Missing data in a multi-item instrument were best handled by multiple imputation at the item score level

Iris Eekhout; Henrica C W de Vet; Jos W R Twisk; Jaap P L Brand; Michiel R de Boer; Martijn W Heymans

doi:10.1016/j.jclinepi.2013.09.009

Missing data in a multi-item instrument were best handled by multiple imputation at the item score level

J Clin Epidemiol. 2014 Mar;67(3):335-42. doi: 10.1016/j.jclinepi.2013.09.009. Epub 2013 Dec 2.

Authors

Iris Eekhout¹, Henrica C W de Vet², Jos W R Twisk³, Jaap P L Brand⁴, Michiel R de Boer⁵, Martijn W Heymans³

Affiliations

¹ Department of Epidemiology and Biostatistics, VU University Medical Center, P.O. box 7057, 1007 MB Amsterdam, The Netherlands; EMGO Institute for Health and Care Research, VU University Medical Center, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands; Department of Methodology and Applied Biostatistics, Faculty of Earth and Life Sciences, Institute for Health Sciences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands. Electronic address: i.eekhout@vumc.nl.
² Department of Epidemiology and Biostatistics, VU University Medical Center, P.O. box 7057, 1007 MB Amsterdam, The Netherlands; EMGO Institute for Health and Care Research, VU University Medical Center, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands.
³ Department of Epidemiology and Biostatistics, VU University Medical Center, P.O. box 7057, 1007 MB Amsterdam, The Netherlands; EMGO Institute for Health and Care Research, VU University Medical Center, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands; Department of Methodology and Applied Biostatistics, Faculty of Earth and Life Sciences, Institute for Health Sciences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands.
⁴ Skyline Diagnostics, Marconistraat 16, 3029 AK Rotterdam, The Netherlands.
⁵ Department of Methodology and Applied Biostatistics, Faculty of Earth and Life Sciences, Institute for Health Sciences, VU University, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands; Department of Public Health, University Medical Center Groningen, PO box 196, 9700 AD Groningen, The Netherlands.

PMID: 24291505
DOI: 10.1016/j.jclinepi.2013.09.009

Abstract

Objectives: Regardless of the proportion of missing values, complete-case analysis is most frequently applied, although advanced techniques such as multiple imputation (MI) are available. The objective of this study was to explore the performance of simple and more advanced methods for handling missing data in cases when some, many, or all item scores are missing in a multi-item instrument.

Study design and setting: Real-life missing data situations were simulated in a multi-item variable used as a covariate in a linear regression model. Various missing data mechanisms were simulated with an increasing percentage of missing data. Subsequently, several techniques to handle missing data were applied to decide on the most optimal technique for each scenario. Fitted regression coefficients were compared using the bias and coverage as performance parameters.

Results: Mean imputation caused biased estimates in every missing data scenario when data are missing for more than 10% of the subjects. Furthermore, when a large percentage of subjects had missing items (>25%), MI methods applied to the items outperformed methods applied to the total score.

Conclusion: We recommend applying MI to the item scores to get the most accurate regression model estimates. Moreover, we advise not to use any form of mean imputation to handle missing data.

Keywords: Bias; Item imputation; Methods; Missing data; Multi-item questionnaire; Multiple imputation; Simulation.

MeSH terms

Bias
Computer Simulation
Data Interpretation, Statistical*
Humans
Linear Models
Research Design