A comparison of multiple imputation strategies for handling missing data in multi‐item scales: Guidance for longitudinal studies

Mainzer, Rheanna; Apajee, Jemishabye; Nguyen, Cattram; Lee, Katherine J.

doi:10.1002/sim.9088

Cited by 18 publications

(13 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The primary analysis was intention-to-treat, with children analyzed in their randomized group, irrespective of whether they completed the trial. We used multiple imputation with chained equations (MICE) approach to compensate for data missingness to create ten imputed data sets and combined these results with Rubin's rules [30]. As multiple imputation has consistently been shown to be a valid approach in handing missing data, only estimates obtained by multiple imputation are reported [31].…”

Section: Discussionmentioning

confidence: 99%

Cost-effectiveness of an Intensive Upper Limb Rehabilitation Therapy for Children With Unilateral Cerebral Palsy: an Economic Evaluation of a Randomized Controlled Trial

David

Higashi

2022

Preprint

View full text Add to dashboard Cite

BackgroundUnilateral cerebral palsy (UCP) is a major cause of childhood disability and a substantial economic burden. Intensive group-based therapy of either hybrid constraint-induced movement therapy (mCIMT) or bimanual therapy has been shown to be effective in improving specific domains of the Quality of Life (QoL) in children with this disability. However, there is a paucity of economic evaluation on its role as an adjunct to other therapies currently in use. This study aimed to address this paucity by evaluating whether an intensive group-based therapy model is cost-effective compared to individualized distributed standard care occupational therapy (SC).MethodsAn open-label parallel randomized controlled trial (RCT) and an embedded economic evaluation of intensive group-based hybrid-CIMT (combined mCIMT and bimanual therapy) were conducted. A total of 47 children were randomized to the hybrid-CIMT group (n=27) or the SC (n=20) group. Effectiveness was assessed using the Cerebral Palsy Quality of Life (Child) questionnaire across seven domains: 1) Social well-being and acceptance; 2) Functioning; 3) Participation and physical health; 4) Emotional well-being; 5) Access to services; 6) Pain; and 7) Impact of disability and family health. Using trial data, costs in Australian dollars (AUD) were assessed from a societal perspective. Non-parametric bootstrap analysis was used to quantify uncertainty intervals (UIs; around point estimates) accompanying incremental cost-effectiveness ratio estimates, the cost per unit of beneficial change in the quality of life.ResultsOnly for the domains of Pain (Incremental cost-effectiveness ratio: $273; 95% Uncertainty interval: $107 - $945) and to a lesser extent, Access to services ($1732; -$6448 - $8775) and the Impact of disability and family health ($1071; -$5718 - $4606), was hybrid-CIMT therapy found to be cost-effective when compared to SC. The probability that hybrid-CIMT was cost-effective for the Pain domain was estimated to be 0.50.ConclusionsHybrid-CIMT was not found to be cost-effective when compared to SC over a 13-week time horizon. In order to assess benefits and costs more fully, we recommend that future research utilize time horizons that extend sufficiently into the future. Irrespective of study duration, therapy comparisons should also consider acceptance, equity and feasibility.Trial registrationACTRN12609000912280. 21/10/2009, https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?ACTRN=12609000912280

show abstract

Section: Discussionmentioning

confidence: 99%

Cost-effectiveness of an Intensive Upper Limb Rehabilitation Therapy for Children With Unilateral Cerebral Palsy: an Economic Evaluation of a Randomized Controlled Trial

David

Higashi

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In addition, many patients had at least one or more measurements recorded during follow-up, suggesting that a sensible imputation approach to explore the longitudinal data structure is Multiple Imputation with chained equation (MICE), 19 while incorporating the partially available measurements. Several detailed accounts of appropriate multiple imputation methods are available for different clinical settings including online advice from Van Buuren 20 and guidance from Mainzer et al 21 Two limitations of our analyses require further consideration. First, we only considered a dataset from a single trial.…”

Section: Discussionmentioning

confidence: 99%

Characterization of missing data patterns and mechanisms in longitudinal composite outcome trial in rheumatoid arthritis

Ibrahim

Tom

Scott

et al. 2022

Therapeutic Advances in Musculoskeletal

View full text Add to dashboard Cite

Background: Composite measures, like the Disease Activity Score for 28 joints (DAS28), are key primary outcomes in rheumatoid arthritis (RA) trials. DAS28 combines four different components in a continuous measure. When one or more of these components are missing the overall composite score is also missing at intermediate or trial endpoint assessments. Objectives: This study examined missing data patterns and mechanisms in a longitudinal RA trial to evaluate how best to handle missingness when analysing composite outcomes. Design: The Tumour-Necrosis-Factor Inhibitors against Combination Intensive Therapy (TACIT) trial was an open label, pragmatic randomized multicentre two arm non-inferiority study. Patients were followed up for 12 months, with monthly measurement of the composite outcome and its components. Active RA patients were randomized to conventional disease modifying drugs (cDMARDs) or Tumour Necrosis Factor-α inhibitors (TNFis). Methods: The TACIT trial was used to explore the extent of missing data in the composite outcome, DAS28. Patterns of missing data in components and the composite outcome were examined graphically. Longitudinal multivariable logistic regression analysis assessed missing data mechanisms during follow-up. Results: Two hundred and five patients were randomized: at 12 months 59/205 (29%) had unobserved composite outcome and 146/205 (71%) had an observed DAS28 outcome; however, 34/146 had one or more intermediate assessments missing. We observed mixed missing data patterns, especially for the missing composite outcome due to one component missing rather than patient not attending thier visit. Age and gender predicted missingness components, providing strong evidence the missing observations were unlikely to be Missing Completely at Random (MCAR). Conclusion: Researchers should undertake detailed evaluations of missing data patterns and mechanisms at the final and intermediate time points, whether or not the outcome variable is a composite outcome. In addition, the impact on treatment estimates in patients who only provide data at milestone assessments need to be assessed. Trial Registration ISRCTN Number: 37438295

show abstract

“…We have shown previously that individuals with missing data on one or more variables in the analysis model had lower mean HRQoL and SEP, and were more likely to be Aboriginal or Torres Strait Islander or come from a non-English speaking background than individuals with complete data on all variables. [38] Furthermore, a large number of potential auxiliary variables were available in the dataset. The assumption that non-response in HRQoL does not depend on HRQoL itself after conditioning on variables in the imputation model (under which MI is valid) is more plausible when the conditioning set includes auxiliary variables, and even more so if there are a large number of these.…”

Section: Motivating Example: the Longitudinal Study Of Australian Chi...mentioning

confidence: 99%

“…for missing data in more than one variable, MVNI could be used instead of multiple imputation by chained equations [46]; or changing the level at which variables are imputed (applicable for derived variables or scale scores). [13,38] However, these approaches also have limitations. Perhaps the most effective approach for overcoming challenges with the MI algorithm would be to identify and then remove from the imputation model the problematic variables.…”

Section: Lsac Examplementioning

confidence: 99%

See 1 more Smart Citation

A comparison of strategies for selecting auxiliary variables for multiple imputation

Mainzer¹,

Nguyen²,

Moreno‐Betancur³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Multiple imputation (MI) is a popular method for handling missing data. Auxiliary variables can be added to the imputation model(s) to improve MI estimates. However, the choice of which auxiliary variables to include in the imputation model is not always straightforward.Including too few may lead to important information being discarded, but including too many can cause problems with convergence of the estimation procedures for imputation models.Several data-driven auxiliary variable selection strategies have been proposed. This paper uses a simulation study and a case study to provide a comprehensive comparison of the performance of eight auxiliary variable selection strategies, with the aim of providing practical advice to users of MI. A complete case analysis and an MI analysis with all auxiliary variables included in the imputation model (the full model) were also performed for comparison.Our simulation study results suggest that the full model outperforms all auxiliary variable selection strategies, providing further support for adopting an inclusive auxiliary variable strategy where possible. Auxiliary variable selection using the Least Absolute Selection and Shrinkage Operator (LASSO) was the best performing auxiliary variable selection strategy overall and is a promising alternative when the full model fails. All MI analysis strategies that we were able to apply to the case study led to similar estimates.

show abstract

A comparison of multiple imputation strategies for handling missing data in multi‐item scales: Guidance for longitudinal studies

Cited by 18 publications

References 43 publications

Cost-effectiveness of an Intensive Upper Limb Rehabilitation Therapy for Children With Unilateral Cerebral Palsy: an Economic Evaluation of a Randomized Controlled Trial

Cost-effectiveness of an Intensive Upper Limb Rehabilitation Therapy for Children With Unilateral Cerebral Palsy: an Economic Evaluation of a Randomized Controlled Trial

Characterization of missing data patterns and mechanisms in longitudinal composite outcome trial in rheumatoid arthritis

A comparison of strategies for selecting auxiliary variables for multiple imputation

Contact Info

Product

Resources

About