Multiple imputation of missing values was not necessary before performing a longitudinal mixed-model analysis
TL;DR: It is not necessary to handle missing data using multiple imputations before performing a mixed-model analysis on longitudinal data, and the results of the mixed- model analysis with multiple imputation were quite unstable.
About: This article is published in Journal of Clinical Epidemiology.The article was published on 2013-09-01. It has received 324 citations till now. The article focuses on the topics: Imputation (statistics) & Missing data.
Citations
More filters
TL;DR: Evidence is provided demonstrating the potential positive effect of gamification and online social support on health and behavioral outcomes and the Web-based intervention had a positive impact on intervention groups compared to the control group.
Abstract: Background: Rheumatoid arthritis (RA) is chronic systematic disease that affects people during the most productive period of their lives. Web-based health interventions have been effective in many studies; however, there is little evidence and few studies showing the effectiveness of online social support and especially gamification on patients’ behavioral and health outcomes. Objective: The aim of this study was to look into the effects of a Web-based intervention that included online social support features and gamification on physical activity, health care utilization, medication overuse, empowerment, and RA knowledge of RA patients. The effect of gamification on website use was also investigated. Methods: We conducted a 5-arm parallel randomized controlled trial for RA patients in Ticino (Italian-speaking part of Switzerland). A total of 157 patients were recruited through brochures left with physicians and were randomly allocated to 1 of 4 experimental conditions with different types of access to online social support and gamification features and a control group that had no access to the website. Data were collected at 3 time points through questionnaires at baseline, posttest 2 months later, and at follow-up after another 2 months. Primary outcomes were physical activity, health care utilization, and medication overuse; secondary outcomes included empowerment and RA knowledge. All outcomes were self-reported. Intention-to-treat analysis was followed and multilevel linear mixed models were used to study the change of outcomes over time.
215 citations
Cites background from "Multiple imputation of missing valu..."
...Recent research [70] showed that for all types of missing data (missing completely at random, missing at random, and missing not at random), multiple imputation is not necessary before performing longitudinal mixed model analysis....
[...]
TL;DR: Among patients with rheumatoid arthritis previously treated with anti-T NF drugs but with inadequate primary response, a non-TNF biologic agent was more effective in achieving a good or moderate disease activity response at 24 weeks than was the second anti- TNF medication.
Abstract: Importance One-third of patients with rheumatoid arthritis show inadequate response to tumor necrosis factor α (TNF-α) inhibitors; little guidance on choosing the next treatment exists. Objective To compare the efficacy of a non–TNF-targeted biologic (non-TNF) vs a second anti-TNF drug for patients with insufficient response to a TNF inhibitor. Design, Setting, and Participants A total of 300 patients (conducted between 2009-2012) with rheumatoid arthritis, with persistent disease activity (disease activity score in 28 joints–erythrocyte sedimentation rate [DAS28-ESR] ≥ 3.2 [range, 0-9.3]) and an insufficient response to anti-TNF therapy were included in a 52-week multicenter, pragmatic, open-label randomized clinical trial. The final follow-up date was in August 2013. Interventions Patients were randomly assigned (1:1) to receive a non–TNF-targeted biologic agent or an anti-TNF that differed from their previous treatment. The choice of the biologic prescribed within each randomized group was left to the treating clinician. Main Outcomes and Measures The primary outcome was the proportion of patients with good or moderate response according to the European League Against Rheumatism (EULAR) scale at week 24. Secondary outcomes included the EULAR response at weeks 12 and 52; at weeks 12, 24, and 52; DAS28ESR, low disease activity (DAS28 ≤3.2), remission (DAS28 ≤2.6); serious adverse events; and serious infections. Results Of the 300 randomized patients (243 [83.2%] women; mean [SD] age, 57.1 [12.2] years; baseline DAS28-ESR, 5.1 [1.1]), 269 (89.7%) completed the study. At week 24, 101 of 146 patients (69%) in the non-TNF group and 76 (52%) in the second anti-TNF group achieved a good or moderate EULAR response (OR, 2.06; 95% CI, 1.27-3.37; P = .004, with imputation of missing data; absolute difference, 17.2%; 95% CI, 6.2% to 28.2%). The DAS28-ESR was lower in the non-TNF group than in the second anti-TNF group (mean difference adjusted for baseline differences, −0.43; 95% CI, −0.72 to −0.14; P = .004). At weeks 24 and 52, more patients in the non-TNF group vs the second anti-TNF group showed low disease activity (45% vs 28% at week 24; OR, 2.09; 95% CI, 1.27 to 3.43; P = .004 and 41% vs 23% at week 52; OR, 2.26; 95% CI, 1.33 to 3.86; P = .003). Conclusions and Relevance Among patients with rheumatoid arthritis previously treated with anti-TNF drugs but with inadequate primary response, a non-TNF biologic agent was more effective in achieving a good or moderate disease activity response at 24 weeks than was the second anti-TNF medication. Trial Registration clinicaltrials.gov Identifier:NCT01000441
152 citations
TL;DR: At one year the effectiveness of microdecompression is equivalent to laminectomy in the surgical treatment of central stenosis of the lumbar spine and no difference was found in quality of life (EQ-5D) one year after surgery.
Abstract: Objective To test the equivalence for clinical effectiveness between microdecompression and laminectomy in patients with central lumbar spinal stenosis. Design Multicentre observational study. Setting Prospective data from the Norwegian Registry for Spine Surgery. Participants 885 patients with central stenosis of the lumbar spine who underwent surgery at 34 Norwegian orthopaedic or neurosurgical departments. Patients were treated from October 2006 to December 2011. Interventions Laminectomy and microdecompression. Main outcome measures The primary outcome was change in Oswestry disability index score one year after surgery. Secondary endpoints were quality of life (EuroQol EQ-5D), perioperative complications, and duration of surgical procedures and hospital stays. A blinded biostatistician performed predefined statistical analyses in unmatched and propensity matched cohorts. Results The study was powered to detect a difference between the groups of eight points on the Oswestry disability index at one year. 721 patients (81%) completed the one year follow-up. Equivalence between microdecompression and laminectomy was shown for the Oswestry disability index (difference 1.3 points, 95% confidence interval −1.36 to 3.92, P Conclusion At one year the effectiveness of microdecompression is equivalent to laminectomy in the surgical treatment of central stenosis of the lumbar spine. Favourable outcomes were observed at one year in both treatment groups. Trial registration ClinicalTrials.gov NCT02006901.
127 citations
TL;DR: The study sought to characterize caregiver burden over a 3‐year period and identify predictors of this burden and found that environmental and social factors play a role in this burden.
Abstract: Objectives Dementia, with its progressive cognitive and functional decline and associated neuropsychiatric symptoms, places a large burden on caregivers. While frequently studied, longitudinal findings about the overall trajectory of burden are mixed. The study sought to characterize caregiver burden over a 3-year period and identify predictors of this burden. Methods Seven-hundred-and-eighty-one patients with dementia were recruited from nine memory clinics around Australia. Measures of caregiver burden, cognition, function, and neuropsychiatric symptoms were completed with patients and their caregivers at regular intervals over a 3-year period. Patients' level of services and medication use were also recorded. Results Of the 720 patients with measures of caregiver burden at baseline, 47.4% of caregivers had clinically significant levels of burden. This proportion increased over time, with 56.8% affected at 3 years. Overall levels of burden increased for caregivers of patients without services, though did not change for caregivers of patients receiving services or residential care after controlling for other variables. Patient characteristics-including greater neuropsychiatric symptoms, lower functional ability, fewer medications, lack of driving ability-and female sex of caregivers were associated with greater burden. Conclusions High levels of caregiver burden are present in a large proportion of caregivers of people with dementia and this increases over time for those without services. Clinical characteristics of patients (including neuropsychiatric symptoms, function, overall health, driving status), level of services, and caregiver sex appear to be the best predictors of this burden. These characteristics may help identify caregivers at greater risk of burden to target for intervention.
89 citations
Cites methods from "Multiple imputation of missing valu..."
...We used linear mixed models to analyze the data; these have the advantage of being able to handle missing data, including attrition, without the need for imputation.(31) Based on previous research, we expected that a range of patient characteristics—such as higher levels of neuropsychiatric symptoms, greater cognitive and functional impairments, and poorer health—and background features—such as longer duration of symptoms, level of services, spousal relationships, and cohabitation—would predict caregiver burden....
[...]
...In this regard, a strength of our approach was the use of linear mixed models, which can provide estimates of trajectories of burden over the course of the disease and handle missing data in the analyses.(31) Nevertheless, significant variability in changes in burden over time was evident, indicating the need for further research to identify drivers of longitudinal trends....
[...]
TL;DR: Addition of a 48 week course of pegylated interferon to nucleos(t)ide analogue therapy in patients with HBeAg-negative chronic hepatitis B with undetectable HBV DNA for a least 1 year was poorly tolerated and did not result in a significant increase of HBsAg clearance.
80 citations
References
More filters
Book•
01 Jan 1987
TL;DR: This work states that maximum Likelihood for General Patterns of Missing Data: Introduction and Theory with Ignorable Nonresponse and large-Sample Inference Based on Maximum Likelihood Estimates is likely to be high.
Abstract: Preface.PART I: OVERVIEW AND BASIC APPROACHES.Introduction.Missing Data in Experiments.Complete-Case and Available-Case Analysis, Including Weighting Methods.Single Imputation Methods.Estimation of Imputation Uncertainty.PART II: LIKELIHOOD-BASED APPROACHES TO THE ANALYSIS OF MISSING DATA.Theory of Inference Based on the Likelihood Function.Methods Based on Factoring the Likelihood, Ignoring the Missing-Data Mechanism.Maximum Likelihood for General Patterns of Missing Data: Introduction and Theory with Ignorable Nonresponse.Large-Sample Inference Based on Maximum Likelihood Estimates.Bayes and Multiple Imputation.PART III: LIKELIHOOD-BASED APPROACHES TO THE ANALYSIS OF MISSING DATA: APPLICATIONS TO SOME COMMON MODELS.Multivariate Normal Examples, Ignoring the Missing-Data Mechanism.Models for Robust Estimation.Models for Partially Classified Contingency Tables, Ignoring the Missing-Data Mechanism.Mixed Normal and Nonnormal Data with Missing Values, Ignoring the Missing-Data Mechanism.Nonignorable Missing-Data Models.References.Author Index.Subject Index.
18,201 citations
"Multiple imputation of missing valu..." refers background or result in this paper
...In addition to this distinction related to the missing data pattern, there is classic categorization of missing data mechanisms that describes relationships among missing values and their dependency on observed and unobserved variables in the data set [4,6]....
[...]
...In addition to the fact that this conclusion was expected theoretically [6], it was also found by Peters et al....
[...]
Book•
01 Jan 1987
TL;DR: In this article, a survey of drinking behavior among men of retirement age was conducted and the results showed that the majority of the participants reported that they did not receive any benefits from the Social Security Administration.
Abstract: Tables and Figures. Glossary. 1. Introduction. 1.1 Overview. 1.2 Examples of Surveys with Nonresponse. 1.3 Properly Handling Nonresponse. 1.4 Single Imputation. 1.5 Multiple Imputation. 1.6 Numerical Example Using Multiple Imputation. 1.7 Guidance for the Reader. 2. Statistical Background. 2.1 Introduction. 2.2 Variables in the Finite Population. 2.3 Probability Distributions and Related Calculations. 2.4 Probability Specifications for Indicator Variables. 2.5 Probability Specifications for (X,Y). 2.6 Bayesian Inference for a Population Quality. 2.7 Interval Estimation. 2.8 Bayesian Procedures for Constructing Interval Estimates, Including Significance Levels and Point Estimates. 2.9 Evaluating the Performance of Procedures. 2.10 Similarity of Bayesian and Randomization--Based Inferences in Many Practical Cases. 3. Underlying Bayesian Theory. 3.1 Introduction and Summary of Repeated--Imputation Inferences. 3.2 Key Results for Analysis When the Multiple Imputations are Repeated Draws from the Posterior Distribution of the Missing Values. 3.3 Inference for Scalar Estimands from a Modest Number of Repeated Completed--Data Means and Variances. 3.4 Significance Levels for Multicomponent Estimands from a Modest Number of Repeated Completed--Data Means and Variance--Covariance Matrices. 3.5 Significance Levels from Repeated Completed--Data Significance Levels. 3.6 Relating the Completed--Data and Completed--Data Posterior Distributions When the Sampling Mechanism is Ignorable. 4. Randomization--Based Evaluations. 4.1 Introduction. 4.2 General Conditions for the Randomization--Validity of Infinite--m Repeated--Imputation Inferences. 4.3Examples of Proper and Improper Imputation Methods in a Simple Case with Ignorable Nonresponse. 4.4 Further Discussion of Proper Imputation Methods. 4.5 The Asymptotic Distibution of (Qm,Um,Bm) for Proper Imputation Methods. 4.6 Evaluations of Finite--m Inferences with Scalar Estimands. 4.7 Evaluation of Significance Levels from the Moment--Based Statistics Dm and Dm with Multicomponent Estimands. 4.8 Evaluation of Significance Levels Based on Repeated Significance Levels. 5. Procedures with Ignorable Nonresponse. 5.1 Introduction. 5.2 Creating Imputed Values under an Explicit Model. 5.3 Some Explicit Imputation Models with Univariate YI and Covariates. 5.4 Monotone Patterns of Missingness in Multivariate YI. 5.5 Missing Social Security Benefits in the Current Population Survey. 5.6 Beyond Monotone Missingness. 6. Procedures with Nonignorable Nonresponse. 6.1 Introduction. 6.2 Nonignorable Nonresponse with Univariate YI and No XI. 6.3 Formal Tasks with Nonignorable Nonresponse. 6.4 Illustrating Mixture Modeling Using Educational Testing Service Data. 6.5 Illustrating Selection Modeling Using CPS Data. 6.6 Extensions to Surveys with Follow--Ups. 6.7 Follow--Up Response in a Survey of Drinking Behavior Among Men of Retirement Age. References. Author Index. Subject Index. Appendix I. Report Written for the Social Security Administration in 1977. Appendix II. Report Written for the Census Bureau in 1983.
14,574 citations
7,643 citations
Book•
01 Aug 1997TL;DR: The Normal Model Methods for Categorical Data Loglinear Models Methods for Mixed Data and Inference by Data Augmentation Methods for Normal Data provide insights into the construction of categorical and mixed data models.
Abstract: Introduction Assumptions EM and Inference by Data Augmentation Methods for Normal Data More on the Normal Model Methods for Categorical Data Loglinear Models Methods for Mixed Data Further Topics Appendices References Index
6,704 citations
TL;DR: This work focuses on the development of Imputation Models for Social Security Benefit Reconciliation in the context of a Finite Population and examines the role of Bayesian and Randomization--Based Inferences in these models.
Abstract: Tables and Figures. Glossary. 1. Introduction. 1.1 Overview. 1.2 Examples of Surveys with Nonresponse. 1.3 Properly Handling Nonresponse. 1.4 Single Imputation. 1.5 Multiple Imputation. 1.6 Numerical Example Using Multiple Imputation. 1.7 Guidance for the Reader. 2. Statistical Background. 2.1 Introduction. 2.2 Variables in the Finite Population. 2.3 Probability Distributions and Related Calculations. 2.4 Probability Specifications for Indicator Variables. 2.5 Probability Specifications for (X,Y). 2.6 Bayesian Inference for a Population Quality. 2.7 Interval Estimation. 2.8 Bayesian Procedures for Constructing Interval Estimates, Including Significance Levels and Point Estimates. 2.9 Evaluating the Performance of Procedures. 2.10 Similarity of Bayesian and Randomization--Based Inferences in Many Practical Cases. 3. Underlying Bayesian Theory. 3.1 Introduction and Summary of Repeated--Imputation Inferences. 3.2 Key Results for Analysis When the Multiple Imputations are Repeated Draws from the Posterior Distribution of the Missing Values. 3.3 Inference for Scalar Estimands from a Modest Number of Repeated Completed--Data Means and Variances. 3.4 Significance Levels for Multicomponent Estimands from a Modest Number of Repeated Completed--Data Means and Variance--Covariance Matrices. 3.5 Significance Levels from Repeated Completed--Data Significance Levels. 3.6 Relating the Completed--Data and Completed--Data Posterior Distributions When the Sampling Mechanism is Ignorable. 4. Randomization--Based Evaluations. 4.1 Introduction. 4.2 General Conditions for the Randomization--Validity of Infinite--m Repeated--Imputation Inferences. 4.3Examples of Proper and Improper Imputation Methods in a Simple Case with Ignorable Nonresponse. 4.4 Further Discussion of Proper Imputation Methods. 4.5 The Asymptotic Distibution of (Qm,Um,Bm) for Proper Imputation Methods. 4.6 Evaluations of Finite--m Inferences with Scalar Estimands. 4.7 Evaluation of Significance Levels from the Moment--Based Statistics Dm and Dm with Multicomponent Estimands. 4.8 Evaluation of Significance Levels Based on Repeated Significance Levels. 5. Procedures with Ignorable Nonresponse. 5.1 Introduction. 5.2 Creating Imputed Values under an Explicit Model. 5.3 Some Explicit Imputation Models with Univariate YI and Covariates. 5.4 Monotone Patterns of Missingness in Multivariate YI. 5.5 Missing Social Security Benefits in the Current Population Survey. 5.6 Beyond Monotone Missingness. 6. Procedures with Nonignorable Nonresponse. 6.1 Introduction. 6.2 Nonignorable Nonresponse with Univariate YI and No XI. 6.3 Formal Tasks with Nonignorable Nonresponse. 6.4 Illustrating Mixture Modeling Using Educational Testing Service Data. 6.5 Illustrating Selection Modeling Using CPS Data. 6.6 Extensions to Surveys with Follow--Ups. 6.7 Follow--Up Response in a Survey of Drinking Behavior Among Men of Retirement Age. References. Author Index. Subject Index. Appendix I. Report Written for the Social Security Administration in 1977. Appendix II. Report Written for the Census Bureau in 1983.
5,436 citations
"Multiple imputation of missing valu..." refers background or methods in this paper
...In addition to this distinction related to the missing data pattern, there is classic categorization of missing data mechanisms that describes relationships among missing values and their dependency on observed and unobserved variables in the data set [4,6]....
[...]
...The (un)stability of the multiple-imputation method is in contrast with most basic multiple-imputation literature [3,4], which suggests that five imputations should be sufficient to obtain valid inference....
[...]
...Within a longitudinal framework, different patterns of missing data can be distinguished: intermittent missing data (also known as nonmonotone missing data) and missing data resulting from dropout (also known as monotone missing data) [4,5]....
[...]