Kwonsang Lee

Understanding causal interactions is critical in observational studies but often challenging due to confounding and methodological limitations. While these concepts have been extensively studied in randomized experiments, their application in observational data remains limited, particularly in settings requiring stratification to evaluate interactions. We propose a novel randomization-based inference framework utilizing matching methods to investigate causal interactions. Additionally, we provide a comprehensive review of causal interaction, explaining its unique focus on joint causal effects and distinguishing it from statistical interaction and effect modification. Using a real-world dataset, we analyze the joint effects of two treatments: residential fire safety equipment and fire response time. Our approach demonstrates how matching can mitigate confounding and identify interaction effects, offering a robust alternative to traditional methods. The findings highlight the importance of considering causal interaction in public safety interventions and provide actionable insights for fire safety policy and fire response optimization.

Dahai Jung, Yongho Yoon, Kwonsang Lee. Korean J Appl Stat 2024;37:675-90. https://doi.org/10.5351/KJAS.2024.37.5.675

Observational studies are frequently used to estimate the effect of an exposure or treatment on an outcome. To obtain an unbiased estimate of the treatment effect, it is crucial to measure the exposure accurately. A common type of exposure misclassification is recall bias, which occurs in retrospective cohort studies when study subjects may inaccurately recall their past exposure. Particularly challenging is differential recall bias in the context of self-reported binary exposures, where the bias may be directional rather than random and its extent varies according to the outcomes experienced. This paper makes several contributions: (1) it establishes bounds for the average treatment effect even when a validation study is not available; (2) it proposes multiple estimation methods across various strategies predicated on different assumptions; and (3) it suggests a sensitivity analysis technique to assess the robustness of the causal conclusion, incorporating insights from prior research. The effectiveness of these methods is demonstrated through simulation studies that explore various model misspecification scenarios. These approaches are then applied to investigate the effect of childhood physical abuse on mental health in adulthood.

In health and social sciences, it is critically important to identify interpretable subgroups of the study population where a treatment has notable heterogeneity in the causal effects with respect to the average treatment effect (ATE). Several approaches have already been proposed for heterogeneous treatment effect (HTE) discovery, either estimating first the conditional average treatment effect (CATE) and identifying heterogeneous subgroups in a second stage (Bargagli-Stoffi et al., 2020, 2022; Foster et al., 2011; Hahn et al., 2020), either estimating directly these subgroups in a direct data-driven procedure (Nagpal et al., 2020; Wang & Rudin, 2022). Many of these methodologies are decision tree-based methodologies. Tree-based approaches are based on efficient and easily implementable recursive mathematical programming (e.g., HTE maximization), they can be easily tweaked and adapted to different scenarios depending on the research question of interest, and they guarantee a high degree of interpretability—i.e., the degree to which a human can understand the cause of a decision (Lakkaraju et al., 2016). Despite these appealing features, single-tree heterogeneity discovery is characterized by two main limitations: instability in the identification of the subgroups and reduced exploration of the potential heterogeneity. To accommodate these shortcomings, Bargagli-Stoffi et al. (2023) proposed Causal Rule Ensemble, a new method for interpretable HTE characterization in terms of decision rules, via an extensive exploration of heterogeneity patterns by an ensemble-of-trees approach. CRE is an R package providing a flexible implementation of Causal Rule Ensemble. The package allows for multiple variants of Causal Rule Ensemble algorithm, also including different internal individual average treatment effect (IATE) estimators—i.e., AIPW (Robins et al., 1994), Causal Forest (Athey et al., 2019), Causal BART (Hill, 2011), S-Learner (Hill, 2011), T-Learner (Hansotia & Rukstales, 2002), X-Learner (Künzel et al., 2019).

Several studies have provided strong evidence that long-term exposure to air pollution, even at low levels, increases risk of mortality. As regulatory actions are becoming prohibitively expensive, robust evidence to guide the development of targeted interventions to protect the most vulnerable is needed. In this article, we introduce a novel statistical method that (i) discovers subgroups whose effects substantially differ from the population mean, and (ii) uses randomization-based tests to assess discovered heterogeneous effects. Also, we develop a sensitivity analysis method to assess the robustness of the conclusions to unmeasured confounding bias. Via simulation studies and theoretical arguments, we demonstrate that hypothesis testing focusing on the discovered subgroups can substantially increase statistical power to detect heterogeneity of the exposure effects. We apply the proposed de novo method to the data of 1,612,414 Medicare beneficiaries in the New England region in the United States for the period 2000–2006. We find that seniors aged between 81 and 85 with low income and seniors aged 85 and above have statistically significant greater causal effects of long-term exposure to PM_2.5 on 5-year mortality rate compared to the population mean.

We investigate the efficacy of surgical versus non-surgical management for two gastrointestinal conditions, colitis and diverticulitis, using observational data. We deploy an instrumental variable design with surgeons’ tendencies to operate as an instrument. Assuming instrument validity, we find that non-surgical alternatives can reduce both hospital length of stay and the risk of complications, with estimated effects larger for septic patients than for non-septic patients. The validity of our instrument is plausible but not ironclad, necessitating a sensitivity analysis. Existing sensitivity analyses for IV designs assume effect homogeneity, unlikely to hold here because of patient-specific physiology. We develop a new sensitivity analysis that accommodates arbitrary effect heterogeneity and exploits components explainable by observed features. We find that the results for non-septic patients prove more robust to hidden bias despite having smaller estimated effects. For non-septic patients, two individuals with identical observed characteristics would have to differ in their odds of assignment to a high tendency to operate surgeon by a factor of 2.34 to overturn our finding of a benefit for non-surgical management in reducing length of stay. For septic patients, this value is only 1.64. Simulations illustrate that this phenomenon may be explained by differences in within-group heterogeneity.

Many cities and countries have implemented heat wave warning systems to combat the health effects of extreme heat. Little is known about whether these systems actually reduce heat-related morbidity and mortality. We examined the effectiveness of heat wave alerts and health plans in reducing the mortality risk of heat waves in Korea by utilizing the discrepancy between the alerts and the monitored temperature. A difference-in-differences analysis combined with propensity score weighting was used. Mortality, weather monitoring, and heat wave alert announcement data were collected for 7 major cities during 2009–2014. Results showed evidence of risk reduction among people aged 19–64 without education (−0.144 deaths/1,000,000 people, 95% CI: −0.227, −0.061) and children aged 0–19 (−0.555 deaths/1,000,000 people, 95% CI: −0.993, −0.117). Decreased cardiovascular and respiratory mortality was found in several subgroups including single persons, widowed people, blue-collar workers, people with no education or the highest level of education (university or higher). No evidence was found for decreased all-cause mortality in the population (1.687 deaths/1,000,000 people per day; 95% CI: 1.118, 2.255). In conclusion, heat wave alerts may reduce mortality for several causes and subpopulations of age and socio-economic status. Further work needs to examine the pathways through which the alerts impact subpopulations differently.

Two problems that arise in making causal inferences for nonmortality outcomes such as bronchopulmonary dysplasia (BPD) are unmeasured confounding and censoring by death, ie, the outcome is observed only when subjects survive. In randomized experiments with noncompliance and no censoring by death, instrumental variable (IV) methods can be used to control for the unmeasured confounding. But, when there is censoring by death, the average causal treatment effect cannot be identified under usual assumptions but can be studied for a specific subpopulation by using sensitivity analysis with additional assumptions. However, evaluating the local average treatment effect (LATE) in observational studies with censoring by death problems while controlling for unmeasured confounding is not well studied. We develop a novel sensitivity analysis method based on IV models for studying the LATE. Specifically, we present the identification results under an additional assumption and propose a three‐step procedure for the LATE estimation. Also, we propose an improved two‐step procedure by simultaneously estimating the instrument propensity score (ie, the probability of instrument given covariates) and the parameters induced by the assumption. We show with simulation studies that the two‐step procedure can be more robust and efficient than the three‐step procedure. Finally, we apply our sensitivity analysis methods to a study on the effect of delivery at high‐level neonatal intensive care units on the risk of BPD.