly applied to the clinically-assessed Expanded Disability Status Score (EDSS), but not to patient-reported Patient Determined Disease Steps (PDDS); both measure disease progression and disability, and neither have been analyzed using the Test of Lumpability (TOL). It is common practice to aggregate these scores for computational or inferential convenience; in the case of Markov Chains (MC), combining states is referred to as lumping. The resulting chain must be evaluated using the TOL to ensure retention of the Markov property. Extending the TOL, we developed a goodness of fit (GOF) test with Pearson and likelihood ratio formulations to compare lumping schemes that pass the TOL; both were shown to follow a Chi-squared distribution. Performance was evaluating using simulated and patient data. Lumping schemes were identified for each disability scale. Using semi-annual surveys from the North American Research Committee on Multiple Sclerosis (NARCOMS), PDDS was predicted using Markov models with and without covariates for multiple lumping schemes; using semi-annual follow-up data from the CombiRx trial, EDSS scores were similarly modeled. Disability scores were lumped with scientifically supported schemes. Schemes were assessed for parsimony, clinical usefulness, and adherence to Markov property (lumpability); covariates were selected with scientific justification. Novel application to NARCOMS PDDS data will benefit from a larger sample size and wider range of disease statuses than are observed in clinical trials, enhancing generalizability; novel evaluation of lumpability to EDSS outcomes will extend current work. Implementation of Markov methodology has the potential to provide fresh insight into MS disease progression.
Recommended Citation
Hartzes, Anastasia, "Applying Markov Processes To Model Disability In Relapsing Multiple Sclerosis Patients" (2019). All ETDs from UAB. 1886.
https://digitalcommons.library.uab.edu/etd-collection/1886
