EUVIMEDSkopri EUVIMED
Open access

Adjusting for Selection Bias Due to Missing Data in Electronic Health Records‐Based Research by Blending Multiple Imputation and Inverse Probability Weighting

Tanayott Thaweethai, Rajarshi Mukherjee, David Arterburn, Heidi Fischer, Catherine Lee, Susan M Shortreed, Sebastien Haneuse iD

Statistics in Medicine · 2025

DOI: 10.1002/sim.70151 ↗

At a glance

Journal article in Statistics in Medicine, 2025. Written by Tanayott Thaweethai and 6 co-authors. Institution: Massachusetts General Hospital Biostatistics, Boston, Massachusetts, USA. Major MEDLINE topics: Electronic Health Records. Cited 2 times so far (Crossref); 64 references. The full text is freely available in PubMed Central. Funded by: NIDDK NIH HHS, NIH HHS, National Institutes of Health.

AstrattSubjectsDettalji biblijografiċiCiteRelated studies

Abstract (original language)

Due to the complex process by which electronic health records (EHR) are generated and collected, missing data is a significant challenge when conducting large observational studies using such data. However, most standard methods that seek to adjust for the potential selection bias induced by restricting to individuals with complete data fail to address the heterogeneous structure of EHR. To address this, a framework was previously proposed that modularizes the data provenance that gives rise to the observed data as a sequence of decisions made by patients, healthcare providers, and the health system. In this work, we formalize analyses with this framework, specifically by proposing a pragmatic, flexible, and scalable framework for estimation and inference that blends inverse probability weighting and multiple imputation. The proposed framework allows better alignment between consideration of missingness assumptions to the complexity of EHR data. In addition to formal theoretical justification and simulation studies, we illustrate the proposed framework with a motivating data application in which EHR data are used to investigate weight loss outcomes following bariatric surgery, and whether differences between two surgery types exhibit effect modification by presence/absence of chronic kidney disease.

Subjects

MeSH terms (MEDLINE)

Electronic Health RecordsHumansSelection BiasComputer SimulationModels, StatisticalProbabilityData Interpretation, StatisticalRenal Insufficiency, ChronicWeight Loss

Author keywords

electronic health recordsinverse probability weightingmissing datamultiple imputationreal‐world dataselection bias

Authors and institutions

  • Massachusetts General Hospital Biostatistics, Boston, Massachusetts, USA.
  • Department of Medicine, Harvard Medical School, Boston, Massachusetts, USA.
  • Department of Biostatistics, Harvard T. H. Chan School of Public Health, Boston, Massachusetts, USA.
  • Kaiser Permanente Washington Health Research Institute, Seattle, Washington, USA.
  • Department of Medicine, University of Washington, Seattle, Washington, USA.
  • Department of Research & Evaluation, Kaiser Permanente Southern California, Pasadena, California, USA.

ORCID profiles: Sebastien Haneuse

Funding

  • NIDDK NIH HHS
  • NIH HHS
  • National Institutes of Health

Cite

APA

Thaweethai, T., Mukherjee, R., Arterburn, D., Fischer, H., Lee, C., Shortreed, S. M., & Haneuse, S. (2025). Adjusting for Selection Bias Due to Missing Data in Electronic Health Records‐Based Research by Blending Multiple Imputation and Inverse Probability Weighting. Statistics in Medicine. https://doi.org/10.1002/sim.70151

Vancouver

Thaweethai T, Mukherjee R, Arterburn D, Fischer H, Lee C, Shortreed SM, et al. Adjusting for Selection Bias Due to Missing Data in Electronic Health Records‐Based Research by Blending Multiple Imputation and Inverse Probability Weighting. Statistics in Medicine. 2025. doi:10.1002/sim.70151

BibTeX
@article{Thaweethai2025,
  author = {Thaweethai, Tanayott and Mukherjee, Rajarshi and Arterburn, David and Fischer, Heidi and Lee, Catherine and Shortreed, Susan M and Haneuse, Sebastien},
  title = {Adjusting for Selection Bias Due to Missing Data in Electronic Health Records‐Based Research by Blending Multiple Imputation and Inverse Probability Weighting},
  journal = {Statistics in Medicine},
  year = {2025},
  doi = {10.1002/sim.70151}
}
Start for free →