Conditional Treatment Effects Research Articles

Estimation of the Local Conditional Tail Average Treatment Effect*

The conditional tail average treatment effect (CTATE) is defined as a difference between the conditional tail expectations (CTE’s) of potential outcomes, which can capture heterogeneity and deliver aggregated local information on treatment effects over different quantile levels and is closely related to the notion of second-order stochastic dominance and the Lorenz curve. These properties render it a valuable tool for policy evaluation. In this paper, we study estimation of the CTATE locally for a group of compliers (local CTATE or LCTATE) under the two-sided noncompliance framework. We consider a semiparametric treatment effect framework under endogeneity for the LCTATE estimation using a newly introduced class of consistent loss functions jointly for the CTE and quantile. We establish the asymptotic theory of our proposed LCTATE estimator and provide an efficient algorithm for its implementation. We then apply the method to evaluate the effects of participating in programs under the Job Training Partnership Act in the US.

Learning optimal biomarker-guided treatment policy for chronic disorders.

Electroencephalogram (EEG) provides noninvasive measures of brain activity and is found to be valuable for the diagnosis of some chronic disorders. Specifically, pre-treatment EEG signals in the alpha and theta frequency bands have demonstrated some association with antidepressant response, which is well-known to have a low response rate. We aim to design an integrated pipeline that improves the response rate of patients with major depressive disorder by developing a treatment policy guided by the resting state pre-treatment EEG recordings and other treatment effects modifiers. First, we design an innovative automatic site-specific EEG preprocessing pipeline to extract features with stronger signals than raw data. We then estimate the conditional average treatment effect (CATE) using causal forests and use a doubly robust technique to improve efficiency in the estimation of the average treatment effect. We present evidence of heterogeneity in the treatment effect and the modifying power of the EEG features, as well as a significant average treatment effect, a result that cannot be obtained with conventional methods. Finally, we employ an efficient policy learning algorithm to learn an optimal depth-2 treatment assignment decision tree and compare its performance with Q-Learning and outcome-weighted learning via simulation studies and an application to a large multi-site, double-blind, randomized controlled clinical trial, EMBARC.

Assessing external validity in practice

We review, from a practical standpoint, the evolving literature on assessing external validity (EV) of estimated treatment effects. We review existing EV measures, and focus on methods that permit multiple datasets (Hotz et al., 2005). We outline criteria for practical usage, evaluate the existing approaches, and identify a gap in potential methods. Our practical considerations motivate a novel method utilizing the Group Lasso (Yuan and Lin, 2006) to estimate a tractable regression-based model of the conditional average treatment effect (CATE). This approach can perform better when settings have differing covariate distributions and allows for easily extrapolating the average treatment effect to new settings. We apply these measures to a set of identical field experiments upgrading slum dwellings in three different countries (Galiani et al., 2017).

Quantifying uncertainty of uplift: Trees and T-learners

Uplift modeling refers to the task of estimating the causal effect of a treatment on an individual, also known as the conditional average treatment effect. However, uplift models do not usually provide uncertainty estimates of the predictions. We explain why estimating uncertainty of the treatment effect is particularly important in many common use cases and we show how epistemic uncertainty of the uplift estimates can be quantified for T-learners and trees. We tested the methods on three empirical datasets and evaluated them on a simulated dataset. We found that high uncertainty might be the result of both modeling choices and properties of the data. Sometimes there is not enough data or the data is simply not rich enough to identify the treatment effect well resulting in high uncertainty. In addition, our results suggest that one commonly used dataset might not be suitable for benchmarking.

Nonparametric estimation of conditional incremental effects

Abstract Conditional effect estimation has great scientific and policy importance because interventions may impact subjects differently depending on their characteristics. Most research has focused on estimating the conditional average treatment effect (CATE). However, identification of the CATE requires that all subjects have a non-zero probability of receiving treatment, or positivity, which may be unrealistic in practice. Instead, we propose conditional effects based on incremental propensity score interventions, which are stochastic interventions where the odds of treatment are multiplied by some factor. These effects do not require positivity for identification and can be better suited for modeling scenarios in which people cannot be forced into treatment. We develop a projection approach and a flexible nonparametric estimator that can each estimate all the conditional effects we propose and derive model-agnostic error guarantees showing that both estimators satisfy a form of double robustness. Further, we propose a summary of treatment effect heterogeneity and a test for any effect heterogeneity based on the variance of a conditional derivative effect and derive a nonparametric estimator that also satisfies a form of double robustness. Finally, we demonstrate our estimators by analyzing the effect of intensive care unit admission on mortality using a dataset from the (SPOT)light study.

Finding the best subgroup with differential treatment effect with multiple outcomes.

Precision medicine aims to identify specific patient subgroups that may benefit the most from a particular treatment than the whole population. Existing definitions for the best subgroup in subgroup analysis are based on a single outcome and do not consider multiple outcomes; specifically, outcomes of different types. In this article, we introduce a definition for the best subgroup under a multiple-outcome setting with continuous, binary, and censored time-to-event outcomes. Our definition provides a trade-off between the subgroup size and the conditional average treatment effects (CATE) in the subgroup with respect to each of the outcomes while taking the relative contribution of the outcomes into account. We conduct simulations to illustrate the proposed definition. By examining the outcomes of urinary tract infection and renal scarring in the RIVUR clinical trial, we identify a subgroup of children that would benefit the most from long-term antimicrobial prophylaxis.

Estimating conditional average treatment effects with heteroscedasticity by model averaging and matching

We propose a model averaging approach, combined with a partition and matching method to estimate the conditional average treatment effects under heteroskedastic error settings. The proposed approach has asymptotic optimality and consistency of weights and estimator. Numerical studies show that our method has good finite-sample performances.

Heterogeneity of the effect of the COVID-19 pandemic on the incidence of Metabolic Syndrome onset at a Japanese campus.

The coronavirus disease 2019 (COVID-19) outbreak began in China in December 2019, with the World Health Organization declaring a state of emergency in January 2020. Worldwide implementation of lockdown measures to slow the spread of the virus led to reduced physical activity, disrupted eating habits, mental health issues, and sleep disturbances, which increased the risk of lifestyle-related diseases such as metabolic syndrome (MetS). During the COVID-19 pandemic, healthcare workers, especially intensive care workers, experienced longer working hours and burnout, which further increased the risk of lifestyle-related diseases. Accordingly, it is important to identify individuals at a risk of new-onset MetS during a pandemic, which could direct preventive interventions. This study aimed to assess the heterogeneous impact of the COVID-19 pandemic on the incidence of new-onset MetS based on the conditional average treatment effect (CATE) and to identify at-risk populations. This study analyzed health checkup data obtained from Okayama University Shikata Campus workers using paired baseline and follow-up years. Baseline data encompassed 2017 to 2019, with respective follow-up data from 2018 to 2020. Furthermore, as the COVID-19 pandemic in Japan began in January 2020, workers who underwent follow-up health checkups in 2018 to 2019 and 2020 were considered as "unexposed" and "exposed," respectively. As the Shikata campus has several departments, comparisons among departments were made. The primary outcome was new-onset MetS at follow-up. Predictor variables included baseline health checkup results, sex, age, and department (administrative, research, medical, or intensive care department). X-learner was used to calculate the CATE. This study included 3,572 eligible individuals (unexposed, n=2,181; exposed, n=1,391). Among them, 1,544 (70.8%) and 866 (62.3%) participants in the unexposed and exposed groups, respectively, were females. The mean age (±standard deviation) of the unexposed and exposed groups was 48.2 ± 8.2 and 47.8 ± 8.3 years, respectively. The COVID-19 pandemic increased the average probability of new-onset MetS by 4.4% in the overall population. According to the department, the intensive care department showed the highest CATE, with a 15.4% increase. Moreover, there was large heterogeneity according to the department. The high-CATE group was characterized by older age, urinary protein, elevated liver enzymes, higher triglyceride levels, and a history of hyperlipidemia treatment. This study demonstrated that the COVID-19 pandemic increased the incidence of new-onset MetS, with this effect showing heterogeneity at a single Japanese campus. Regarding specific populations, workers in the intensive care department showed an increased risk of new-onset MetS. At-risk populations require specific preventive interventions in case the current COVID-19 pandemic persists or a new pandemic occurs.

Bayesian Multivariate Logistic Regression for Superiority and Inferiority Decision-Making under Observable Treatment Heterogeneity

The effects of treatments may differ between persons with different characteristics. Addressing such treatment heterogeneity is crucial to investigate whether patients with specific characteristics are likely to benefit from a new treatment. The current paper presents a novel Bayesian method for superiority decision-making in the context of randomized controlled trials with multivariate binary responses and heterogeneous treatment effects. The framework is based on three elements: a) Bayesian multivariate logistic regression analysis with a Pólya-Gamma expansion; b) a transformation procedure to transfer obtained regression coefficients to a more intuitive multivariate probability scale (i.e., success probabilities and the differences between them); and c) a compatible decision procedure for treatment comparison with prespecified decision error rates. Procedures for a priori sample size estimation under a non-informative prior distribution are included. A numerical evaluation demonstrated that decisions based on a priori sample size estimation resulted in anticipated error rates among the trial population as well as subpopulations. Further, average and conditional treatment effect parameters could be estimated unbiasedly when the sample was large enough. Illustration with the International Stroke Trial dataset revealed a trend toward heterogeneous effects among stroke patients: Something that would have remained undetected when analyses were limited to average treatment effects.

다중비교 인과포레스트를 활용한 대학입학전형에 따른 대학생활 성과 분석

Over the years, university admission types have primarily been the subject of social discussions on the
 dimension of fairness. However, considering the original purpose of university admissions system, it is also
 crucial to evaluate how well each type performs in selecting students who can successfully navigate academic
 life at each university. Accordingly, this study aimed to analyze the university life outcomes (sense of
 belonging, adaptability, intention to reselect the university, intention to reselect the major, GPA) according
 to different university admissions types: the admission officer(AO) system, high school grade-focused(HSG)
 selection system, and College Scholastic Ability Test-focused(CSAT) selection system, using the data from the
 Korea Education Longitudinal Study(KELS) 2013 collected in ’20~’21(N=2,011). Assuming the effects may
 vary according to the baseline characteristics, the study employs the Conditional Average Treatment Effects
 (CATE) perspective through the Multi-arm Causal Forest technique(Athey, Tibshirani, & Wager, 2019). The
 main findings are summarized as follows: First, on average, students admitted through AO system exhibit
 superior outcomes across all considered aspects of college life, with a notable emphasis on higher intentions
 to reselect both the university and major. Second, it appears that parental income and education, high
 school academic achievement levels and their university-level average were significant conditioning variables
 predicting heterogeneous treatment effects on college life outcomes. Last, the effect heterogeneity of CSAT
 system, depending on the values of major conditioning variables, was more prominent compared to that of
 AO system. Based on these findings, suggestions for interpreting the research results and directions for
 future studies were provided.

Treatment Allocation with Strategic Agents

There is increasing interest in allocating treatments based on observed individual characteristics: examples include targeted marketing, individualized credit offers, and heterogeneous pricing. Treatment personalization introduces incentives for individuals to modify their behavior to obtain a better treatment. Strategic behavior shifts the joint distribution of covariates and potential outcomes. The optimal rule without strategic behavior allocates treatments only to those with a positive conditional average treatment effect. With strategic behavior, we show that the optimal rule can involve randomization, allocating treatments with less than 100% probability even to those who respond positively on average to the treatment. We propose a sequential experiment based on Bayesian optimization that converges to the optimal treatment rule without parametric assumptions on individual strategic behavior. This paper was accepted by Vivek Farias, data science. Supplemental Material: The data files are available at https://doi.org/10.1287/mnsc.2022.01629 .

Confounder-dependent Bayesian mixture model: Characterizing heterogeneity of causal effects in air pollution epidemiology.

Several epidemiological studies have provided evidence that long-term exposure to fine particulate matter (pm2.5) increases mortality rate. Furthermore, some population characteristics (e.g., age, race, and socioeconomic status) might play a crucial role in understanding vulnerability to air pollution. To inform policy, it is necessary to identify groups of the population that are more or less vulnerable to air pollution. In causal inference literature, the group average treatment effect (GATE) is a distinctive facet of the conditional average treatment effect. This widely employed metric serves to characterize the heterogeneity of a treatment effect based on some population characteristics. In this paper, we introduce a novel Confounder-Dependent Bayesian Mixture Model (CDBMM) to characterize causal effect heterogeneity. More specifically, our method leverages the flexibility of the dependent Dirichlet process to model the distribution of the potential outcomes conditionally to the covariates and the treatment levels, thus enabling us to: (i) identify heterogeneous and mutually exclusive population groups defined by similar GATEs in a data-driven way, and (ii) estimate and characterize the causal effects within each of the identified groups. Through simulations, we demonstrate the effectiveness of our method in uncovering key insights about treatment effects heterogeneity. We apply our method to claims data from Medicare enrollees in Texas. We found six mutually exclusive groups where the causal effects of pm2.5 on mortality rate are heterogeneous.

Heterogeneous Association of Tooth Loss with Functional Limitations

Tooth loss is prevalent in older adults and associated with functional capacity decline. Studies on the susceptibility of some individuals to the effects of tooth loss are lacking. This study aimed to investigate the heterogeneity of the association between tooth loss and higher-level functional capacity in older Japanese individuals employing a machine learning approach. This is a prospective cohort study using the data of adults aged ≥65 y in Japan (N = 16,553). Higher-level functional capacity, comprising instrumental independence, intellectual activity, and social role, was evaluated using the Tokyo Metropolitan Institute of Gerontology Index of Competence (TMIG-IC). The scale ranged from 0 (lowest function) to 13 (highest function). Doubly robust targeted maximum likelihood estimation was used to estimate the population-average association between tooth loss (having <20 natural teeth) and TMIG-IC total score after 6 y. The heterogeneity of the association was evaluated by estimating conditional average treatment effects (CATEs) using the causal forest algorithm. The result showed that tooth loss was statistically significantly associated with lower TMIG-IC total scores (population-average effect: −0.14; 95% confidence interval, −0.18 to −0.09). The causal forest analysis revealed the heterogeneous associations between tooth loss and lower TMIG-IC total score after 6 y (median of estimated CATEs = −0.13; interquartile range = 0.12). The high-impact subgroup (i.e., individuals with estimated CATEs of the bottom 10%) were significantly more likely to be older and male, had a lower socioeconomic status, did not have a partner, and had poor health conditions compared with the low-impact subgroup (i.e., individuals with estimated CATEs of the top 10%). This study found that heterogeneity exists in the association between tooth loss and lower scores on functional capacity. Implementing tooth loss prevention policy and clinical measures, especially among vulnerable subpopulations significantly affected by tooth loss, may reduce its burden more effectively.

Towards Safe Policy Learning under Partial Identifiability: A Causal Approach

Learning personalized treatment policies is a formative challenge in many real-world applications, including in healthcare, econometrics, artificial intelligence. However, the effectiveness of candidate policies is not always identifiable, i.e., it is not uniquely computable from the combination of the available data and assumptions about the generating mechanisms. This paper studies policy learning from data collected in various non-identifiable settings, i.e., (1) observational studies with unobserved confounding; (2) randomized experiments with partial observability; and (3) their combinations. We derive sharp, closed-formed bounds from observational and experimental data over the conditional treatment effects. Based on these novel bounds, we further characterize the problem of safe policy learning and develop an algorithm that trains a policy from data guaranteed to achieve, at least, the performance of the baseline policy currently deployed. Finally, we validate our proposed algorithm on synthetic data and a large clinical trial, demonstrating that it guarantees safe behaviors and robust performance.

Simultaneous treatment effect estimation and variable selection for observational data

Due to the difficulties inherent in conducting controlled experiments, recent causal inference studies have focused on developing treatment effect estimation using observational data. One major difficulty in causal inference from observational data is that the underlying causal structure is unknown. This may result in the misidentification of potential sources of causal estimation bias, such as confounders, which must be controlled for accurate estimation. To consider all possible confounders and other relevant information, conventional methods predominantly use all observed covariates indiscriminately. However, previous studies have shown that including all covariates without considering their causal relationships may deteriorate the estimation performance. Although several data-driven variable selection methods have been proposed for treatment effect estimation, they cannot distinguish the confounders from other outcome-related covariates and are limited to simple regression forms. In this study, we propose a method called the Variable Selection Causal Estimation Network (VSCEN) that performs treatment effect estimation and causal variable selection simultaneously. Through end-to-end differentiable training, the VSCEN selects only the covariates beneficial for effect estimation and uses only those selected for effect estimation. Experimental evaluations on fully synthetic, semi-synthetic, and real datasets demonstrate the VSCEN’s superior performance in conditional average treatment effect estimation and competitive performance in average treatment effect estimation along with its accurate variable selection capabilities.

Phenotype-based targeted treatment of SGLT2 inhibitors and GLP-1 receptor agonists in type 2 diabetes

Aims/hypothesisA precision medicine approach in type 2 diabetes could enhance targeting specific glucose-lowering therapies to individual patients most likely to benefit. We aimed to use the recently developed Bayesian causal forest (BCF) method to develop and validate an individualised treatment selection algorithm for two major type 2 diabetes drug classes, sodium–glucose cotransporter 2 inhibitors (SGLT2i) and glucagon-like peptide-1 receptor agonists (GLP1-RA).MethodsWe designed a predictive algorithm using BCF to estimate individual-level conditional average treatment effects for 12-month glycaemic outcome (HbA1c) between SGLT2i and GLP1-RA, based on routine clinical features of 46,394 people with type 2 diabetes in primary care in England (Clinical Practice Research Datalink; 27,319 for model development, 19,075 for hold-out validation), with additional external validation in 2252 people with type 2 diabetes from Scotland (SCI-Diabetes [Tayside & Fife]). Differences in glycaemic outcome with GLP1-RA by sex seen in clinical data were replicated in clinical trial data (HARMONY programme: liraglutide [n=389] and albiglutide [n=1682]). As secondary outcomes, we evaluated the impacts of targeting therapy based on glycaemic response on weight change, tolerability and longer-term risk of new-onset microvascular complications, macrovascular complications and adverse kidney events.ResultsModel development identified marked heterogeneity in glycaemic response, with 4787 (17.5%) of the development cohort having a predicted HbA1c benefit >3 mmol/mol (>0.3%) with SGLT2i over GLP1-RA and 5551 (20.3%) having a predicted HbA1c benefit >3 mmol/mol with GLP1-RA over SGLT2i. Calibration was good in hold-back validation, and external validation in an independent Scottish dataset identified clear differences in glycaemic outcomes between those predicted to benefit from each therapy. Sex, with women markedly more responsive to GLP1-RA, was identified as a major treatment effect modifier in both the UK observational datasets and in clinical trial data: HARMONY-7 liraglutide (GLP1-RA): 4.4 mmol/mol (95% credible interval [95% CrI] 2.2, 6.3) (0.4% [95% CrI 0.2, 0.6]) greater response in women than men. Targeting the two therapies based on predicted glycaemic response was also associated with improvements in short-term tolerability and long-term risk of new-onset microvascular complications.Conclusions/interpretationPrecision medicine approaches can facilitate effective individualised treatment choice between SGLT2i and GLP1-RA therapies, and the use of routinely collected clinical features for treatment selection could support low-cost deployment in many countries.Graphical

The impact of early childhood education and medical care on children

Causal inference is a statistical approach that aims to understand and quantify the causal relationship between variables, allowing us to determine the impact of one variable on another while accounting for potential confounding factors. In this study, we chose the Infant Health and Development Program (IHDP) dataset to test whether high-quality early childhood education and medical care will enhance their cognitive and academic ability. We use some common and classic algorithms to achieve this study by calculating the CATE (Conditional Average Treatment Effect) of the dataset; if the result is a negative number, that means early treatment doesn’t improve children’s further ability; if the result is a positive number, that means treatment improves it. After fitting the targeted models, BART(Bayesian Additive Regression Trees) and Metalearners, into this dataset, we found out that all the models will get positive CATE, which means that these early treatments positively affect children’s future development. Also, we can judge which model gives us a more accurate result according to the standard variance of TE(treatment effect). The result of this study will provide us with an insightful idea about whether we should give children some treatments regarding education and medical care and which model is more suitable for this casual inference test.

Model-based standardization using multiple imputation

BackgroundWhen studying the association between treatment and a clinical outcome, a parametric multivariable model of the conditional outcome expectation is often used to adjust for covariates. The treatment coefficient of the outcome model targets a conditional treatment effect. Model-based standardization is typically applied to average the model predictions over the target covariate distribution, and generate a covariate-adjusted estimate of the marginal treatment effect.MethodsThe standard approach to model-based standardization involves maximum-likelihood estimation and use of the non-parametric bootstrap. We introduce a novel, general-purpose, model-based standardization method based on multiple imputation that is easily applicable when the outcome model is a generalized linear model. We term our proposed approach multiple imputation marginalization (MIM). MIM consists of two main stages: the generation of synthetic datasets and their analysis. MIM accommodates a Bayesian statistical framework, which naturally allows for the principled propagation of uncertainty, integrates the analysis into a probabilistic framework, and allows for the incorporation of prior evidence.ResultsWe conduct a simulation study to benchmark the finite-sample performance of MIM in conjunction with a parametric outcome model. The simulations provide proof-of-principle in scenarios with binary outcomes, continuous-valued covariates, a logistic outcome model and the marginal log odds ratio as the target effect measure. When parametric modeling assumptions hold, MIM yields unbiased estimation in the target covariate distribution, valid coverage rates, and similar precision and efficiency than the standard approach to model-based standardization.ConclusionWe demonstrate that multiple imputation can be used to marginalize over a target covariate distribution, providing appropriate inference with a correctly specified parametric outcome model and offering statistical performance comparable to that of the standard approach to model-based standardization.

Impact of access to cash remittances on cocoa yield in Southwestern Nigeria

Access to cash remittances has a positive impact on the food security of people living in developing countries. However, little is known about the impact of access to cash remittances concerning cocoa yield. Therefore, the study investigates the impact of access to cash remittances on cocoa yield. A multistage sampling technique was used to select 300 cocoa farming households for the study. The data were analysed using the exponential double hurdle model (Churdle) estimation process, inverse probability weighted regression adjustment (IPWRA), and endogenous switching regression model (ESRM). According to the ATU estimate, the average farmer would produce a yield of 0.824 times more when he/she has access to cash remittances. The conditional treatment effect (ATT), which assesses the impact of access to cash remittances on cocoa yield, was correspondingly approximately 0.828 and statistically significant at 1 %. As a result, the typical farmer who has access to cash remittances would produce a yield that is 0.828 times more than it would be if he/she did not have access to cash remittances. The study found that access to cash remittances has a positive impact on cocoa yield among farmers in rural Nigeria after adjusting for both observable and unobserved factors.

Harnessing Causal Forests for Epidemiologic Research: Key Consideration.

Assessing heterogeneous treatment effects (HTEs) is an essential task in epidemiology. The recent integration of machine learning into causal inference has provided a new, flexible tool for evaluating complex HTEs: causal forest. Jawadekar et al. (Am J Epidemiol. 2023) introduce this innovative approach and offer practical guidelines for applied users. Building on their work, this commentary provides additional insights and guidance to promote the understanding and application of causal forest in epidemiologic research. We start with conceptual clarifications, differentiating between honesty and cross-fitting, and exploring the interpretation of estimated conditional average treatment effects. We then delve into the following practical considerations not addressed by Jawadekar et al., including motivations for estimating HTEs, calibration approaches, and ways to leverage causal forest output with examples from simulated data. We conclude by outlining challenges to consider for future advancements and applications of causal forest in epidemiological research.