Estimating generalized propensity scores with survey and attrition weighted data.

Prior work in causal inference has shown that using survey sampling weights in the propensity score estimation stage and the outcome model stage for binary treatments can result in a more robust estimator of the effect of the binary treatment being analyzed. However, to date, extending this work to continuous treatments and exposures has not been explored nor has consideration been given for how to handle attrition weights in the propensity score model. Nonetheless, generalized propensity score (GPS) analyses are being used for estimating continuous treatment effects on outcomes when researchers have observational data, and those data sets often have survey or attrition weights that need to be accounted for in the analysis. Here, we extend prior work and show with analytic results that using survey sampling or attrition weights in the GPS estimation stage and the outcome model stage for continuous treatments can result in a more robust estimator than one that does not. Simulation study results show that, although using weights in both estimation stages is sufficient for robust estimation, it is not necessary and unbiased estimation is possible in some cases under various approaches to using weights in estimation. Analysts do not know if the conditions of our simulation studies hold, so use of weights in both estimation stages might provide insurance for reducing potential bias. We discuss the implications of our results in the context of an empirical example.

Sampling weighting strategies in causal mediation analysis.

BACKGROUND: Causal mediation analysis plays a crucial role in examining causal effects and causal mechanisms. Yet, limited work has taken into consideration the use of sampling weights in causal mediation analysis. In this study, we compared different strategies of incorporating sampling weights into causal mediation analysis. METHODS: We conducted a simulation study to assess 4 different sampling weighting strategies-1) not using sampling weights, 2) incorporating sampling weights into mediation "cross-world" weights, 3) using sampling weights when estimating the outcome model, and 4) using sampling weights in both stages. We generated 8 simulated population scenarios comprising an exposure (A), an outcome (Y), a mediator (M), and six covariates (C), all of which were binary. The data were generated so that the true model of A given C and the true model of A given M and C were both logit models. We crossed these 8 population scenarios with 4 different sampling methods to obtain 32 total simulation conditions. For each simulation condition, we assessed the performance of 4 sampling weighting strategies when calculating sample-based estimates of the total, direct, and indirect effects. We also applied the four sampling weighting strategies to a case study using data from the National Survey on Drug Use and Health (NSDUH). RESULTS: Using sampling weights in both stages (mediation weight estimation and outcome models) had the lowest bias under most simulation conditions examined. Using sampling weights in only one stage led to greater bias for multiple simulation conditions. DISCUSSION: Using sampling weights in both stages is an effective approach to reduce bias in causal mediation analyses under a variety of conditions regarding the structure of the population data and sampling methods.

Why Are Adolescent Cannabis Use Disorder Treatment Admissions Declining in the US? The Mediated Pathway of State Treatment Admissions Rates before and after Recreational Cannabis Legalization.

Purpose: State-level prevalence data are used to investigate whether recreational cannabis legalization moderates the mediated pathway from the perception of low risk of harm, to cannabis use, to cannabis use disorder (CUD) treatment admissions, among adolescents (age 12-17) in the US. Methods: Annual state prevalence measures of perception of low risk, cannabis use, and CUD treatment admissions between 2008 and 2019 (N = 542 state-year observations) were collected from the National Survey on Drug Use and Health (NSDUH) and the Treatment Episode Dataset - Admissions (TEDS-A). A two-way fixed effects (state and year) moderated mediation model was used to test whether recreational legalization moderated the indirect effect of perception of low risk on treatment admissions via cannabis use. Results: A positive indirect effect of perceiving cannabis as low risk on CUD treatment admissions via cannabis use was observed prior to legalization but not afterwards. After legalization, the positive association of perceiving cannabis as low risk with cannabis use was strengthened, and the positive association of cannabis use with treatment admissions was suppressed, as compared to before legalization. Discussion: Recreational legalization may alter the social acceptability and medical self-administration of cannabis, potentially leading to CUD treatment utilization decline among adolescents even as risk factors for CUD increase. Linking recreational cannabis legalization to advancing awareness of the health risks associated with adolescent cannabis use and promoting adolescent CUD treatment engagement through mHealth approaches and primary care providers are key to addressing potential adolescent health challenges brought about by expanding cannabis legalization.

A Tutorial for Propensity Score Weighting for Moderation Analysis With Categorical Variables: An Application Examining Smoking Disparities Among Sexual Minority Adults.

OBJECTIVE: To provide step-by-step guidance and STATA and R code for using propensity score (PS) weighting to estimate moderation effects with categorical variables. RESEARCH DESIGN: Tutorial illustrating the key steps for estimating and testing moderation using observational data. Steps include: (1) examining covariate overlap across treatment groups within levels of the moderator; (2) estimating the PS weights; (3) evaluating whether PS weights improved covariate balance; (4) estimating moderated treatment effects; and (5) assessing the sensitivity of findings to unobserved confounding. Our illustrative case study uses data from 41,832 adults from the 2019 National Survey on Drug Use and Health to examine if gender moderates the association between sexual minority status (eg, lesbian, gay, or bisexual [LGB] identity) and adult smoking prevalence. RESULTS: For our case study, there were no noted concerns about covariate overlap, and we were able to successfully estimate the PS weights within each level of the moderator. Moreover, balance criteria indicated that PS weights successfully achieved covariate balance for both moderator groups. PS-weighted results indicated there was significant evidence of moderation for the case study, and sensitivity analyses demonstrated that results were highly robust for one level of the moderator but not the other. CONCLUSIONS: When conducting moderation analyses, covariate imbalances across levels of the moderator can cause biased estimates. As demonstrated in this tutorial, PS weighting within each level of the moderator can improve the estimated moderation effects by minimizing bias from imbalance within the moderator subgroups.

Age-related differences in eating location, food source location, and timing of snack intake among U.S. children 1-19 years.

BACKGROUND: Snacking is nearly universal and contributes significant energy to U.S. children's diets. Little is known, however, about where and when snacking intake occurs and if such patterns change with age. This research evaluated age-related differences in eating location, food source location, and timing of snacking among U.S. children aged 1-19 years (y). METHODS: A cross-sectional analysis of snacking among 14,666 children in the 2007-2018 U.S. National Health and Nutrition Examination Survey was conducted using a single 24-hour dietary recall. Snacking was participant-defined and included all eating occasions outside of meals. Linear regression and analysis of covariance were used to examine the effects of age (toddler 1-2 y, preschooler 3-5 y, school-age 6-11 y, adolescent 12-19 y) on the percentage of daily snack energy consumed by eating location (at home vs. away from home), food source location (grocery store, convenience store, school/childcare, restaurant, from someone else (i.e. "socially"), and other), and time of day (morning, 6am-12pm; early afternoon, 12pm-3pm; late afternoon/afterschool 3pm-6pm; evening 6pm-9pm, late-night 9pm-12am, and overnight 12am-6am). RESULTS: On average, U.S. children consumed most of their daily snacking energy at home (71%), from foods and beverages obtained from grocery stores (75%), and in the late afternoon/afterschool (31%). Toddlers and preschoolers consumed a greater percentage of their daily snack energy during the morning hours compared to school-age children and adolescents (both p < 0.001); school-age children consumed the most in the evening (27%, p < 0.001), and adolescents consumed the most in the late-night period (22%, p < 0.001). Age-related increases were seen in the percentage of daily snacking energy eaten outside the home (p < 0.001), and obtained socially (p < 0.001), from restaurants (p < 0.001), and convenience stores (p < 0.001). CONCLUSION: Findings reveal age-related differences in eating location, food source location, and timing of snack intake among U.S. children aged 1-19 y. Younger children consume a greater percentage of snacking calories in the morning and at home relative to older children. School-age children and adolescents show greater snacking in the evening and at night and from foods obtained and eaten outside the home. Efforts to promote healthy snacking behaviors among children should consider developmental differences in snacking patterns.

The Impact of Smoking Status and Smoking-Related Comorbidities on Coronavirus Disease 2019 Patient Outcomes: A Causal Mediation Analysis.

INTRODUCTION: Smoking history is a known risk factor for significant chronic diseases as well as pulmonary infections; however, the impact of smoking status on coronavirus disease 2019 (COVID-19) outcomes has not been conclusively characterized. This study aims to evaluate the association of smoking status on COVID-19 outcomes, and to explore the mechanism by which smoking and smoking-related comorbidities relate to COVID-19 outcomes. AIMS AND METHODS: Patients admitted with SARS-CoV-2 infection from November 2020 through January 2021 were included in this study. Causal mediation models investigating the associations between smoking status and the outcomes of mortality, intensive care unit (ICU) admission, advanced respiratory support, mechanical ventilation, ICU length of stay, and hospital length of stay, through mediation via smoking-related comorbidities, were examined. RESULTS: Active smokers did not experience worse COVID-19 outcomes once hospitalized. Former smokers had a higher odds of mortality (total effect OR 1.59, 95% CI 1.07 to 2.38, p = .01; indirect effect OR 1.45, 95% CI 1.09 to 1.93, p < .001), and advanced respiratory support (total effect OR 1.31, 95% CI 1.04 to 1.67, p = .02; indirect effect OR 1.26, 95% CI 1.03 to 1.54, p = .02), which were mediated by smoking-related comorbidities. While there was a nonsignificant increase in the total effect for mechanical ventilation, smoking-related comorbidities were significant mediators for their increased need (total effect OR 1.40, 95% CI 0.92 to 2.14, p = .13; indirect effect OR 1.47, 95% CI 1.10 to 1.87, p < .001). CONCLUSIONS: Although active smokers did not experience worse COVID-19 outcomes compared to never smokers, these results should be interpreted with caution. Compared to never smokers, former smokers had greater odds of mortality, advanced respiratory support, and mechanical ventilation which was significantly mediated through smoking-related comorbidities. IMPLICATIONS: Previous studies have linked smoking status with worse COVID-19 outcomes, and have inferred that smoking-related comorbidities may play a role in these findings. This causal mediation analysis provides statistical evidence supporting this hypothesis, clarifying the risk that smoking-related comorbidities impart on COVID-19 outcomes in those with a smoking history.

The Role of Pre-Treatment Traumatic Stress Symptoms in Adolescent Substance Use Treatment Outcomes.

Background: Prominent theories suggest that individuals with co-occurring traumatic stress symptoms (TSS) and substance use (SU) may be less responsive to SU treatment compared to those with SU only. However, empirical findings in adult samples are mixed, and there has been limited work among adolescents. This study assesses the association between TSS and SU treatment outcomes among trauma-exposed adolescents, using statistical methods to reduce potential confounding from important factors such as baseline SU severity. Method: 2,963 adolescents with lifetime history of victimization received evidence-based SU treatment in outpatient community settings. At baseline, 3- and 6-months, youth were assessed using the Global Appraisal of Individual Needs Traumatic Stress Scale and the Substance Frequency Scale. Propensity score weighting was used to mitigate potential confounding due to baseline differences in sociodemographic characteristics and SU across youth with varying levels of TSS. Results: Propensity score weighting successfully balanced baseline differences in sociodemographic factors and baseline SU across youth. Among all youth, mean SU was lower at both 3- and 6- month follow-up relative to baseline, indicating declining use. After adjusting for potential confounders, we observed no statistically significant relationship between TSS and SU at either 3- or 6-month follow-up. Conclusions: Based on this investigation, conducted among a large sample of trauma-exposed youth receiving evidence-based outpatient SU treatment, baseline TSS do not appear to be negatively associated with SU treatment outcomes. However, future research should examine whether youth with TSS achieve better outcomes through integrative treatment for both SU and TSS.

The results of this study provide keenly needed evidence that, among youth with prior victimization, presence and level of traumatic stress symptoms at substance use treatment initiation does not lead to significantly worse treatment outcomes for youth in outpatient treatment. This suggests that evidence-based outpatient substance use treatment modalities may be effective at improving substance use outcomes even when co-existing traumatic stress symptoms are present.

A Causal Approach to Functional Mediation Analysis with Application to a Smoking Cessation Intervention.

The increase in the use of mobile and wearable devices now allows dense assessment of mediating processes over time. For example, a pharmacological intervention may have an effect on smoking cessation via reductions in momentary withdrawal symptoms. We define and identify the causal direct and indirect effects in terms of potential outcomes on the mean difference and odds ratio scales, and present a method for estimating and testing the indirect effect of a randomized treatment on a distal binary variable as mediated by the nonparametric trajectory of an intensively measured longitudinal variable (e.g., from ecological momentary assessment). Coverage of a bootstrap test for the indirect effect is demonstrated via simulation. An empirical example is presented based on estimating later smoking abstinence from patterns of craving during smoking cessation treatment. We provide an R package, funmediation, available on CRAN at https://cran.r-project.org/web/packages/funmediation/index.html, to conveniently apply this technique. We conclude by discussing possible extensions to multiple mediators and directions for future research.

Sensitivity and specificity of information criteria.

Information criteria (ICs) based on penalized likelihood, such as Akaike's information criterion (AIC), the Bayesian information criterion (BIC) and sample-size-adjusted versions of them, are widely used for model selection in health and biological research. However, different criteria sometimes support different models, leading to discussions about which is the most trustworthy. Some researchers and fields of study habitually use one or the other, often without a clearly stated justification. They may not realize that the criteria may disagree. Others try to compare models using multiple criteria but encounter ambiguity when different criteria lead to substantively different answers, leading to questions about which criterion is best. In this paper we present an alternative perspective on these criteria that can help in interpreting their practical implications. Specifically, in some cases the comparison of two models using ICs can be viewed as equivalent to a likelihood ratio test, with the different criteria representing different alpha levels and BIC being a more conservative test than AIC. This perspective may lead to insights about how to interpret the ICs in more complex situations. For example, AIC or BIC could be preferable, depending on the relative importance one assigns to sensitivity versus specificity. Understanding the differences and similarities among the ICs can make it easier to compare their results and to use them to make informed decisions.

Randomized controlled trial to enhance children's sleep, eating, and weight.

BACKGROUND: The present study assessed the efficacy of a behavioral intervention to enhance children's sleep and reduce caloric intake and body mass index (BMI) change. METHODS: Seventy-eight children 8-11 years old who slept 9.5 h/night or less were randomized to the sleep intervention or to no treatment control. The primary outcome was 2-month change in the actigraph-estimated sleep period; changes in reported caloric intake, percent calories from fat, and BMI/BMI z-score (BMIz) were assessed. RESULTS: Children randomized to intervention enhanced their sleep period by 40 ± 7 min/night relative to control (p < 0.001), and were more likely to increase their sleep period by 30 min/night or more (52% versus 15%, p = 0.003). No differences were observed for reported dietary intake or BMI/BMIz. However, in post-hoc analyses collapsing across groups, those who increased sleep by 30 min/night or more had lower BMI (-0.31 kg/m2, p = 0.01) and BMIz (-0.07, p = 0.03) and reported fewer percent calories from fat at 2 months (-2.2%, p = 0.04). CONCLUSIONS: A brief behavioral intervention can enhance children's sleep, but did not result in changes in caloric intake or weight status. Enhancing sleep by 30 min/night or more may be beneficial for weight regulation. IMPACT: A brief behavioral intervention improved children's nocturnal sleep relative to no treatment control. Given the many benefits of a good night's sleep across domains of functioning, findings have significant implications for children's health and wellbeing. There were no differences between groups on eating behaviors or BMI. However, across groups, children who increased their sleep period by at least 30 min/night, reported reduced intake from fat and evidenced lower BMI at 2 months. Thus, a brief intervention can improve sleep and may have potential benefits for weight regulation.

Estimation and inference for the mediation effect in a time-varying mediation model.

BACKGROUND: Traditional mediation analysis typically examines the relations among an intervention, a time-invariant mediator, and a time-invariant outcome variable. Although there may be a total effect of the intervention on the outcome, there is a need to understand the process by which the intervention affects the outcome (i.e., the indirect effect through the mediator). This indirect effect is frequently assumed to be time-invariant. With improvements in data collection technology, it is possible to obtain repeated assessments over time resulting in intensive longitudinal data. This calls for an extension of traditional mediation analysis to incorporate time-varying variables as well as time-varying effects. METHODS: We focus on estimation and inference for the time-varying mediation model, which allows mediation effects to vary as a function of time. We propose a two-step approach to estimate the time-varying mediation effect. Moreover, we use a simulation-based approach to derive the corresponding point-wise confidence band for the time-varying mediation effect. RESULTS: Simulation studies show that the proposed procedures perform well when comparing the confidence band and the true underlying model. We further apply the proposed model and the statistical inference procedure to data collected from a smoking cessation study. CONCLUSIONS: We present a model for estimating time-varying mediation effects that allows both time-varying outcomes and mediators. Simulation-based inference is also proposed and implemented in a user-friendly R package.

Time-Varying Mediation of Pharmacological Smoking Cessation Treatments on Smoking Lapse via Craving, Cessation Fatigue, and Negative Mood.

INTRODUCTION: The addictive nature of nicotine makes smoking cessation an extremely challenging process. With prolonged exposure, tobacco smoking transforms from being a positive reinforcer to a negative one, as smoking is used to mitigate aversive withdrawal symptoms. Studying the variations in withdrawal symptoms, especially during their peak in the first week of a quit attempt, could help improve cessation treatment for the future. The time-varying mediation model effectively studies whether altering withdrawal symptoms act as mediators in the pathway between treatment and cessation. AIMS AND METHODS: This secondary data analysis of a randomized clinical smoking cessation trial of three pharmacotherapy regimens (nicotine patch, varenicline, and nicotine patch + mini-lozenge) analyzes ecological momentary assessment (EMA) data from the first 4 weeks post-target quit day (TQD). We assess whether withdrawal symptoms (eg, negative mood, cessation fatigue, and craving) mediate the pathway between pharmacotherapy and daily smoking status and whether this effect varies over time. RESULTS: We found a statistically significant time-varying mediation effect of varenicline on smoking status through craving, which shows decreasing risk of lapse via reduction in craving. We did not find significant time-varying mediation effects through negative mood and cessation fatigue. CONCLUSIONS: This study supports the importance of craving suppression in the smoking cessation process. It also helped identify specific timepoints when withdrawal symptoms increased that would likely benefit from targeted cessation intervention strategies. IMPLICATIONS: This study aimed to understand the underlying dynamic mechanisms of the smoking cessation process using a new analytical approach that capitalizes on the intensive longitudinal data collected via EMAs. The findings from this study further elucidate the smoking cessation process and provide insight into behavioral intervention targets and the timing of such interventions through the estimation of time-varying mediation effects.

Relationships Between E-cigarette Use and Subsequent Cigarette Initiation Among Adolescents in the PATH Study: an Entropy Balancing Propensity Score Analysis.

This study aimed to examine the relationship between electronic cigarette use and subsequent combustible cigarette use, controlling for confounding by using a propensity score method approach. Data from the first three annual waves of the Population Assessment of Tobacco and Health study were analyzed (n = 6309). Participants were tobacco-naïve at Wave 1; used e-cigarettes exclusively (n = 414), used combustible cigarettes exclusively (n = 46), or not used any tobacco products (n = 5849) at Wave 2. We conducted entropy balancing propensity score analysis to examine the association between exclusive e-cigarette or cigarette initiation and subsequent cigarette use at Wave 3, adjusting for non-response bias, sampling bias, and confounding. Among tobacco-naïve youth, exclusive e-cigarette use was associated with greater risk for subsequent combustible cigarette smoking initiation (OR = 3.42, 95% CI = (1.99, 5.93)) and past 30-day combustible cigarette use (OR = 2.88, 95% CI = (1.22, 6.86)) in the following year. However, the latter risk was comparatively lower than the risk if youth started with a combustible cigarette (OR = 25.79, 95% CI = (9.68, 68.72)). Results of sensitivity analyses indicated that estimated effects were robust to unmeasured confounding. Use of e-cigarettes in tobacco-naïve youth is associated with increased risk of subsequent past 30-day combustible cigarette use but the risk is an order of magnitude higher if they start with a combustible cigarette.

The effects of visual context on visual-vestibular mismatch revealed by electrodermal and postural response measures.

BACKGROUND: No objective criteria exist for diagnosis and treatment of visual-vestibular mismatch (VVM). OBJECTIVE: To determine whether measures of electrodermal activity (EDA) and trunk acceleration will identify VVM when exposed to visual-vestibular conflict. METHODS: A modified VVM questionnaire identified the presence of VVM (+ VVM) in 13 of 23 young adults (34 ± 8 years) diagnosed with vestibular migraine. Rod and frame tests and outcome measures for dizziness and mobility were administered. Participants stood on foam while viewing two immersive virtual environments. Trunk acceleration in three planes and electrodermal activity (EDA) were assessed with wearable sensors. Linear mixed effect (LME) models were used to examine magnitude and smoothness of trunk acceleration and tonic and phasic EDA. Welch's t-test and associations between measures were assessed with a Pearson Correlation Coefficient. Effect sizes of group mean differences were calculated. RESULTS: Greater than 80% of all participants were visually dependent. Outcome measures were significantly poorer in the + VVM group: tonic EDA was lower (p < 0.001) and phasic EDA higher (p < 0.001). Postural accelerations varied across groups; LME models indicated a relationship between visual context, postural, and ANS responses in the + VVM group. CONCLUSIONS: Lower tonic EDA with + VVM suggests canal-otolith dysfunction. The positive association between vertical acceleration, tonic EDA, and visual dependence suggests that increased vertical segmental adjustments are used to compensate. Visual context of the spatial environment emerged as an important control variable when testing or treating VVM.

Impact of topical budesonide on prevention of esophageal stricture after mucosal resection.

BACKGROUND AND AIMS: EMR and endoscopic submucosal dissection (ESD) are treatment modalities for Barrett's esophagus involving high-grade dysplasia or early cancer. Injectional corticosteroid therapy decreases the risk of procedure-related esophageal stricture (ES) formation. Our aim was to assess the efficacy of topical budesonide on the rate of ES formation after EMR or ESD. METHODS: Patients included prospectively from 3 tertiary endoscopy centers received 3 mg budesonide orally twice a day for 8 weeks after esophageal EMR or ESD of 50% or more of the esophageal circumference between January 1, 2014 and June 30, 2018. These patients were matched (1:3 ratio) retrospectively with a consecutive patient cohort who underwent EMR or ESD of 50% or more of the esophageal circumference without concomitant corticosteroid therapy. The primary endpoint was the presence of ES at the 12-week follow-up. RESULTS: Twenty-five patients (budesonide) were matched with 75 patients (no budesonide). Most underwent EMR for Barrett's esophagus with biopsy-proven high-grade dysplasia or suspected T1a cancer. Although most baseline characteristics did not differ significantly, patients in the budesonide cohort tended to have a higher proportion of circumferential EMR. The proportion of patients with ES was not significantly lower in the budesonide cohort (16% vs 28%). On logistic regression analysis, budesonide remained associated with a lower incidence of ES (P = .023); however, when controlling for baseline characteristics with a propensity score weighted logistic regression model, there was no significant effect on ES formation (P = .176). CONCLUSIONS: Topical budesonide might be associated with a reduction of ES after EMR or ESD; however, further studies are needed to verify our results.

Comparison of methods for handling covariate missingness in propensity score estimation with a binary exposure.

BACKGROUND: Causal effect estimation with observational data is subject to bias due to confounding, which is often controlled for using propensity scores. One unresolved issue in propensity score estimation is how to handle missing values in covariates. METHOD: Several approaches have been proposed for handling covariate missingness, including multiple imputation (MI), multiple imputation with missingness pattern (MIMP), and treatment mean imputation. However, there are other potentially useful approaches that have not been evaluated, including single imputation (SI) + prediction error (PE), SI + PE + parameter uncertainty (PU), and Generalized Boosted Modeling (GBM), which is a nonparametric approach for estimating propensity scores in which missing values are automatically handled in the estimation using a surrogate split method. To evaluate the performance of these approaches, a simulation study was conducted. RESULTS: Results suggested that SI + PE, SI + PE + PU, MI, and MIMP perform almost equally well and better than treatment mean imputation and GBM in terms of bias; however, MI and MIMP account for the additional uncertainty of imputing the missingness. CONCLUSIONS: Applying GBM to the incomplete data and relying on the surrogate split approach resulted in substantial bias. Imputation prior to implementing GBM is recommended.

Long-term consequences of early sexual initiation on young adult health: A causal inference approach.

Although early sexual initiation has been linked to negative outcomes, it is unknown whether these effects are causal. In this study, we use propensity score methods to estimate the causal effect of early sexual initiation on young adult sexual risk behaviors and health outcomes using data from the National Longitudinal Study of Adolescent to Adult Health. We found that early sexual initiation predicted having 2 or more partners (for both males and females) and having a sexually transmitted infection in the past year (females only) but did not predict depressive symptoms in the past week (for either gender). These results underscore the importance of continued programmatic efforts to delay age of sexual initiation, particularly for females.

Restrictive educational placements increase adolescent risks for students with early-starting conduct problems.

BACKGROUND: Students with early-starting conduct problems often do poorly in school; they are disproportionately placed in restrictive educational placements outside of mainstream classrooms. Although intended to benefit students, research suggests that restrictive placements may exacerbate the maladjustment of youth with conduct problems. Mixed findings, small samples, and flawed designs limit the utility of existing research. METHODS: This study examined the impact of restrictive educational placements on three adolescent outcomes (high school noncompletion, conduct disorder, depressive symptoms) in a sample of 861 students with early-starting conduct problems followed longitudinally from kindergarten (age 5-6). Causal modeling with propensity scores was used to adjust for confounding factors associated with restrictive placements. Analyses explored the timing of placement (elementary vs. secondary school) and moderation of impact by initial problem severity. RESULTS: Restrictive educational placement in secondary school (but not in elementary school) was iatrogenic, increasing the risk of high school noncompletion and the severity of adolescent conduct disorder. Negative effects were amplified for students with conduct problem behavior with less cognitive impairment. CONCLUSIONS: To avoid harm to students and to society, schools must find alternatives to restrictive placements for students with conduct problems in secondary school, particularly when these students do not have cognitive impairments that might warrant specialized educational supports.

Penalized regression procedures for variable selection in the potential outcomes framework.

A recent topic of much interest in causal inference is model selection. In this article, we describe a framework in which to consider penalized regression approaches to variable selection for causal effects. The framework leads to a simple 'impute, then select' class of procedures that is agnostic to the type of imputation algorithm as well as penalized regression used. It also clarifies how model selection involves a multivariate regression model for causal inference problems and that these methods can be applied for identifying subgroups in which treatment effects are homogeneous. Analogies and links with the literature on machine learning methods, missing data, and imputation are drawn. A difference least absolute shrinkage and selection operator algorithm is defined, along with its multiple imputation analogs. The procedures are illustrated using a well-known right-heart catheterization dataset.