<![CDATA[Early clinical assessment in Emergency Departments (EDs) frequently relies on laboratory indicators that are often unavailable at the time of patient arrival. The White Blood Cell (WBC) count is a key marker for infection, inflammation, and acute physiological stress, yet missing or delayed WBC measurements are common in emergency care workflows. Conventional imputation methods typically replace missing values with single-point estimates, implicitly assuming full reliability and ignoring the uncertainty inherent in the imputation process. This paper investigates the use of multiple imputation strategies to address missing WBC measurements in emergency care data. A multivariate, tree-based imputation approach is employed to generate multiple plausible WBC values for each missing observation, capturing the inherent variability of laboratory uncertainty. Instead of focusing solely on point estimates, this study emphasizes the role of imputation variability as an indicator of confidence in reconstructed WBC values. Experiments are conducted on a synthetic emergency care dataset designed to mimic early ED scenarios, using three multiple imputations (M = 3) to compare single imputation and tree-based multiple imputation in terms of variability and uncertainty representation. Through analytical comparison and illustrative experiments on emergency care data with realistic missingness patterns, the proposed approach demonstrates that multiple imputation provides more informative and robust representations of missing WBC measurements compared to traditional single-imputation techniques. The results highlight how uncertainty-aware WBC reconstruction can better support early emergency assessment, particularly in high-risk and time-sensitive clinical scenarios. By focusing on WBC as a representative and clinically critical laboratory variable, this work underscores the importance of treating missing laboratory data as uncertain rather than deterministic. The proposed perspective offers practical insights for improving the reliability of data-driven decision support systems in emergency medicine and lays the groundwork for future integration with predictive modeling frameworks.]]>
Copyrights © 2026
