Comparison of Imputation Strategies for Incomplete Electronic Health Data
4 affiliations
- 1Department of Computer Science, Tianjin University of Technology, Tianjin, China .
- 2National Institute of Health Data Science, Peking University, Beijing, China .
- 3DCST, BNRist, RIIT, Institute of Internet Industry, Tsinghua University, Beijing, China .
- 4Department of Clinical Epidemiology and Biostatistics, Peking University People’s Hospital, Beijing, China .
Abstract
Background: Missing data are a common challenge in clinical tabular datasets and can substantially affect both statistical analyses and predictive modeling in healthcare. Methods: In this study, we conducted an empirical evaluation of 8 imputation strategies—median imputation, median imputation with missing indicator, multiple imputation by chained equations (MICE), MissForest, generative adversarial imputation network (GAIN), missing data imputation using autoencoders (MIWAE), self-attention-based imputation for time series, and gated recurrent unit with decay—across 3 heterogeneous real-world clinical datasets. The methods were assessed under controlled missingness simulations covering different mechanisms (missing completely at random, missing at random [MAR], and missing not at random [MNAR]) and missingness rates ranging from 10% to 90%. We evaluated imputation quality using multiple statistical metrics and examined its relationship with downstream classification performance across multiple classifiers. Results: Across the considered settings, MICE and MissForest showed consistently strong and stable performance, while deep generative approaches such as GAIN and MIWAE exhibited higher variability, particularly under the MAR and MNAR mechanisms and at higher missingness rates. Importantly, we observed that improvements in imputation accuracy did not necessarily translate into better downstream predictive performance. Conclusion: Rather than drawing general conclusions about method superiority, the results highlight the sensitivity of imputation outcomes to data characteristics, missingness mechanisms, and evaluation objectives. These findings underscore the importance of aligning imputation strategy selection with the intended downstream task and data regime when working with heterogeneous clinical tabular data.
Related
- D3MI: an efficient and powerful federated imputation method for bias reduction in the analysis of distributed incomplete data by accounting for within-site correlation and between-site heterogeneity.
- Comparison of parametric versus machine-learning multiple imputation in clinical trials with missing continuous outcomes.
- How to Train Your Intercept: A Comparison of Intercept Update Strategies in Coordinate Descent
- Portfolio Diversification Revisited: An Out-of-Sample Comparison of Twenty-One Strategies Across Asset Universes and Rebalancing Frequencies
- Enhancing Spectral Embedding through Robust and Flexible Knowledge Transfer in Electronic Health Records
BibTeX
@article{zhang2026comparison,
title = {{Comparison of Imputation Strategies for Incomplete Electronic Health Data}},
author = {Shuo Zhang and Zhilong Zhang and Yuxi Zhou and Huixin Liu},
journal = {Health data science},
year = {2026},
month = nov,
doi = {10.34133/hds.0493},
url = {https://europepmc.org/article/PMC/PMC13724834},
note = {CC BY},
}
RIS
TY - JOUR
TI - Comparison of Imputation Strategies for Incomplete Electronic Health Data
AU - Shuo Zhang
AU - Zhilong Zhang
AU - Yuxi Zhou
AU - Huixin Liu
PY - 2026
DA - 2026/11/02
JO - Health data science
DO - 10.34133/hds.0493
UR - https://europepmc.org/article/PMC/PMC13724834
AB - Background: Missing data are a common challenge in clinical tabular datasets and can substantially affect both statistical analyses and predictive modeling in healthcare. Methods: In this study, we conducted an empirical evaluation of 8 imputation strategies—median imputation, median imputation with missing indicator, multiple imputation by chained equations (MICE), MissForest, generative adversarial imputation network (GAIN), missing data imputation using autoencoders (MIWAE), self-attention-based imputation for time series, and gated recurrent unit with decay—across 3 heterogeneous real-world clinical datasets. The methods were assessed under controlled missingness simulations covering different mechanisms (missing completely at random, missing at random [MAR], and missing not at random [MNAR]) and missingness rates ranging from 10% to 90%. We evaluated imputation quality using multiple statistical metrics and examined its relationship with downstream classification performance across multiple classifiers. Results: Across the considered settings, MICE and MissForest showed consistently strong and stable performance, while deep generative approaches such as GAIN and MIWAE exhibited higher variability, particularly under the MAR and MNAR mechanisms and at higher missingness rates. Importantly, we observed that improvements in imputation accuracy did not necessarily translate into better downstream predictive performance. Conclusion: Rather than drawing general conclusions about method superiority, the results highlight the sensitivity of imputation outcomes to data characteristics, missingness mechanisms, and evaluation objectives. These findings underscore the importance of aligning imputation strategy selection with the intended downstream task and data regime when working with heterogeneous clinical tabular data.
SN - 2097-1095
SN - 2765-8783
ER -