Should multiple imputation be stratified by exposure group when estimating causal effects via outcome regression in observational studies?

BMC Medical Research Methodology

Research

Jiaxin Zhang¹  Margarita Moreno-Betancur¹  Katherine J. Lee¹  John B. Carlin¹  S Ghazaleh Dashti¹ 

[1] Clinical Epidemiology and Biostatistics Unit, Department of Paediatrics, University of Melbourne, Parkville, Australia;Clinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute, 50 Flemington Road, 3052, Parkville, Australia;

关键词: Causal inference;    Multiple imputation;    Outcome regression;    Observational study;    Missing data;    Target trial;   

DOI  :  10.1186/s12874-023-01843-6

 received in 2022-05-25, accepted in 2023-01-16,  发布年份 2023

来源: Springer

【 摘 要 】

BackgroundDespite recent advances in causal inference methods, outcome regression remains the most widely used approach for estimating causal effects in epidemiological studies with a single-point exposure and outcome. Missing data are common in these studies, and complete-case analysis (CCA) and multiple imputation (MI) are two frequently used methods for handling them. In randomised controlled trials (RCTs), it has been shown that MI should be conducted separately by treatment group. In observational studies, causal inference is now understood as the task of emulating an RCT, which raises the question of whether MI should be conducted by exposure group in such studies.MethodsWe addressed this question by evaluating the performance of seven methods for handling missing data when estimating causal effects with outcome regression. We conducted an extensive simulation study based on an illustrative case study from the Victorian Adolescent Health Cohort Study, assessing a range of scenarios, including seven outcome generation models with exposure-confounder interactions of differing strength.ResultsThe simulation results showed that MI by exposure group led to the least bias when the size of the smallest exposure group was relatively large, followed by MI approaches that included the exposure-confounder interactions.ConclusionsThe findings from our simulation study, which was designed based on a real case study, suggest that current practice for the conduct of MI in causal inference may need to shift to stratifying by exposure group where feasible, or otherwise including exposure-confounder interactions in the imputation model.

【 授权许可】

CC BY   
© The Author(s) 2023

【 图 表 】

Fig. 1

Fig. 2

Fig. 1

Fig. 3

Fig. 2

Fig. 5

Fig. 2

Fig. 1

Fig. 1

Fig. 3

42004_2023_830_Article_IEq16.gif

Fig. 1

Fig. 3

42004_2023_830_Article_IEq12.gif

Fig. 6

Fig. 1

Fig. 2

Fig. 1

Fig. 3

Fig. 2

40854_2023_458_Article_IEq23.gif

Fig. 1

Fig. 7

Fig. 5

40854_2023_458_Article_IEq19.gif

Fig. 2

Fig. 6

Fig. 2

Fig. 6

Fig. 2

Fig. 1

12302_2023_718_Article_IEq70.gif

Fig. 3

Fig. 2

Fig. 2

Fig. 4

Fig. 4

Fig. 2

42004_2023_830_Article_IEq4.gif

42004_2023_830_Article_IEq3.gif

Fig. 5

Fig. 2

12302_2023_718_Article_IEq58.gif

Fig. 3

12302_2023_718_Article_IEq56.gif

Fig. 1

Fig. 1

12302_2023_718_Article_IEq54.gif

Fig. 2

Fig. 2

12302_2023_718_Article_IEq52.gif

Fig. 3

Fig. 1

12302_2023_718_Article_IEq50.gif

40854_2023_458_Article_IEq14.gif

12302_2023_718_Article_IEq49.gif

Fig. 4

Fig. 1

12302_2023_718_Article_IEq44.gif

Fig. 1

Fig. 1

Fig. 5

Fig. 2

Scheme 1

Fig. 3

Fig. 1

Fig. 6

Fig. 6

13046_2023_2621_Article_IEq2.gif

Fig. 8

Fig. 1

Fig. 1

Fig. 1

Fig. 5

Fig. 7

Fig. 1

Fig. 3

Fig. 2

Fig. 1

Fig. 1

Fig. 4

Fig. 7

Fig. 1

Fig. 2

Fig. 6

12888_2023_4583_Article_IEq4.gif

Fig. 5

Fig. 8

Fig. 5

Fig. 2