Examining Effect of Age, Education, and Race on the Montreal Cognitive Assessment and the Mini-Mental State Exam Open Access

Schneider, Brandon (Spring 2018)

Permanent URL: https://etd.library.emory.edu/concern/etds/3484zg90n?locale=en


Introduction: The Mini-Mental State Exam (MMSE) has begun to be replaced by the Montreal Cognitive Assessment (MoCA) to detect the presence of cognitive impairment. As such, there have been multiple attempts to equate the two tests, usually using equipercentile weighting. This method does not adequately control for confounding variables such as race, education, and age.  Therefore, we used regression methods to control for these potential confounding variables.


Methods: We used a set of quantile regression models to evaluate potential nonlinear effects between each of the tests and our three variables of interest: age, education, and race. If there was no nonlinear effect present, then we used parametric methods to equate MoCA and MMSE. First, we proposed a linear model with splines to control for potential ceiling effects in the MMSE. This model does not restrict the predictions to be within scientific boundaries. If there were a large number of scientifically absurd predictions we used a non-linear link function. 


Results: There were 927 subjects, split into training (n=648) and validation (n=279) datasets. Our population was largely white and highly educated. The quantile regression showed different effects at the tails, with larger effect sizes for the MoCA. However, these estimates were not determined to be different from the OLS (SPELL OUT) estimates, so parametric regression was adequate for equating the two tests. Linear Regression produced significant effects for age and race, as well as the two splines. Education was not significant at any level. Predictions resulted in only one scientifically implausible value, and the majority of predictions were within 2 points of the true values.


Discussion: We conclude that there is no substantial nonlinear effect as determined by the quantile regression, and parametric assumptions should be adequate. In addition, race and age are two previously uncontrolled for confounders that are significant. Limitations include low diversity of the subject population, no restrictions on validity of the predictions, and insufficient controlling for socioeconomic status and cerebrovascular disease. 

Table of Contents


Introduction. 1

Methods. 8

Results Section. 11

Discussion. 18

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research field
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files