Multiple Imputation for Multivariate Missing Data: The Joint Modeling Approach

2021 ◽  
pp. 143-180
Author(s):  
Yulei He ◽  
Guangyu Zhang ◽  
Chiu-Hsieh Hsu
Biostatistics ◽  
2017 ◽  
Vol 19 (4) ◽  
pp. 479-496 ◽  
Author(s):  
Margarita Moreno-Betancur ◽  
John B Carlin ◽  
Samuel L Brilleman ◽  
Stephanie K Tanamas ◽  
Anna Peeters ◽  
...  

2017 ◽  
Vol 43 (3) ◽  
pp. 316-353 ◽  
Author(s):  
Simon Grund ◽  
Oliver Lüdtke ◽  
Alexander Robitzsch

Multiple imputation (MI) can be used to address missing data at Level 2 in multilevel research. In this article, we compare joint modeling (JM) and the fully conditional specification (FCS) of MI as well as different strategies for including auxiliary variables at Level 1 using either their manifest or their latent cluster means. We show with theoretical arguments and computer simulations that (a) an FCS approach that uses latent cluster means is comparable to JM and (b) using manifest cluster means provides similar results except in relatively extreme cases with unbalanced data. We outline a computational procedure for including latent cluster means in an FCS approach using plausible values and provide an example using data from the Programme for International Student Assessment 2012 study.


Author(s):  
Simon Grund ◽  
Oliver Lüdtke ◽  
Alexander Robitzsch

AbstractMultilevel models often include nonlinear effects, such as random slopes or interaction effects. The estimation of these models can be difficult when the underlying variables contain missing data. Although several methods for handling missing data such as multiple imputation (MI) can be used with multilevel data, conventional methods for multilevel MI often do not properly take the nonlinear associations between the variables into account. In the present paper, we propose a sequential modeling approach based on Bayesian estimation techniques that can be used to handle missing data in a variety of multilevel models that involve nonlinear effects. The main idea of this approach is to decompose the joint distribution of the data into several parts that correspond to the outcome and explanatory variables in the intended analysis, thus generating imputations in a manner that is compatible with the substantive analysis model. In three simulation studies, we evaluate the sequential modeling approach and compare it with conventional as well as other substantive-model-compatible approaches to multilevel MI. We implemented the sequential modeling approach in the R package and provide a worked example to illustrate its application.


Rheumatology ◽  
2021 ◽  
Vol 60 (Supplement_1) ◽  
Author(s):  
Alice Gottlieb ◽  
Frank Behrens ◽  
Peter Nash ◽  
Joseph F Merola ◽  
Pascale Pellet ◽  
...  

Abstract Background/Aims  Psoriatic arthritis (PsA) is a heterogeneous disease comprising musculoskeletal and dermatological manifestations, especially plaque psoriasis. Secukinumab, an interleukin17A inhibitor, provided significantly greater PASI75/100 responses in two head-to-head trials versus etanercept or ustekinumab, a tumour necrosis factor inhibitor (TNFi), in patients with moderate-to-severe plaque psoriasis. The EXCEED study (NCT02745080) investigated whether secukinumab was superior to adalimumab, another TNFi, as monotherapy in biologic-naive active PsA patients with active plaque psoriasis (defined as having ≥1 psoriatic plaque of ≥ 2 cm diameter, nail changes consistent with psoriasis or documented history of plaque psoriasis). Here we report the pre-specified skin outcomes from the EXCEED study in the subset of patients with ≥3% body surface area (BSA) affected with psoriasis at baseline. Methods  In this head-to-head, Phase 3b, randomised, double-blind, active-controlled, multicentre, parallel-group trial, patients were randomised to receive subcutaneous secukinumab 300 mg at baseline and Weeks 1-4, followed by dosing every 4 weeks until Week 48, or subcutaneous adalimumab 40 mg at baseline followed by the same dosing every 2 weeks until Week 50. The primary endpoint was superiority of secukinumab versus adalimumab on ACR20 response at Week 52. Pre-specified outcomes included the proportion of patients achieving a combined ACR50 and PASI100 response, PASI100 response, and absolute PASI score ≤3. Missing data were handled using multiple imputation. Results  Overall, 853 patients were randomised to receive secukinumab (n = 426) or adalimumab (n = 427). At baseline, 215 and 202 patients had at least 3% BSA affected with psoriasis in the secukinumab and adalimumab groups, respectively. At Week 52, more patients achieved simultaneous improvement in ACR50 and PASI100 response with secukinumab versus adalimumab (30.7% versus 19.2%, respectively; P = 0.0087). Greater efficacy was demonstrated for secukinumab versus adalimumab for PASI100 responses and for the proportion of patients achieving absolute PASI score ≤3 (Table 1). Conclusion  In this pre-specified analysis, secukinumab provided higher responses compared with adalimumab in achievement of combined improvement in joint and skin disease (combined ACR50 and PASI100 response) and in skin-specific endpoints (PASI100 and absolute PASI score ≤3) at Week 52. P189 Table 1:Skin-specific outcomes at Week 52Endpoints, % responseSEC 300 mg (N = 215)ADA 40 mg (N = 202)P value (unadjusted)PASI10046300.0007Combined ACR50 and PASI10031190.0087Absolute PASI score ≤379650.0015P value vs ADA; unadjusted P values are presented. Multiple imputation was used for handling missing data. ADA, adalimumab; ACR, American College of Rheumatology; N, number of patients in the psoriasis subset; PASI, Psoriasis Area and Severity Index; SEC, secukinumab. Disclosure  A. Gottlieb: Grants/research support; A.G. has received research support, consultation fees or speaker honoraria from Pfizer, AbbVie, BMS, Lilly, MSD, Novartis, Roche, Sanofi, Sandoz, Nordic, Celltrion and UCB. F. Behrens: Consultancies; F.B. is a consultant for Pfizer, AbbVie, Sanofi, Lilly, Novartis, Genzyme, Boehringer Ingelheim, Janssen, MSD, Celgene, Roche and Chugai. Grants/research support; F.B. has received grant/research support from Pfizer, Janssen, Chugai, Celgene, Lilly and Roche. P. Nash: Consultancies; P.N. is a consultant for AbbVie, Bristol Myers Squibb, Celgene, Eli Lilly, Gilead, Janssen, MSD, Novartis, Pfizer Inc., Roche, Sanofi and UCB. Member of speakers’ bureau; for AbbVie, Bristol Myers Squibb, Celgene, Eli Lilly, Gilead, Janssen, MSD, Novartis, Pfizer Inc., Roche, Sanofi and UCB. Grants/research support; P.N. has received research support from AbbVie, Bristol Myers Squibb, Celgene, Eli Lilly and Company, Gilead, Janssen, MSD, Novartis, Pfizer Inc, Roche, Sanofi and UCB. J. Merola: Consultancies; J.F.M. is a consultant for Merck, AbbVie, Dermavant, Eli Lilly, Novartis, Janssen, UCB Pharma, Celgene, Sanofi, Regeneron, Arena, Sun Pharma, Biogen, Pfizer, EMD Sorono, Avotres and LEO Pharma. P. Pellet: Corporate appointments; P.P. is an employee of Novartis. Shareholder/stock ownership; P.P. is a shareholder of Novartis. L. Pricop: Corporate appointments; L.P. is an employee of Novartis. Shareholder/stock ownership; L.P. is a shareholder of Novartis. I. McInnes: Consultancies; I.M. is a consultant for AbbVie, Bristol Myers Squibb, Celgene, Eli Lilly and Company, Gilead, Janssen, Novartis, Pfizer and UCB. Grants/research support; I.M. has received grant/research support from Bristol Myers Squibb, Celgene, Eli Lilly and Company, Janssen and UCB.


Psychometrika ◽  
2011 ◽  
Vol 76 (3) ◽  
pp. 487-503 ◽  
Author(s):  
T. Loeys ◽  
Y. Rosseel ◽  
K. Baten

2018 ◽  
Vol Volume 10 ◽  
pp. 1869-1877 ◽  
Author(s):  
John M. Dennis ◽  
Beverley M. Shields ◽  
Angus G. Jones ◽  
Ewan R. Pearson ◽  
Andrew T. Hattersley ◽  
...  

2021 ◽  
Vol 50 (Supplement_1) ◽  
Author(s):  
Jiaxin Zhang ◽  
S. Ghazaleh Dashti ◽  
John B. Carlin ◽  
Katherine J. Lee ◽  
Margarita Moreno-Betancur

Abstract Background Outcome regression remains widely applied for estimating causal effects in observational studies, in which causal inference is conceptualised as emulating a randomized controlled trial (RCT). Multiple imputation (MI) is a commonly used method for handling missing data, but while in RCTs it has been shown that MI should be conducted by treatment group to reduce bias, whether imputation should be conducted by exposure group in observational studies has not been studied. Methods We conducted a simulation study to evaluate the performance of seven methods for handling missing data: Complete-case analysis (CCA), MI of main effect, MI with interactions (between exposure and: outcome, a strong confounder, outcome and a strong confounder, all incomplete), and MI conducted by exposure group. We simulated data based on an example from the Victorian Adolescent Health Cohort Study. Three exposure prevalences and seven outcome generation models were considered, the latter ranging from no interaction to strong-positive or negative exposure-confounder interaction. Various missingness scenarios were examined: with incomplete outcome only or also incomplete confounders, and three levels of complexity regarding the missingness mechanism. Results For all scenarios, MI by exposure led to the least bias, followed by MI approaches that included exposure-confounder interactions. Conclusions If MI is adopted in outcome regression, we recommend conducting MI by exposure group and, when not feasible, including exposure-confounder interactions in the imputation model. Key messages Similar to RCTs, MI should be conducted by exposure group when estimating average causal effects using outcome regression in observational studies.


Sign in / Sign up

Export Citation Format

Share Document