596Two approaches for estimating propensity score weights for examining neighbourhood built environment and walking changes

Background Walking is associated with the built environment, however, this association may be biased by residential self-selection. This study examined how walking duration changed with residential relocation, while accounting for unbalanced covariates that may contribute to residential self-selection, using two different propensity score inverse probability weight (IPW) methods. Methods Urban participants (n = 703) of Alberta’s Tomorrow Project with pre- and post-relocation neighbourhood built environment and walking data were included. A walkability index was created by aggregating estimates for population density, street connectivity, and destination diversity. Participants were categorized into three groups based on change in residential walkability (decreased, minimal change, or increased). The association between changes in walkability and walking duration (min/week) was modelled with linear regression. Two types of IPWs were applied: 1) manually generated from multinomial regression models, and 2) generated from generalized boosted models. Results All three groups increased walking duration from pre- to post-relocation, however the largest increase was among participants who had increased walkability (M = 73.2, SD = 388), followed by those with minimal change (M = 60.0, SD = 382) and decreased (M = 50.2, SD = 374) walkability. Longitudinal associations between walkability change and walking were not statistically significant (p < 0.05) in models with or without IPWs. Conclusions Changes in neighbourhood walkability were not associated with changes in walking, regardless of how the sample was weighted. Further research should examine changes in the neighbourhood environment with different types of walking and physical activity behaviours. Key messages IPW methods can be used to account for unbalanced covariates in analyses that involve possible self-selection bias.

Predicting Higher Education Grades using Strategies Correcting for Panel Attrition

This study aims to forecast the final grade of the first higher education degree which can be of considerable interest for higher education institutions to implement early warning systems, students themselves, or potential employers. The analysis is based on the National Education Panel Study (NEPS), a large German dataset covering many aspects of students’ (educational) life. Since panel attrition concerns 35% of participants the Heckman correction and the inverse probability weight (IPW) estimator are used to reduce the estimation bias. A distinction is made between two scenarios, excluding dropout students and including them with a grade of 5.0. Some predictors reveal significant parameter estimates in the first but not in the second scenario, or vice versa, which means that dropout and study performance is not driven by the same variables. To get an early prediction of grades only variables of a pre-university episode were included in the first step. Afterward, variables of the early study phase are included. For the IPW estimator, the R2 improves from 0.202 to 0.593 (dropouts included) when adding the additional variables. The best predictors are the grades at secondary school, grades in the first exams, and the type of institution.

Ten-Year Clinical Outcomes of Late-Acquired Stent Malapposition After Coronary Stent Implantation

Objective: The goal of this study was to determine the impact of late-acquired stent malapposition (LASM) on long-term clinical outcomes in patients treated with coronary stent implantation. Approach and Results: We investigated major adverse cardiac event during 10 years after 6-month intravascular ultrasound examination using our previous studies database. A total of 732 patients treated with bare-metal stent (54 LASM versus 678 non-LASM) and 529 patients treated with first-generation drug-eluting stent (82 LASM versus 447 non-LASM), who did not have clinical event or censoring at the time of follow-up intravascular ultrasound, were included for the present analysis. major adverse cardiac event was defined as the composite of cardiac death, target vessel–related myocardial infarction, target lesion revascularization and stent thrombosis. Multivariable adjustment and inverse probability weight were performed to consider baseline differences. After multivariable adjustment, LASM was related to a greater risk of major adverse cardiac event (hazard ratio, 1.666 [95% CI, 1.041–2.665]; P =0.0333) and very-late stent thrombosis (hazard ratio, 3.529 [95% CI, 1.153–10.798]; P =0.0271) than non-LASM in patients treated with first-generation drug-eluting stent, but not in those treated with bare-metal stent. Results were consistent after inverse probability weight. Among patients with LASM of first-generation drug-eluting stent, no late stent thrombosis occurred in patients who continued to receive dual antiplatelet therapy. Conclusions: The relationship between LASM and major adverse cardiac event might depend on the type of implanted stents during the long-term follow-up, highlighting the clinical significance of polymers and drugs in drug-eluting stent system.

Assessing balance in outcome regressions

An outcome regression controlling for observed confounders remains a popular way to assess the causal effect of an exposure in epidemiology, despite more modern causal techniques for adjusting for observed confounders, such as inverse probability weighting. A feature of inverse probability weighting is that checking balance of confounders in the control and exposure groups after confounder adjustment is simple. However, researchers using outcome regressions commonly do not check confounder balance after controlling for confounders. Although outcome regressions will balance any confounder specified in the model, the confounder value the model balances at is not transparent. We show that a matrix representation of an outcome regression reveals that an outcome regression includes a weight similar to an inverse probability weight. We also show that outcome regressions may not be balancing at the sample mean of the confounders particularly if interactions are not included with the exposure, which is typically the case in outcome regressions. Finally, we show that the coefficient of the exposure in an outcome regression is simply the difference between two weighted counterfactuals. Thus, there is an important connection between traditional outcome regression and modern causal techniques.