A Novel Modeling Focused Missing Data Imputation Framework

Missing data is a pervasive challenge in critical fields such as healthcare, often arising due to machinery failures, human error, or incomplete data collection. Integrating established imputation techniques such as MICE, iterative imputation, and KNN imputation, we propose a novel framework that combines the strengths of distinct imputation methods to enhance downstream model performance. Trained on several years of NHANES survey data, an ensemble of neural networks predicts biological age using common bioindicators found in blood & urine. Our results demonstrate that this integrated approach improves age prediction accuracy, providing valuable insights into the importance of imputation strategies in age prediction research.