Mark Jones1, Gita D Mishra2, Annette Dobson2. 1. School of Public Health, University of Queensland, Public Health Building, Herston Road, Herston, Brisbane, Qld 4006, Australia. Electronic address: m.jones@sph.uq.edu.au. 2. School of Public Health, University of Queensland, Public Health Building, Herston Road, Herston, Brisbane, Qld 4006, Australia.
Abstract
OBJECTIVES: To compare methods for analysis of longitudinal studies with missing data due to participant dropout and follow-up truncated by death. STUDY DESIGN AND SETTING: We analyzed physical functioning in an Australian longitudinal study of elderly women where the missing data mechanism could either be missing at random (MAR) or missing not at random (MNAR). We assumed either an immortal cohort where deceased participants are implicitly included after death or a mortal cohort where the target of inference is surviving participants at each survey wave. To illustrate the methods a covariate was included. Simulation was used to assess the effect of the assumptions. RESULTS: Ignoring attrition or restricting analysis to participants with complete follow up led to biased estimates. Linear mixed model was appropriate for an immortal cohort under MAR but not MNAR. Linear increment model and joint modeling of longitudinal outcome and time to death were the most robust to MNAR. For a mortal cohort, inverse probability weighting and multiple imputation could be used, but care is needed in specifying dropout and imputation models, respectively. CONCLUSION: Appropriate analysis methodology to deal with attrition in longitudinal studies depends on the target of inference and the missing data mechanism.
OBJECTIVES: To compare methods for analysis of longitudinal studies with missing data due to participant dropout and follow-up truncated by death. STUDY DESIGN AND SETTING: We analyzed physical functioning in an Australian longitudinal study of elderly women where the missing data mechanism could either be missing at random (MAR) or missing not at random (MNAR). We assumed either an immortal cohort where deceased participants are implicitly included after death or a mortal cohort where the target of inference is surviving participants at each survey wave. To illustrate the methods a covariate was included. Simulation was used to assess the effect of the assumptions. RESULTS: Ignoring attrition or restricting analysis to participants with complete follow up led to biased estimates. Linear mixed model was appropriate for an immortal cohort under MAR but not MNAR. Linear increment model and joint modeling of longitudinal outcome and time to death were the most robust to MNAR. For a mortal cohort, inverse probability weighting and multiple imputation could be used, but care is needed in specifying dropout and imputation models, respectively. CONCLUSION: Appropriate analysis methodology to deal with attrition in longitudinal studies depends on the target of inference and the missing data mechanism.
Authors: Yu-Tzu Wu; Christina Daskalopoulou; Graciela Muniz Terrera; Albert Sanchez Niubo; Fernando Rodríguez-Artalejo; Jose Luis Ayuso-Mateos; Martin Bobak; Francisco Félix Caballero; Javier de la Fuente; Alejandro de la Torre-Luque; Esther García-Esquinas; Jose Maria Haro; Seppo Koskinen; Ilona Koupil; Matilde Leonardi; Andrzej Pajak; Demosthenes Panagiotakos; Denes Stefler; Beata Tobias-Adamczyk; Martin Prince; A Matthew Prina Journal: Lancet Public Health Date: 2020-07
Authors: D Leann Long; George Howard; Dustin M Long; Suzanne Judd; Jennifer J Manly; Leslie A McClure; Virginia G Wadley; Monika M Safford; Ronit Katz; M Maria Glymour Journal: Am J Epidemiol Date: 2019-03-01 Impact factor: 4.897
Authors: Samuel M Brown; Emily Wilson; Angela P Presson; Chong Zhang; Victor D Dinglas; Tom Greene; Ramona O Hopkins; Dale M Needham Journal: Thorax Date: 2016-07-20 Impact factor: 9.139
Authors: Andreea Monica Rawlings; Yingying Sang; Albert Richey Sharrett; Josef Coresh; Michael Griswold; Anna Maria Kucharska-Newton; Priya Palta; Lisa Miller Wruck; Alden Lawrence Gross; Jennifer Anne Deal; Melinda Carolyn Power; Karen Jean Bandeen-Roche Journal: Eur J Epidemiol Date: 2016-09-12 Impact factor: 8.082