The Health and Aging Brain Study–Health Disparities (HABS–HD) aims to understand factors impacting brain aging in diverse communities. A critical challenge is missing data, hindering accurate machine learning (ML). Common imputation methods may lead to biased outcomes. Thus, developing a new imputation methodology has become an urgent task for HABS–HD.
We devised a three-step workflow: 1) evaluating missing data; 2) ML-based multiple imputation; and 3) imputation evaluation. Embedded is ML-based multiple imputation (MLMI).
The MLMI excelled, demonstrating superior prediction and maintaining distribution and correlation. This workflow is effective for HABS–HD, robustly handling missing values, especially in Alzheimer's disease models, and is applicable to other disease data analyses. View this paper