Comparing single and multiple imputation strategies for harmonizing substance use data across HIV-related cohort studies

Estimating the Prevalence of Injection Drug Use Among Acute Hepatitis C Cases From a National Surveillance System: Application of Random Forest-Based Multiple Imputation

BACKGROUND

Injection drug use (IDU) is a major contributor to the syndemic of viral hepatitis, human immunodeficiency virus, and drug overdose. However, information on IDU is frequently missing in national viral hepatitis surveillance data, which limits our understanding of the full extent of IDU-associated infections. Multiple imputation by chained equations (MICE) has become a popular approach to address missing data, but its application for IDU imputation is less studied.

METHODS

Using the 2019-2021 National Notifiable Diseases Surveillance System acute hepatitis C case data and publicly available county-level measures, we evaluated listwise deletion (LD) and 3 models imputing missing IDU data through MICE: parametric logistic regression, semi-parametric predictive mean matching (PMM), and nonparametric random forest (RF) (both standard RF [sRF] and fast implementation of RF [fRF]).

RESULTS

The estimated IDU prevalence among acute hepatitis C cases increased from 63.5% by LD to 65.1% by logistic regression, 66.9% by PMM, 76.0% by sRF, and 85.1% by fRF. Evaluation studies showed that RF-based MICE imputation, especially fRF, has the highest accuracy (as measured by smallest raw bias, percent bias, and root mean square error) and highest efficiency (as measured by smallest 95% confidence interval width) compared to LD and other models. Sensitivity analyses indicated that fRF remained robust when data were missing not at random.

CONCLUSION

Our analysis suggested that RF-based MICE imputation, especially fRF, could be a valuable approach for addressing missing IDU data in the context of population-based surveillance systems like National Notifiable Diseases Surveillance System. The inclusion of imputed IDU data may enhance the effectiveness of future surveillance and prevention efforts for the IDU-driven syndemic.

Narcissism Moderates the Association Between Autonomy-Supportive Parenting and Adolescents' Prosocial Behavior

Prior research has separately investigated the associations of autonomy-supportive parenting and narcissism with adolescents' prosocial behavior, but their joint relationships with prosocial behavior have been rarely examined. The present research aimed to expand the existing literature by scrutinizing the main and interactive associations of autonomy-supportive parenting and narcissism with adolescents' prosocial behavior. In so doing, a series of four studies (collectively N = 2023), combining cross-sectional, longitudinal, and experimental designs, were conducted. The adolescents' mean age varied from 12.42 to 15.70 years, with a balanced representation of the sexes in those studies. Converging results across four studies showed that high narcissism magnified the positive association between autonomy-supportive parenting and adolescents' prosocial behavior. The interaction pattern presented also suggested adolescents with high narcissism scores were more affected than others-both for better and for worse-by autonomy-supportive parenting, although this interaction might be specific to particular facets of prosocial behavior. These results were robust after adjusting for a few key covariates and survived a set of additional analyses. The present findings provide a novel avenue to explain individual differences linking prosocial behavior with those two factors and further advance precise, individualized strategies to promote adolescents' prosocial behavior.

Studying missingness in spinal cord injury data: challenges and impact of data imputation

BACKGROUND

In the last decades, medical research fields studying rare conditions such as spinal cord injury (SCI) have made extensive efforts to collect large-scale data. However, most analysis methods rely on complete data. This is particularly troublesome when studying clinical data as they are prone to missingness. Often, researchers mitigate this problem by removing patients with missing data from the analyses. Less commonly, imputation methods to infer likely values are applied.

OBJECTIVE

Our objective was to study how handling missing data influences the results reported, taking the example of SCI registries. We aimed to raise awareness on the effects of missing data and provide guidelines to be applied for future research projects, in SCI research and beyond.

METHODS

Using the Sygen clinical trial data (n = 797), we analyzed the impact of the type of variable in which data is missing, the pattern according to which data is missing, and the imputation strategy (e.g. mean imputation, last observation carried forward, multiple imputation).

RESULTS

Our simulations show that mean imputation may lead to results strongly deviating from the underlying expected results. For repeated measures missing at late stages (> = 6 months after injury in this simulation study), carrying the last observation forward seems the preferable option for the imputation. This simulation study could show that a one-size-fit-all imputation strategy falls short in SCI data sets.

CONCLUSIONS

Data-tailored imputation strategies are required (e.g., characterisation of the missingness pattern, last observation carried forward for repeated measures evolving to a plateau over time). Therefore, systematically reporting the extent, kind and decisions made regarding missing data will be essential to improve the interpretation, transparency, and reproducibility of the research presented.

Risk of non-fatal suicide attempt in individuals with substance use disorder: the roles of aggregate genetic liability and environmental exposures in a Swedish population-based cohort

BACKGROUND AND AIMS

Substance use disorder (SUD) is related to widespread adverse consequences, including higher suicidality. Shared genetic liability has been demonstrated between SUD and suicidality. Here, we measured the factors that contribute to increased risk of non-fatal suicide attempt among individuals with SUD by focusing upon aggregate genetic risks and both childhood and past-year environmental factors.

DESIGN

Longitudinal study. Family genetic risk scores and environmental factors (childhood, aged from 0 to 15 years, and the year preceding SUD registration) were used to predict the relative risk of non-fatal suicide attempt using Cox proportional hazards models. Additional analyses employed a co-relative design, accounting for genetic factors and shared familial environment, to test for potential causality.

SETTING AND PARTICIPANTS

Based on longitudinal Swedish registry data, 228 617 individuals with SUD registrations from 1991 to 2015 were included.

MEASUREMENTS

SUD and suicide attempts were identified using medical records (International Classification of Diseases codes). SUD was also identified using pharmacy and criminal registries.

FINDINGS

In multivariable analyses that jointly accounted for all the selected potential predictors, individuals with SUD were at higher risk for non-fatal suicide attempt if they had experienced a parental death [hazard ratio (HR) = 1.58; 95% confidence interval (CI) = 1.30, 1.93], were female (HR = 1.53, 95% CI = 1.49, 1.57), had low educational attainment (HR = 1.50, 95% CI = 1.46, 1.55), received social welfare (HR = 1.21, 95% CI = 1.17, 1.25) or had lived in a non-intact family (HR = 1.11, 95% CI = 1.08, 1.14). In co-relative analyses, low education was supported as a possible causal factor for suicide attempt. Aggregate genetic risks interacted with low education and being raised in a non-intact family, with increased prevalence of suicide attempt in people with high genetic risks and unfavorable environmental exposures.

CONCLUSIONS

Aggregate genetic liability, childhood environmental experiences and specific socio-economic indicators are important risk factors for non-fatal suicide attempt among individuals with substance use disorder.