Dynamic lifetime risk prediction of Alzheimer's disease with longitudinal cognitive assessment measurements

Abstract

Introduction: The progressive nature of Alzheimer's disease (AD) highlights the importance of predicting lifetime risk and updating assessments as new data emerge. This study aimed to develop a dynamic model using longitudinal cognitive assessments for updated risk predictions.

Methods: This study used data from the Religious Orders Study and the Rush Memory and Aging Project (ROSMAP) to develop a dynamic risk prediction model based on five cognitive domains, updated annually over 10 years.

Results: The lifetime prediction models based on 2384 participants showed improved area under the curve (AUC) over time, rising from 0.578 at baseline to 0.765 with 10 years of data. The models predicting AD onset before ages 85 and 90 showed superior performance, with AUCs increasing from 0.761 to 0.932 and 0.658 to 0.876, respectively.

Discussion: Incorporating longitudinal cognitive assessments improves AD risk prediction as more data become available. Future research should integrate diverse data types to further boost predictive accuracy.

Highlights: Developed a dynamic lifetime risk prediction model. The area under the curve (AUC) increased from 0.578 at baseline to 0.765 with 10 years of data. The models predicting pre-85 and pre-90 risks demonstrated superior performance.

Keywords: Alzheimer's disease; cognitive assessment; dynamic risk prediction.