Multiple Imputation Approach Research Articles

The association between tobacco use and COVID-19 diagnoses in three Nordic countries: a pooled analysis.

Previous research has suggested an unexpected negative association between smoking and susceptibility to COVID-19. This study, drawing on population-based data from three Nordic countries-Sweden, Norway, and Finland-aims to investigate this association further, capitalizing on diversity introduced by different containment measures. The objective of this research was to examine the association between cigarette smoking and snus (smokeless tobacco) use and the risk of confirmed COVID-19 infection. A pooled analysis integrating original data from 547,685 participants across three countries. We used a multiple imputation approach based on conditional probabilities to impute the systematically missing covariates. The associations between tobacco use and COVID-19 infection were assessed, controlling for potential confounding factors. Current cigarette smokers had a lower risk of a confirmed COVID-19 case, whereas there was an increased risk among snus users. Our sensitivity analysis confirmed that the associations between tobacco use and COVID-19 infection risk are robust, remaining consistent regardless of whether covariate imputation was applied. Findings support a negative association between smoking and SARS-CoV-2 infection, but not the hypothesis that nicotine may be protective against the risk of contracting SARS-CoV-2 infection.

A hybridization of multiple imputation and one-class bagging ensemble approach for missing value and class imbalance problem

A hybridization of multiple imputation and one-class bagging ensemble approach for missing value and class imbalance problem

Assessing the impact of missing data in youth overweight and obesity research: complete case analysis versus multiple imputation

ABSTRACT Youth overweight and obesity (OWOB) surveillance often uses body mass index (BMI) derived from self-reported height and weight, but these measures can suffer from high proportions of missing data. Complete case analysis (CCA) is the most common approach to handle missing data, but this approach can introduce bias if missing data are not missing completely at random. Using BMI and related covariate data from 36,546 female and 37,126 male youth aged 12–19 years who participated in the COMPASS study in 2018/19, where approximately 30% of BMI data were missing, results and inference were compared between CCA and multiple imputation (MI) approaches to examine associations with youth BMI. Results of regression joint models showed contrasting findings between MI and CCA, highlighting that appropriate methodological choices in the handling of missing data are essential in youth OWOB research and that choices can impact research inference and thereby associated policy and programming recommendations.

The Effects of Missing Data Handling Methods on Reliability Coefficients: A Monte Carlo Simulation Study

This study holds significant implications as it examines the impact of different missing data handling methods on the internal consistency coefficients. With Monte Carlo simulations, we manipulated number of items, true reliability, sample size, missing data ratio and mechanisms, to compare relative bias of reliability coefficients. The reliability coefficients under scrutiny in this study encompass Cronbach's Alpha, Heise & Bohrnsted's Omega, Hancock & Mueller's H, Gölbaşı-Şimşek & Noyan's Theta G, Armor's Theta, and Gilmer-Feldt coefficients. Our arsenal of techniques includes single imputation methods like zero, mean, median, and regression imputation, as well as multiple imputation approaches like expectation maximization and random forest. We also employ the classic deletion method known as listwise deletion. The findings suggest that, for missing completely at random (MCAR) or missing at random (MAR) data, single imputation approaches (excluding zero imputation) may still be preferable to expectation maximization and random forest imputation, thereby underscoring the importance of our research.

Filling the gaps: A multiple imputation approach to estimating aging curves in baseball

In sports, an aging curve depicts the relationship between average performance and age in athletes’ careers. This paper investigates the aging curves for offensive players in Major League Baseball. We study this problem in a missing data context and account for different types of dropouts of baseball players during their careers. We employ a multiple imputation framework for multilevel data to impute the player performance associated with the missing seasons, and estimate the aging curves based on the imputed datasets. We then evaluate the effects of different dropout mechanisms on the aging curves through simulation, before applying our method to analyze MLB player data from past seasons. Results suggest an overestimation of the aging curves constructed without considering the unobserved seasons, whereas estimates obtained from multiple imputation address this shortcoming.

Evaluation of data imputation approaches for multi-stream building systems data1

Increasing advancements in building digitization, smart sensing, and metering technologies have allowed large amounts of timeseries data to be collected for monitoring, analyzing, and controlling building systems. However, due to sensor or communication failures, the data collected are often incomplete and poor in quality. Data imputation approaches to replace the missing values, specifically based on either statistical or predictive models have been widely adopted for multivariate datasets in other domains. It is hence of interest to find an effective way to impute timeseries data collected from a building system. In this paper, we evaluate multiple data imputation approaches using data collected from a medium sized building situated in Stockholm, Sweden and a small commercial building from the ASHRAE RP-1312 research project. Sensors with widely varying characteristics from the case study buildings were selected to evaluate the imputation methods. The imputation accuracy and the impact of each chosen imputation method on information entropy, short-term building forecasting model performance, and fault detection strategy were evaluated. Results demonstrate that incorporating time-lagged cross correlations within a k -nearest neighbor ( k NN) model provide the most accurate imputations without affecting the quality of subsequent data analysis.

On the Use of Multiple Imputation Approach in Pediatric Research

On the Use of Multiple Imputation Approach in Pediatric Research

Recoverability and estimation of causal effects under typical multivariable missingness mechanisms.

In the context of missing data, the identifiability or "recoverability" of the average causal effect (ACE) depends not only on the usual causal assumptions but also on missingness assumptions that can be depicted by adding variable-specific missingness indicators to causal diagrams, creating missingness directed acyclic graphs (m-DAGs). Previous research described canonical m-DAGs, representing typical multivariable missingness mechanisms in epidemiological studies, and examined mathematically the recoverability of the ACE in each case. However, this work assumed no effect modification and did not investigate methods for estimation across such scenarios. Here, we extend this research by determining the recoverability of the ACE in settings with effect modification and conducting a simulation study to evaluate the performance of widely used missing data methods when estimating the ACE using correctly specified g-computation. Methods assessed were complete case analysis (CCA) and various implementations of multiple imputation (MI) with varying degrees of compatibility with the outcome model used in g-computation. Simulations were based on an example from the Victorian Adolescent Health Cohort Study (VAHCS), where interest was in estimating the ACE of adolescent cannabis use on mental health in young adulthood. We found that the ACE is recoverable when no incomplete variable (exposure, outcome, or confounder) causes its own missingness, and nonrecoverable otherwise, in simplified versions of 10 canonical m-DAGs that excluded unmeasured common causes of missingness indicators. Despite this nonrecoverability, simulations showed that MI approaches that are compatible with the outcome model in g-computation may enable approximately unbiased estimation across all canonical m-DAGs considered, except when the outcome causes its own missingness or causes the missingness of a variable that causes its own missingness. In the latter settings, researchers may need to consider sensitivity analysis methods incorporating external information (e.g., delta-adjustment methods). The VAHCS case study illustrates the practical implications of thesefindings.

Assessing efficacy in non-inferiority trials with non-adherence to interventions: Are intention-to-treat and per-protocol analyses fit for purpose?

Non-inferiority trials comparing different active drugs are often subject to treatment non-adherence. Intention-to-treat (ITT) and per-protocol (PP) analyses have been advocated in such studies but are not guaranteed to be unbiased in the presence of differential non-adherence. The REMoxTB trial evaluated two 4-month experimental regimens compared with a 6-month control regimen for newly diagnosed drug-susceptible TB. The primary endpoint was a composite unfavorable outcome of treatment failure or recurrence within 18 months post-randomization. We conducted a simulation study based on REMoxTB to assess the performance of statistical methods for handling non-adherence in non-inferiority trials, including: ITT and PP analyses, adjustment for observed adherence, multiple imputation (MI) of outcomes, inverse-probability-of-treatment weighting (IPTW), and a doubly-robust (DR) estimator. When non-adherence differed between trial arms, ITT, and PP analyses often resulted in non-trivial bias in the estimated treatment effect, which consequently under- or over-inflated the type I error rate. Adjustment for observed adherence led to similar issues, whereas the MI, IPTW and DR approaches were able to correct bias under most non-adherence scenarios; they could not always eliminate bias entirely in the presence of unobserved confounding. The IPTW and DR methods were generally unbiased and maintained desired type I error rates and statistical power. When non-adherence differs between trial arms, ITT and PP analyses can produce biased estimates of efficacy, potentially leading to the acceptance of inferior treatments or efficacious regimens being missed. IPTW and the DR estimator are relatively straightforward methods to supplement ITT and PP approaches.

An evaluation of the replicability of analyses using synthetic health data

Synthetic data generation is being increasingly used as a privacy preserving approach for sharing health data. In addition to protecting privacy, it is important to ensure that generated data has high utility. A common way to assess utility is the ability of synthetic data to replicate results from the real data. Replicability has been defined using two criteria: (a) replicate the results of the analyses on real data, and (b) ensure valid population inferences from the synthetic data. A simulation study using three heterogeneous real-world datasets evaluated the replicability of logistic regression workloads. Eight replicability metrics were evaluated: decision agreement, estimate agreement, standardized difference, confidence interval overlap, bias, confidence interval coverage, statistical power, and precision (empirical SE). The analysis of synthetic data used a multiple imputation approach whereby up to 20 datasets were generated and the fitted logistic regression models were combined using combining rules for fully synthetic datasets. The effects of synthetic data amplification were evaluated, and two types of generative models were used: sequential synthesis using boosted decision trees and a generative adversarial network (GAN). Privacy risk was evaluated using a membership disclosure metric. For sequential synthesis, adjusted model parameters after combining at least ten synthetic datasets gave high decision and estimate agreement, low standardized difference, as well as high confidence interval overlap, low bias, the confidence interval had nominal coverage, and power close to the nominal level. Amplification had only a marginal benefit. Confidence interval coverage from a single synthetic dataset without applying combining rules were erroneous, and statistical power, as expected, was artificially inflated when amplification was used. Sequential synthesis performed considerably better than the GAN across multiple datasets. Membership disclosure risk was low for all datasets and models. For replicable results, the statistical analysis of fully synthetic data should be based on at least ten generated datasets of the same size as the original whose analyses results are combined. Analysis results from synthetic data without applying combining rules can be misleading. Replicability results are dependent on the type of generative model used, with our study suggesting that sequential synthesis has good replicability characteristics for common health research workloads.

Better tools for better estimates: improving approaches to handling missing data in Swiss cancer registries.

Various approaches have been used in the literature to handle missing vital status data in cancer registries. We aimed to compare these approaches to determine which led to the least biased estimates in typical analytic tasks of cancer registries. A simulation study was performed using data from the Swiss National Agency for Cancer Registration for six tumor types. First, 5%, 10% and 15% missingness in the vital status were introduced artificially in the complete data. Second, missing vital status data were handled by applying no, single or multiple imputations. Five-year overall survival estimates, relative survival or standardized incidence ratio were computed. Estimates were compared with the true value. Standardized incidence ratio estimates for colorectal cancer obtained with multiple imputation yielded least biased results (-0.06 to -0.04), but the widest confidence intervals. Single imputation was more biased (-0.32) than using no imputation at all (-0.21). A similar pattern was observed for overall survival and relative survival. This simulation study indicated that often used single imputation (sometimes referred to as simulating follow-up times) techniques to fill in missing vital status data are likely too biased to be useful in practice. Multiple imputation approaches yielded standardized incidence ratio, overall and relative survival estimates with the least bias, indicating reasonable performance that is likely to generalize to other settings.

Multi-omics Integrative Analysis for Incomplete Data Using Weighted p-Value Adjustment Approaches

AbstractThe advancements in high-throughput technologies provide exciting opportunities to obtain multi-omics data from the same individuals in a biomedical study, and joint analyses of data from multiple sources offer many benefits. However, the occurrence of missing values is an inevitable issue in multi-omics data because measurements such as mRNA gene expression levels often require invasive tissue sampling from patients. Common approaches for addressing missing measurements include analyses based on observations with complete data or multiple imputation methods. In this paper, we propose a novel integrative multi-omics analytical framework based on p-value weight adjustment in order to incorporate observations with incomplete data into the analysis. By splitting the data into a complete set with full information and an incomplete set with missing measurements, we introduce mechanisms to derive weights and weight-adjusted p-values from the two sets. Through simulation analyses, we demonstrate that the proposed framework achieves considerable statistical power gains compared to a complete case analysis or multiple imputation approaches. We illustrate the implementation of our proposed framework in a study of preterm infant birth weights by a joint analysis of DNA methylation, mRNA, and the phenotypic outcome. Supplementary materials accompanying this paper appear online.

Handling missing data when estimating causal effects with targeted maximum likelihood estimation.

Targeted maximum likelihood estimation (TMLE) is increasingly used for doubly robust causal inference, but how missing data should be handled when using TMLE with data-adaptive approaches is unclear. Based on data (1992-1998) from the Victorian Adolescent Health Cohort Study, we conducted a simulation study to evaluate 8 missing-data methods in this context: complete-case analysis, extended TMLE incorporating an outcome-missingness model, the missing covariate missing indicator method, and 5 multiple imputation (MI) approaches using parametric or machine-learning models. We considered 6 scenarios that varied in terms of exposure/outcome generation models (presence of confounder-confounder interactions) and missingness mechanisms (whether outcome influenced missingness in other variables and presence of interaction/nonlinear terms in missingness models). Complete-case analysis and extended TMLE had small biases when outcome did not influence missingness in other variables. Parametric MI without interactions had large bias when exposure/outcome generation models included interactions. Parametric MI including interactions performed best in bias and variance reduction across all settings, except when missingness models included a nonlinear term. When choosing a method for handling missing data in the context of TMLE, researchers must consider the missingness mechanism and, for MI, compatibility with the analysis method. In many settings, a parametric MI approach that incorporates interactions and nonlinearities is expected to perform well.

Comparison of common multiple imputation approaches: An application of logistic regression with an interaction

Background Multiple imputation is often used to reduce bias and gain efficiency when there is missing data. The most appropriate imputation method depends on the model the analyst is interested in fitting. We consolidate and compare the performance and ease of use for several commonly implemented imputation approaches. Methods Using 1000 simulations, each with 10,000 observations, under six data-generating mechanisms (DGM), we investigate the performance of four methods: (i) ’passive imputation’, (ii) ’just another variable’ (JAV), (iii) ’stratify-impute-append’ (SIA), and (iv) ’substantive model compatible fully conditional specification’ (SMCFCS). The application of each method is shown in an empirical example using England-based cancer registry data. Results SMCFCS and SIA showed the least biased estimate of the coefficients for the fully, and partially, observed variable and the interaction term. SMCFCS and SIA showed good coverage and low relative error for all DGMs. SMCFCS had a large bias when there was a low prevalence of the fully observed variable in the interaction. SIA performed poorly when the fully observed variable in the interaction had a continuous underlying form. Conclusion SMCFCS and SIA give consistent estimation and either can be used in most analyses. SMCFCS performed better than SIA when the fully observed variable in the interaction had an underlying continuous form. Researchers should be cautious when using SMCFCS when there is a low prevalence of the fully observed variable in the interaction.

Impact of cash transfers on the association between prenatal exposures to high temperatures and low birthweight: Retrospective analysis from the LEAP 1000 study.

To explore the associations between prenatal temperature exposures and low birthweight (LBW) and modification by cash transfer (CT) receipt. Retrospective cohort study. Five rural districts in Northern Ghana. A total of 3016 infants born to women interviewed as part of the Livelihood Empowerment Against Poverty (LEAP 1000) impact evaluation between 2015 and 2017. Birthweight was collected using household surveys administered to LEAP 1000 eligible women. We used a UNICEF-developed multiple imputation approach to address missingness of birthweight and applied an empirical heaping correction to the multiply imputed birthweight data. Survey data were linked to the European Centre for Medium-Range Weather Forecasts Reanalysis 5-hourly temperature averaged to weeks for 2011-2017 using community centroids. Using distributed-lag nonlinear models, we explored the lag-specific associations between weekly average temperatures greater than 30°C and LBW, and stratified by LEAP 1000 treatment. Low birthweight (<2.5 kg). Twelve percent (n = 365) of infants were LBW; the mean ± SD birthweight was 3.02 ± 0.37 kg. Overall, increasing temperatures were associated with increased odds of LBW, with the greatest odds observed in the 3 weeks before birth (odds ratio 1.005-1.025). These positive associations were even larger among comparison infants and null among treatment infants. Our study found increased odds of LBW with high weekly average temperatures throughout pregnancy and the preconception period and demonstrate mitigated effects by the LEAP 1000 CT program. More evidence on the potential of CTs to serve as adaptation interventions in low- and middle-income countries is needed to protect pregnant persons and their infantsfrom the impacts of climate change.

Multiple imputation of missing data in multilevel ecological momentary assessments: an example using smoking cessation study data.

Advances in digital technology have greatly increased the ease of collecting intensive longitudinal data (ILD) such as ecological momentary assessments (EMAs) in studies of behavior changes. Such data are typically multilevel (e.g., with repeated measures nested within individuals), and are inevitably characterized by some degrees of missingness. Previous studies have validated the utility of multiple imputation as a way to handle missing observations in ILD when the imputation model is properly specified to reflect time dependencies. In this study, we illustrate the importance of proper accommodation of multilevel ILD structures in performing multiple imputations, and compare the performance of a multilevel multiple imputation (multilevel MI) approach relative to other approaches that do not account for such structures in a Monte Carlo simulation study. Empirical EMA data from a tobacco cessation study are used to demonstrate the utility of the multilevel MI approach, and the implications of separating participant- and study-initiated EMAs in evaluating individuals' affective dynamics and urge.

Multiple Imputation Through XGBoost

The use of multiple imputation (MI) is becoming increasingly popular for addressing missing data. Although some conventional MI approaches have been well studied and have shown empirical validity, they have limitations when processing large datasets with complex data structures. Their imputation performances usually rely on the proper specification of imputation models, and this requires expert knowledge of the inherent relations among variables. Moreover, these standard approaches tend to be computationally inefficient for medium and large datasets. In this article, we propose a scalable MI framework mixgb, which is based on XGBoost, subsampling, and predictive mean matching. Our approach leverages the power of XGBoost, a fast implementation of gradient boosted trees, to automatically capture interactions and nonlinear relations while achieving high computational efficiency. In addition, we incorporate subsampling and predictive mean matching to reduce bias and to better account for appropriate imputation variability. The proposed framework is implemented in an R package mixgb. Supplementary materials for this article are available online.

Mind–body and art therapies impact on emotional regulation in patients with chronic diseases: a pragmatic mixed-methods randomized controlled trial

BackgroundEffective emotional regulation is recognized as essential to a good mental health of people with chronic diseases, and Mind–body and Art Therapies (MBATs) could have a positive effect on emotional regulation skills in this population. Thus, we aimed to evaluate the effect of MBATs on emotional regulation as measured by the Difficulties in Emotion Regulation Scale (DERS) questionnaire.MethodsA convergent mixed approach nested in a pragmatic superiority two arms parallel randomized controlled trial was conducted. French speaking adults with one or more chronic somatic illnesses and not suffering from a chronic psychiatric disorder unrelated to one of their chronic somatic illness were included. At inclusion, non-directive interviews were conducted, followed by an initial DERS assessment. The same combination of evaluation was implemented after 6 months of activity (T1). After inclusion, each participant was randomized within either the intervention group (G1) or the control group (G2) following a controlled wait-list design by use of a pregenerated randomization list. Staff and patient were blinded to this list until the initial evaluation was completed, after which the trial was conducted in an open-label fashion. Participants chose 2 mediations: one creativity-focused (art-therapy, writing workshop, theatre of life, vocal workshop) and one mind–body-focused (mindfulness meditation, Pilates, shiatsu, ayurvedic massages). G1 started their mediations immediately after inclusion, while G2 started 6 months later. Primary outcome was the change in means at 6 months in the overall DERS score compared between each group. Non-directive interviews were carried out at the inclusion and after 6 months of MBATs. A continuous inductive analysis was carried out on gathered material in G1 to explore the participants' experiences regarding their disease and their perceived changes associated to the intervention.ResultsA total of 150 patients was randomized (75 per groups) at the end of the study. At T1, 133 patients filled out the final questionnaire (67 in G1 vs 66 in G2) and 112 interviews were analysed (54 in G1 vs 58 in G2). All 150 patients were analysed (intention to treat) using a multiple imputation approach. The mean DERS score at T0 was equal to 82.8 ± 21.1 and 85.0 ± 20.2 in G1 and G2 respectively. On average, at T1, the score decreased in the G1 (Δ = -4.8, SD = 21.3) and in G2 (Δ = -0.11, SD = 17.8). The difference in decrease, however, was not statistically significant (p = 0.13). Qualitative analysis underlined some MBATs benefits on emotional regulation, especially on regulation strategies. No harms related to the intervention has been observed.ConclusionsThis study only partially supports benefits on MBAT on emotional regulation skills enhancement in patients with chronic disease receiving MBATs, as measured by the DERS scale.Trial registrationThe protocol was registered on Clinical Trials (NCT02911207).

A multiple imputation approach for the Cox–Aalen cure model with interval-censored data

A multiple imputation approach for the Cox–Aalen cure model with interval-censored data

Multiple imputation for non-monotone missing not at random data using the no self-censoring model.

Although approaches for handling missing data from longitudinal studies are well-developed when the patterns of missingness are monotone, fewer methods are available for non-monotone missingness. Moreover, the conventional missing at random assumption-a natural benchmark for monotone missingness-does not model realistic beliefs about the non-monotone missingness processes (Robins and Gill, 1997). This has provided the impetus for alternative non-monotone missing not at random mechanisms. The "no self-censoring" model is such a mechanism and assumes the probability an outcome variable is missing is independent of its value when conditioning on all other possibly missing outcome variables and their missingness indicators. As an alternative to "weighting" methods that become computationally demanding with increasing number of outcome variables, we propose a multiple imputation approach under no self-censoring. We focus on the case of binary outcomes and present results of simulation and asymptotic studies to investigate the performance of the proposed imputation approach. We describe a related approach to sensitivity analysis to departure from no self-censoring. We discuss the relationship between missing at random and no self-censoring and prove that one is not a special case of the other. Finally, we discuss extensions to non-binary data settings. The proposed methods are illustrated with application to a substance use disorder clinical trial.