Dataset Information

Multiple imputation by predictive mean matching in cluster-randomized trials.

ABSTRACT:

Background

Random effects regression imputation has been recommended for multiple imputation (MI) in cluster randomized trials (CRTs) because it is congenial to analyses that use random effects regression. This method relies heavily on model assumptions and may not be robust to misspecification of the imputation model. MI by predictive mean matching (PMM) is a semiparametric alternative, but current software for multilevel data relies on imputation models that ignore clustering or use fixed effects for clusters. When used directly for imputation, these two models result in underestimation (ignoring clustering) or overestimation (fixed effects for clusters) of variance estimates.

Methods

We develop MI procedures based on PMM that leverage these opposing estimated biases in the variance estimates in one of three ways: weighting the distance metric (PMM-dist), weighting the average of the final imputed values from two PMM procedures (PMM-avg), or performing a weighted draw from the final imputed values from the two PMM procedures (PMM-draw). We use Monte-Carlo simulations to evaluate our newly proposed methods relative to established MI procedures, focusing on estimation of treatment group means and their variances after MI.

Results

The proposed PMM procedures reduce the bias in the MI variance estimator relative to established methods when the imputation model is correctly specified, and are generally more robust to model misspecification than even the random effects imputation methods.

Conclusions

The PMM-draw procedure in particular is a promising method for multiply imputing missing data from CRTs that can be readily implemented in existing statistical software.

SUBMITTER: Bailey BE

PROVIDER: S-EPMC7106802 | biostudies-literature | 2020 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Multiple imputation by predictive mean matching in cluster-randomized trials.

Bailey Brittney E BE Andridge Rebecca R Shoben Abigail B AB

BMC medical research methodology 20200330 1

<h4>Background</h4>Random effects regression imputation has been recommended for multiple imputation (MI) in cluster randomized trials (CRTs) because it is congenial to analyses that use random effects regression. This method relies heavily on model assumptions and may not be robust to misspecification of the imputation model. MI by predictive mean matching (PMM) is a semiparametric alternative, but current software for multilevel data relies on imputation models that ignore clustering or use fi ...[more]

PMID: 32228491

Similar Datasets

Project description:BackgroundIn cluster-randomized controlled trials (C-RCTs), covariate-constrained randomization (CCR) methods efficiently control imbalance in multiple baseline cluster-level variables, but the choice of imbalance metric to define the subset of "adequately balanced" possible allocation schemes for C-RCTs involving more than two arms and continuous variables is unclear. In an ongoing three-armed C-RCT, we chose the min(three Kruskal-Wallis [KW] test P values) > 0.30 as our metric. We use simulation studies to explore the performance of this and other metrics of baseline variable imbalance in CCR.MethodsWe simulated three continuous variables across three arms under varying allocation ratios and assumptions. We compared the performance of min(analysis of variance [ANOVA] P value) > 0.30, min(KW P value) > 0.30, multivariate analysis of variance (MANOVA) P value > 0.30, min(nine possible t test P values) > 0.30, and min(Wilcoxon rank-sum [WRS] P values) > 0.30.ResultsPairwise comparison metrics (t test and WRS) tended to be the most conservative, providing the smallest subset of allocation schemes (10%-13%) meeting criteria for acceptable balance. Sensitivity of the min(t test P values) > 0.30 for detecting non-trivial imbalance was 100% for both hypothetical and resampled simulation scenarios. The KW criterion maintained higher sensitivity than both the MANOVA and ANOVA criteria (89% to over 99%) but was not as sensitive as pairwise criteria.ConclusionsOur criterion, the KW P value > 0.30, to signify "acceptable" balance was not the most conservative, but it appropriately identified imbalance in the majority of simulations. Since all are related, CCR algorithms involving any of these imbalance metrics for continuous baseline variables will ensure robust simultaneous control over multiple continuous baseline variables, but we recommend care in determining the threshold of "acceptable" levels of (im)balance.Trial registrationThis trial is registered on ClinicalTrials.gov (initial post: December 1, 2016; identifier: NCT02979444 ).

Project description:BackgroundRecruitment and engagement of clusters in a cluster randomized controlled trial can sometimes prove challenging. Identification of successful or unsuccessful strategies may be beneficial in guiding future researchers in conducting their cluster randomized controlled trial. This study aimed to identify strategies that could be used to facilitate the delivery of cluster randomized controlled trials in hospitals.MethodsThe study employed the Consolidated Framework for Implementation Research-Expert Recommendations for Implementing Change matching tool. The barriers and enablers to cluster randomized controlled trial conduct identified in our previously conducted studies served as a means of determinant identification for the conduct of cluster randomized controlled trials. These determinants were mapped to Consolidated Framework for Implementation Research constructs and then matched to Expert Recommendations for Implementing Change compilation strategies using the Consolidated Framework for Implementation Research-Expert Recommendations for Implementing Change matching tool.ResultsThe Expert Recommendations for Implementing Change strategies matched to at least one determinant Consolidated Framework for Implementation Research construct were as follows: (1) 'Identify and prepare champions', (2) 'Conduct local needs assessment', (3) 'Conduct educational meetings', (4) 'Inform local opinion leaders', (5) 'Build a coalition', (6) 'Promote adaptability', (7) 'Develop a formal implementation blueprint', (8) 'Involve patients/consumers and family members', (9) 'Obtain and use patients/consumers and family feedback', (10) 'Develop educational materials', (11) 'Promote network weaving', (12) 'Distribute educational materials', (13) 'Access new funding' and (14) 'Develop academic partnerships'.ConclusionThis study was intended as a step in the research agenda aimed at facilitating cluster randomized controlled trial delivery in hospitals and can act as a resource for future researchers when planning their cluster randomized controlled trial, with the expectation that the strategies identified here will be tailored to each context.

Project description:BackgroundIn cluster randomized crossover (CRXO) trials, groups of participants (i.e., clusters) are randomly allocated to receive a sequence of interventions over time (i.e., cluster periods). CRXO trials are becoming more comment when they are feasible, as they require fewer clusters than parallel group cluster randomized trials. However, CRXO trials have not been frequently used in orthopedic fracture trials and represent a novel methodological application within the field. To disseminate the early knowledge gained from our experience initiating two cluster randomized crossover trials, we describe our process for the identification and selection of the orthopedic practices (i.e., clusters) participating in the PREP-IT program and present data to describe their key characteristics.MethodsThe PREP-IT program comprises two ongoing pragmatic cluster randomized crossover trials (Aqueous-PREP and PREPARE) which compare the effect of iodophor versus chlorhexidine solutions on surgical site infection and unplanned fracture-related reoperations in patients undergoing operative fracture management. We describe the process we used to identify and select orthopedic practices (clusters) for the PREP-IT trials, along with their characteristics.ResultsWe identified 58 potential orthopedic practices for inclusion in the PREP-IT trials. After screening each practice for eligibility, we selected 30 practices for participation and randomized each to a sequence of interventions (15 for Aqueous-PREP and 20 for PREPARE). The majority of orthopedic practices included in the Aqueous-PREP and PREPARE trials were situated in level I trauma centers (100% and 87%, respectively). Orthopedic practices in the Aqueous-PREP trial operatively treated a median of 149 open fracture patients per year, included a median of 11 orthopedic surgeons, and had access to a median of 5 infection preventionists. Orthopedic practices in the PREPARE trial treated a median of 142 open fracture and 1090 closed fracture patients per year, included a median of 7.5 orthopedic surgeons, and had access to a median of 6 infection preventionists.ConclusionsThe PREP-IT trials provide an example of how to follow the reporting standards for cluster randomized crossover trials by providing a clear definition of the cluster unit, a thorough description of the cluster identification and selection process, and sufficient description of key cluster characteristics.Trial registrationBoth trials are registered at ClinicalTrials.gov (A-PREP: NCT03385304 December 28, 2017, and PREPARE: NCT03523962 May 14, 2018).

Dataset Information

Multiple imputation by predictive mean matching in cluster-randomized trials.

Background

Methods

Results

Conclusions

Publications

Multiple imputation by predictive mean matching in cluster-randomized trials.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets