Comparison of Random Forest and Parametric Imputation Models for Imputing Missing Data Using MICE: A CALIBER Study

Multivariate imputation by chained equations (MICE) is commonly used for imputing missing data in epidemiologic research. The "true" imputation model may contain nonlinearities which are not included in default imputation models. Random forest imputation is a machine learning technique which can accommodate nonlinearities and interactions and does not require a particular regression model to be specified. We compared parametric MICE with a random forest-based MICE algorithm in 2 simulation studies. The first study used 1,000 random samples of 2,000 persons drawn from the 10,128 stable angina patients in the CALIBER database (Cardiovascular Disease Research using Linked Bespoke Studies and Electronic Records;…

• NI
  National Institute for Social Care and Health Research

  Award: MR/K006584/1
• LS
  London School of Hygiene and Tropical Medicine
• WT
  Wellcome Trust

  Awards: MR/K006584/1, 086091, 086091/Z/08/Z
• CR
  Cancer Research UK
• NI
  National Institute for Health and Care Research

  Award: RP-PG-0407-10314
• BH
  British Heart Foundation
• UC
  University College London
• MR
  Medical Research Council

  Awards: MR/K006584/1, MR/K006584/1, MC_EX_G0800814, K006584/1, MR/K02180X/1, G0902393
• EA
  Engineering and Physical Sciences Research Council
• EA
  Economic and Social Research Council

  Awards: ES/G026300/1, ES/H022252/1