A method for comparing multiple imputation techniques: A case study on the U.S. national COVID cohort collaborative

Public Deposited
Analytics

Downloadable Content

Download PDF
Request Version for Screen Reader
Creator
  • Casiraghi E.
    • Other Affiliation: Università degli Studi di Milano
  • Wong R.
    • Other Affiliation: Stony Brook University
  • Hall M.
    • Other Affiliation: Stony Brook University
  • Coleman B.
    • Other Affiliation: University of Connecticut
  • Notaro M.
    • Other Affiliation: Università degli Studi di Milano
  • Evans M.D.
    • Other Affiliation: University of Minnesota
  • Tronieri J.S.
    • Other Affiliation: University of Pennsylvania
  • Blau H.
    • Other Affiliation: The Jackson Laboratory for Genomic Medicine
  • Laraway B.
    • Other Affiliation: University of Colorado
  • Callahan T.J.
    • Other Affiliation: University of Colorado
  • Chan L.E.
    • Other Affiliation: Oregon State University
  • Bramante C.T.
    • Other Affiliation: University of Minnesota
  • Buse J.B.
    • Affiliation: School of Medicine, Department of Medicine
  • Moffitt R.A.
    • Other Affiliation: Stony Brook University
  • Stürmer T.
    • Affiliation: Gillings School of Global Public Health, Department of Epidemiology
  • Johnson S.G.
    • Other Affiliation: University of Minnesota
  • Shao Y.R.
    • Other Affiliation: UT Southwestern Medical Center
  • Reese J.
    • Other Affiliation: Lawrence Berkeley National Laboratory
  • Robinson P.N.
    • Other Affiliation: University of Connecticut
  • Paccanaro A.
    • Other Affiliation: University of London
  • Valentini G.
    • Other Affiliation: Università degli Studi di Milano
  • Huling J.D.
    • Other Affiliation: University of Minnesota
  • Wilkins K.J.
    • Other Affiliation: National Institutes of Health
Abstract
  • Healthcare datasets obtained from Electronic Health Records have proven to be extremely useful for assessing associations between patients’ predictors and outcomes of interest. However, these datasets often suffer from missing values in a high proportion of cases, whose removal may introduce severe bias. Several multiple imputation algorithms have been proposed to attempt to recover the missing information under an assumed missingness mechanism. Each algorithm presents strengths and weaknesses, and there is currently no consensus on which multiple imputation algorithm works best in a given scenario. Furthermore, the selection of each algorithm's parameters and data-related modeling choices are also both crucial and challenging. In this paper we propose a novel framework to numerically evaluate strategies for handling missing data in the context of statistical analysis, with a particular focus on multiple imputation techniques. We demonstrate the feasibility of our approach on a large cohort of type-2 diabetes patients provided by the National COVID Cohort Collaborative (N3C) Enclave, where we explored the influence of various patient characteristics on outcomes related to COVID-19. Our analysis included classic multiple imputation techniques as well as simple complete-case Inverse Probability Weighted models. Extensive experiments show that our approach can effectively highlight the most promising and performant missing-data handling strategy for our case study. Moreover, our methodology allowed a better understanding of the behavior of the different models and of how it changed as we modified their parameters. Our method is general and can be applied to different research fields and on datasets containing heterogeneous types.
Date of publication
Keyword
DOI
Identifier
Resource type
  • Article
License
  • Attribution-NonCommercial-NoDerivs 4.0 International
Journal title
  • Journal of Biomedical Informatics
Journal volume
  • 139
Language
  • English
Version
  • Publisher
Funder
  • Consejo Nacional de Ciencia y Tecnología, Paraguay, CONACYT, (PINV15–315, 14-INV-088, PINV20-337)
  • Università degli Studi di Milano, UniMi, (2015-17 PSR2015-17)
  • Biotechnology and Biological Sciences Research Council, BBSRC, (BB/F00964X/1, BB/M025047/1, BB/K004131/1)
  • Medical Research Council, MRC, (MR/T001070/1)
  • Fundação Getulio Vargas, FGV
  • Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro, FAPERJ, (260380, E-26/201.079/2021)
  • National Science Foundation Advances in Bio Informatics, (1660648)
ISSN
  • 1532-0464
Publisher
  • Academic Press Inc.

Relations

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details 1-s2.0-S1532046423000163-main.pdf 2023-04-05 Public Download