Predicting Conversion Time from Mild Cognitive Impairment to Dementia with Interval-Censored Models

Area of Science:

• Neurology
• Biostatistics
• Data Science

Background:

• Mild cognitive impairment (MCI) patients face a high risk (over 10% annually) of progressing to Alzheimer's disease and related dementias (ADRD).
• Accurate prediction of MCI-to-dementia conversion time is vital for clinical management and patient care.

Purpose of the Study:

• To develop and validate predictive models for accurately estimating the time from MCI to dementia.
• To identify key clinical measures for predicting conversion using easily available data.

Main Methods:

• Utilized semi-parametric and random forest models designed for interval-censored data.
• Employed a variable selection approach to identify important predictive measures.
• Validated models using two large Alzheimer's disease (AD) cohort datasets.

Main Results:

• The semi-parametric model demonstrated improved prediction accuracy for MCI-to-dementia conversion time.
• The model showed good predictive performance across all patient groups.
• Variable selection effectively identified crucial predictors for conversion.

Conclusions:

• Analyzing interval-censored data with appropriate models enhances prediction performance.
• The developed models offer a valuable tool for predicting dementia conversion in MCI patients.
• Accurate prediction facilitates timely interventions and personalized treatment strategies.