Practical strategies for handling breakdown of multiple imputation procedures

Perceived Neighbourhood Disorder, Alcohol Consumption and Alcohol-Related Problems in Chile

Background: Alcohol misuse is one of the most important preventable public health risk factors. Empirical research shows that alcohol misuse is related to social and economic losses. Both theoretical and empirical evidence suggests that neighborhood disorder impacts alcohol-related behavior. However, there is limited literature in the context of developing countries. Objectives: The aim of this research is to estimate the association between perceived neighborhood disorder and (1) alcohol-related behavior and (2) alcohol-related problems in the context of the Chilean population. Our contribution focuses on the examination of the perception of disorder in urban neighborhoods and alcohol use patterns in a wide age range and sample of Chilean cities. Results: High levels of neighbor disorder perception are associated with higher levels of drinking and hazardous alcohol use. In addition, perceived neighborhood disorder is directly associated with probability of alcohol-related problems (ranging from 2% to 11%). Conclusions/Importance: The results are consistent with empirical and theoretical frameworks. This research could be used to better guide place-based policies in emerging countries with high levels of alcohol consumption to prevent alcohol risk behaviors and alcohol-related problems.

A comparison of strategies for selecting auxiliary variables for multiple imputation

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 is not always straightforward. Several data-driven auxiliary variable selection strategies have been proposed, but there has been limited evaluation of their performance. Using a simulation study we evaluated the performance of eight auxiliary variable selection strategies: (1, 2) two versions of selection based on correlations in the observed data; (3) selection using hypothesis tests of the "missing completely at random" assumption; (4) replacing auxiliary variables with their principal components; (5, 6) forward and forward stepwise selection; (7) forward selection based on the estimated fraction of missing information; and (8) selection via the least absolute shrinkage and selection operator (LASSO). A complete case analysis and an MI analysis using all auxiliary variables (the "full model") were included for comparison. We also applied all strategies to a motivating case study. The full model outperformed all auxiliary variable selection strategies in the simulation study, with the LASSO strategy the best performing auxiliary variable selection strategy overall. All MI analysis strategies that we were able to apply to the case study led to similar estimates, although computational time was substantially reduced when variable selection was employed. This study provides further support for adopting an inclusive auxiliary variable strategy where possible. Auxiliary variable selection using the LASSO may be a promising alternative when the full model fails or is too burdensome.

Epilepsy and long-term risk of arrhythmias

BACKGROUND AND AIMS

Previous evidence has mainly supported transient changes in cardiac function during interictal or peri-ictal phases in people with epilepsy, but the long-term risk of cardiac arrhythmias is poorly described. This study aimed to assess the long-term association of epilepsy with cardiac arrhythmias, considering the potential role of genetic predisposition and antiseizure medications (ASMs) in any associations observed.

METHODS

This population-based study evaluated UK Biobank data for individuals recruited between 2006 and 2010. Cox proportional hazards models and competing risk models were used to examine the association of epilepsy history with the long-term incidence risk of cardiac arrhythmias and arrhythmias subtypes. Polygenic risk scores (PRS) were calculated to investigate the effect of genetic susceptibility. The role of ASMs was also evaluated by integrating observational and drug target Mendelian randomization (MR) evidence.

RESULTS

The study included 329 432 individuals, including 2699 people with epilepsy. Compared with those without epilepsy, people with epilepsy experienced an increased risk of all cardiac arrhythmias [hazard ratio (HR) 1.36, 95% confidence interval (CI) 1.21-1.53], atrial fibrillation (HR 1.26, 95% CI 1.08-1.46), and other cardiac arrhythmias (HR 1.56, 95% CI 1.34-1.81). The associations were not modified by genetic predisposition as indicated by PRS. Competing and sensitivity analyses corroborated these results. Individuals with epilepsy using ASMs, especially carbamazepine and valproic acid, were at a higher risk for cardiac arrhythmias. This observation was further supported by drug target MR results (PSMR < .05 and PHEIDI > .05).

CONCLUSION

This study revealed the higher risk of cardiac arrhythmias persists long term in people with epilepsy, especially among those using carbamazepine and valproic acid. These findings highlight the need for regular heart rhythm monitoring and management in people with epilepsy in order to reduce the risk of further cardiovascular complications.

Financial and Other Life Stressors, Psychological Distress, and Food and Beverage Consumption among Students Attending a Large California State University during the COVID-19 Pandemic

The novel coronavirus disease 2019 (COVID-19) pandemic abruptly disrupted the daily lives and health of college students across the United States. This study investigated several stressors (e.g., financial strain/uncertainty), psychological distress, and dietary behaviors among college students attending a large state university during the pandemic. A cross-sectional online survey was administered to students from the California State University, Los Angeles between April and May 2021 (final analytic sample n = 736). Differences in gender and race/ethnicity were examined using chi-square, t-test, and one-way ANOVA tests. Paired t-tests were performed to compare variables before and during the pandemic. Negative binomial regression models examined the associations between various stressors, psychological distress, and three key dietary outcomes. Descriptive results showed that the consumption of fruits and vegetables, fast food, and sugary beverages, along with psychological distress, all increased during the pandemic. Significant differences in fruit and vegetable and fast food consumption by gender and race/ethnicity were also observed. In the regression models, several stressors, including financial strain and psychological distress, were associated with unfavorable food and beverage consumption, thereby suggesting that college students may need more support in mitigating these stressors so they do not manifest as poor dietary behaviors. Poor diet quality is associated with poor physical health outcomes such as premature development of type 2 diabetes or hypertension.

Methods for handling missing data in serially sampled sputum specimens for mycobacterial culture conversion calculation

BACKGROUND

The occurrence and timing of mycobacterial culture conversion is used as a proxy for tuberculosis treatment response. When researchers serially sample sputum during tuberculosis studies, contamination or missed visits leads to missing data points. Traditionally, this is managed by ignoring missing data or simple carry-forward techniques. Statistically advanced multiple imputation methods potentially decrease bias and retain sample size and statistical power.

METHODS

We analyzed data from 261 participants who provided weekly sputa for the first 12 weeks of tuberculosis treatment. We compared methods for handling missing data points in a longitudinal study with a time-to-event outcome. Our primary outcome was time to culture conversion, defined as two consecutive weeks with no Mycobacterium tuberculosis growth. Methods used to address missing data included: 1) available case analysis, 2) last observation carried forward, and 3) multiple imputation by fully conditional specification. For each method, we calculated the proportion culture converted and used survival analysis to estimate Kaplan-Meier curves, hazard ratios, and restricted mean survival times. We compared methods based on point estimates, confidence intervals, and conclusions to specific research questions.

RESULTS

The three missing data methods lead to differences in the number of participants achieving conversion; 78 (32.8%) participants converted with available case analysis, 154 (64.7%) converted with last observation carried forward, and 184 (77.1%) converted with multiple imputation. Multiple imputation resulted in smaller point estimates than simple approaches with narrower confidence intervals. The adjusted hazard ratio for smear negative participants was 3.4 (95% CI 2.3, 5.1) using multiple imputation compared to 5.2 (95% CI 3.1, 8.7) using last observation carried forward and 5.0 (95% CI 2.4, 10.6) using available case analysis.

CONCLUSION

We showed that accounting for missing sputum data through multiple imputation, a statistically valid approach under certain conditions, can lead to different conclusions than naïve methods. Careful consideration for how to handle missing data must be taken and be pre-specified prior to analysis. We used data from a TB study to demonstrate these concepts, however, the methods we described are broadly applicable to longitudinal missing data. We provide valuable statistical guidance and code for researchers to appropriately handle missing data in longitudinal studies.

Assessing Alternative Imputation Strategies for Infrequently Missing Items on Multi-item Scales

Health-science researchers often measure psychological constructs using multi-item scales and encounter missing items on some participants. Multiple imputation (MI) has emerged as an alternative to ad-hoc methods (e.g., mean substitution) for handling incomplete data on multi-item scales, appealingly reflecting available information while accounting for uncertainty due to missing values in a unified inferential framework. However, MI can be implemented in a variety of ways. When the number of variables to impute gets large, some strategies yield unstable estimates of quantities of interest while others are not technically feasible to implement. These considerations raise pragmatic questions about the extent to which ad-hoc procedures would yield statistical properties that are competitive with theoretically motivated methods. Drawing on an HIV study where depression and anxiety symptoms are measured with multi-item scales, this empirical investigation contrasts ad-hoc methods for handling missing items with various MI implementations that differ as to whether imputation is at the item-level or scale-level and how auxiliary variables are incorporated. While the findings are consistent with previous reports favoring item-level imputation when feasible to implement, we found only subtle differences in statistical properties across procedures, suggesting that weaknesses of ad-hoc procedures may be muted when missing data percentages are modest.

What difference does multiple imputation make in longitudinal modeling of EQ-5D-5L data? Empirical analyses of simulated and observed missing data patterns

PURPOSE

Although multiple imputation is the state-of-the-art method for managing missing data, mixed models without multiple imputation may be equally valid for longitudinal data. Additionally, it is not clear whether missing values in multi-item instruments should be imputed at item or score-level. We therefore explored the differences in analyzing the scores of a health-related quality of life questionnaire (EQ-5D-5L) using four approaches in two empirical datasets.

METHODS

We used simulated (GR dataset) and observed missingness patterns (ABCD dataset) in EQ-5D-5L scores to investigate the following approaches: approach-1) mixed models using respondents with complete cases, approach-2) mixed models using all available data, approach-3) mixed models after multiple imputation of the EQ-5D-5L scores, and approach-4) mixed models after multiple imputation of EQ-5D 5L items.

RESULTS

Approach-1 yielded the highest estimates of all approaches (ABCD, GR), increasingly overestimating the EQ-5D-5L score with higher percentages of missing data (GR). Approach-4 produced the lowest scores at follow-up evaluations (ABCD, GR). Standard errors (0.006-0.008) and mean squared errors (0.032-0.035) increased with increasing percentages of simulated missing GR data. Approaches 2 and 3 showed similar results (both datasets).

CONCLUSION

Complete cases analyses overestimated the scores and mixed models after multiple imputation by items yielded the lowest scores. As there was no loss of accuracy, mixed models without multiple imputation, when baseline covariates are complete, might be the most parsimonious choice to deal with missing data. However, multiple imputation may be needed when baseline covariates are missing and/or more than two timepoints are considered.

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

Medical research often involves using multi-item scales to assess individual characteristics, disease severity, and other health-related outcomes. It is common to observe missing data in the scale scores, due to missing data in one or more items that make up that score. Multiple imputation (MI) is a popular method for handling missing data. However, it is not clear how best to use MI in the context of scale scores, particularly when they are assessed at multiple waves of data collection resulting in large numbers of items. The aim of this article is to provide practical advice on how to impute missing values in a repeatedly measured multi-item scale using MI when inference on the scale score is of interest. We evaluated the performance of five MI strategies for imputing missing data at either the item or scale level using simulated data and a case study based on four waves of the Longitudinal Study of Australian Children (LSAC). MI was implemented using both multivariate normal imputation and fully conditional specification, with two rules for calculating the scale score. A complete case analysis was also performed for comparison. Based on our results, we caution against the use of a MI strategy that does not include the scale score in the imputation model(s) when the scale score is required for analysis.