Related Experiment Video
Updated: Jul 10, 2026
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Improving linear modeling of cognitive decline in patients with mild cognitive impairment: comparison of two methods
S J Teipel1, A J Mitchell, H J Möller
1Alzheimer Memorial Center and Geriatric Psychiatry Branch, Dementia and Neuroimaging Section, Department of Psychiatry, Ludwig-Maximilian University, Munich, Germany. stefan.teipel@med.uni-muenchen.de
Background:
High variability of estimates of cognitive decline in patients with Alzheimer's disease (AD) derived from unbalanced longitudinal designs may result as much from the applied statistical model as from true biological variability.
Objective:
To compare the accuracy of two statistical models, serial subtraction score (SSA) and mixed-effects regression analysis (MEM), to estimate rates of cognitive decline in patients with amnestic mild cognitive impairment (MCI), a group at risk for AD.
Methods:
We recorded serial mini mental state examination (MMSE) scores from 78 MCI patients. Additionally, we derived simulated trajectories of cognitive decline with unequally spaced observation intervals. Rates of change were assessed from clinical and simulated data using SSA and MEM models.
Results:
MEM reduced variability of rates of change significantly compared to SSA. In a polynomial model, overall length of observation time explained a significant amount of variance of SSA, but not of MEM estimates. For simulated data, MEM was significantly more accurate in predicting true rates of change compared to SSA (p < 0.001).
Conclusion:
MEM yields more accurate estimates of cognitive decline from unbalanced longitudinal data. Simulation studies may be useful to select the appropriate statistical model for a given set of clinical data.
