Imputation of missing values for electronic health record laboratory data

Abstract

Laboratory data from Electronic Health Records (EHR) are often used in prediction models where estimation bias and model performance from missingness can be mitigated using imputation methods. We demonstrate the utility of imputation in two real-world EHR-derived cohorts of ischemic stroke from Geisinger and of heart failure from Sutter Health to: (1) characterize the patterns of missingness in laboratory variables; (2) simulate two missing mechanisms, arbitrary and monotone; (3) compare cross-sectional and multi-level multivariate missing imputation algorithms applied to laboratory data; (4) assess whether incorporation of latent information, derived from comorbidity data, can improve the performance of the algorithms. The latter was based on a case study of hemoglobin A1c under a univariate missing imputation framework. Overall, the pattern of missingness in EHR laboratory variables was not at random and was highly associated with patients’ comorbidity data; and the multi-level imputation algorithm showed smaller imputation error than the cross-sectional method.

Predicting mortality among ischemic stroke patients using pathways-derived polygenic risk scores

“…Limitations: (1) Single healthcare system cohort with one ethnic background; (2) Limited sample size (lack of power) for a prediction study in subgroups; and (3) Challenges to the survival analysis using EHR data which are often high-dimensional, censored, have high and not-completely-at-random missingness, and low prevalence for the outcome of interest 20 .…”

“…For the self-reported variables (such as alcohol, smoking) the missing value was replaced by zero. The BMI and systolic and diastolic blood pressures were imputed by MICE 2lpan, an appropriate strategy as we previously demonstrated 20 . No imputation was conducted for NIHSS (missingness at 37.8%) as there is no consensus strategy to impute this variable.…”

Artificial Intelligence: A Shifting Paradigm in Cardio-Cerebrovascular Medicine

“…Currently, the limitations in AI-based models are mostly centered on the lack of sufficient patient representation, balanced cohorts, and biases introduced by cohort definitions or selection of variables, as well as the exclusion of a certain group of patients. Machine learning models pick up biases from the training datasets; therefore, to reach new heights, it is of fundamental importance to increase patient representation and data density and improve data for downstream modeling [ 154 , 155 ]. Finally, in terms of methodologies, both fields are taking advantage of advances in machine learning frameworks and tools.…”

Multimodal biomedical AI

“…Imputation has remarkably improved the statistical power of genome-wide association studies to identify novel genetic risk loci, and is facilitated by large reference datasets with deep genotypic coverage such as 1000 Genomes 149 , the UK10K 150 , the Haplotype reference consortium 151 and, recently, TOPMed 89 . Beyond genomics, imputation has also demonstrated utility for other types of medical data 152 . Different strategies have been suggested to make fewer assumptions.…”