Related Experiment Video
Updated: Aug 28, 2025
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.6K
Using dynamic microsimulation to project cognitive function in the elderly population
Yifan Wei1, Hanke Heun-Johnson1, Bryan Tysinger1
1Leonard D. Schaeffer Center for Health Policy and Economics, University of Southern California, Los Angeles, California, United States of America.
Plos One
|September 15, 2022
Summary
A new model accurately predicts cognitive decline and mortality over 10 years, aiding Alzheimer's disease and dementia therapy economic evaluations. This tool helps assess the societal value of new treatments.
Area of Science:
- Gerontology
- Biostatistics
- Health Economics
Background:
- Accurate long-term projection models are crucial for evaluating Alzheimer's disease and other dementias (ADOD) therapies.
- Economic evaluations of ADOD treatments require models that track disease progression across the continuum.
- Nationally representative data is essential for robust ADOD projection models.
Purpose of the Study:
- To develop and validate a dynamic microsimulation model for projecting cognitive function and health outcomes.
- To utilize the Future Elderly Model (FEM) with an ordered probit transition model for predicting cognitive status.
- To assess the predictive accuracy of the FEM TICS27 model for long-term cognitive and mortality outcomes.
Main Methods:
- Developed an ordered probit transition model to predict future TICS27 scores using Health and Retirement Study (HRS) data.
- Integrated the TICS27 prediction model into the Future Elderly Model (FEM) dynamic microsimulation framework.
- Validated the FEM TICS27 model using a five-fold cross-validation approach against observed HRS data (2006-2016).
Main Results:
- The FEM TICS27 model accurately replicated the distribution of cognitive scores and cognitive/mortality status observed in HRS over ten years.
- Achieved an Area Under the Receiver Operating Characteristic (AUROC) curve of 0.904 for predicting dementia or death with dementia within 10 years.
- Demonstrated an AUROC of 0.722 for predicting significant cognitive decline within two years among individuals with mild cognitive impairment.
Conclusions:
- The FEM TICS27 model exhibits strong predictive accuracy for both short-term (2-year) and long-term (10-year) cognitive outcomes.
- The model's unbiased validation provides a reliable platform for assessing the societal burden of cognitive decline.
- This validated model can translate the benefits of innovative therapies into measurable long-term societal value.
