Estimating heterogeneous survival treatment effect in observational data using machine learning

Hu, Liangyuan; Ji, Jiayi; Li, Fan

doi:10.1002/sim.9090

Cited by 45 publications

(58 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also adapt the popularly used inverse probability weighting method into the setting of clustered and censored survival data to form two comparison methods: inverse probability of treatment weighting with the random-intercept Cox regression model (IPW-riCox) and doubly robust random-intercept additive hazards model (DR-riAH). In addition, we consider another outcome modeling based method, random-intercept generalized additive proportional hazards model (riGAPH), 36,3 that is flexible at capturing nonlinear relationships. We use the counterfactual survival curve as the basis to objectively compare methods.…”

Section: Comparison Methodsmentioning

confidence: 99%

“…We similarly define ( ) as the counterfactual censoring time under treatment . Throughout, we maintained the standard assumptions for drawing causal inference with observational clustered survival data: 3,17,18…”

Section: =1mentioning

confidence: 99%

“…Alternatively, one can define causal estimands based on functionals (e.g., median) of the counterfactual survival curves. 3 In this paper, we focus on the average treatment effect (ATE) defined either over the sample or the population. Consider a pairwise comparison between treatments and ′ .…”

Section: =1mentioning

confidence: 99%

“…The analysis dataset was drawn from the NCDB and included 23058 high-risk localized prostate cancer patients who were older than 65 when diagnosed between 2004 and 2015. Among these patients, 14237 received RP, 6683 undertook EBRT+AD with at least 7920 cGy EBRT dose 3 and 2138 were treated with EBRT+brachy±AD. Included in the dataset are pre-treatment 4.…”

Section: Application To Prostate Cancermentioning

confidence: 99%

See 3 more Smart Citations

A flexible approach for causal inference with multiple treatments and clustered survival outcomes

Ji¹,

Ennis²,

Hogan³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

When drawing causal inferences about the effects of multiple treatments on clustered survival outcomes using observational data, we need to address implications of the multilevel data structure, multiple treatments, censoring and unmeasured confounding for causal analyses. Few off-the-shelf causal inference tools are available to simultaneously tackle these issues. We develop a flexible random-intercept accelerated failure time model, in which we use Bayesian additive regression trees to capture arbitrarily complex relationships between censored survival times and pre-treatment covariates and use the random intercepts to capture cluster-specific main effects. We develop an efficient Markov chain Monte Carlo algorithm to draw posterior inferences about the population survival effects of multiple treatments and examine the variability in cluster-level effects. We further propose an interpretable sensitivity analysis approach to evaluate the sensitivity of drawn causal inferences about treatment effect to the potential magnitude of departure from the causal assumption of no unmeasured confounding. Expansive simulations empirically validate and demonstrate good practical operating characteristics of our proposed methods. Applying the proposed methods to a dataset on older high-risk localized prostate cancer patients drawn from the National Cancer Database, we evaluate the comparative effects of three treatment approaches on patient survival, and assess the ramifications of potential unmeasured confounding. The methods developed in this work are readily available in the R package riAFTBART.

show abstract

Section: Comparison Methodsmentioning

confidence: 99%

Section: =1mentioning

confidence: 99%

Section: =1mentioning

confidence: 99%

Section: Application To Prostate Cancermentioning

confidence: 99%

See 2 more Smart Citations

A flexible approach for causal inference with multiple treatments and clustered survival outcomes

Ji¹,

Ennis²,

Hogan³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Given the value of 𝛽 * is set to log(4), we considered 9 different scenarios by varying the sample size and average treatment effect. In these 9 scenarios, the sizes of survey subjects sampled were 3,000 ( 31 We maintained the sample size of 9,000 and set the values (𝛽 𝑡 , 𝛽 * ) = (−0.2, log(2.46)) for all three levels of overlap in propensity score. We set the value of 𝜓 to 1, 2.5 and 5 to demonstrate when there is strong, moderate, and weak overlap, depicted by the three panels respectively in Figure 1.…”

Section:  Simulation Studymentioning

confidence: 99%

Propensity Score Weighting With Survey Weighted Data When Outcomes Are Binary

Yang

Cuerden

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Background Despite that propensity score methods have been widely adopted in observational studies, few studies have focused on how to properly estimate and incorporate propensity scores in complex survey data settings. Particularly, it is lacking that the research on propensity score-based weighting methods when estimating the treatment effect using complex survey data with binary outcomes. The objective of this study is therefore to compare three propensity score weighting methods for complex survey data when outcomes are binary. MethodsA simulation study was used to compare three propensity score weighting approaches for estimating treatment effects using survey data: 1) no survey weights in the propensity score model or the outcome model; 2) survey weights in the outcome model only; 3) survey weights in both models. Each of the three methods is applied in the context of four measures of the treatment effect: the sample average treatment effect (SATE), the population average treatment effect (PATE), the sample average treatment effect on the treated (SATT), and the population average treatment effect on the treated (PATT). The methods are compared in terms of mean relative bias and coverage probability under different scenarios by varying combinations of sample size and treatment effect, degrees of model misspecification and levels of overlap in propensity score. In addition, using the 2015 National Health Interview Survey (NHIS) data as a real data example, each method is employed to estimate the effect of provider-patient discussion about smoking on smoking cessation. ResultsThe methods which account for the survey weight outperform the unweighted method as the degree of misspecification increases, regardless of the sample size, treatment effect and level of overlap in propensity score. The performance of the two weighted methods, where the survey weight is incorporated in the outcome model or both the propensity score model and the outcome model, are similar. Conclusions The propensity score weighting methods accounting for the survey weight are necessary for estimating the population-level treatment effects. This paper provides guidance for selecting an appropriate propensity score weighting method based on the estimand of interest when using complex survey data with binary outcomes.

show abstract

A new method for clustered survival data: Estimation of treatment effect heterogeneity and variable selection

2023

Biometrical J

Self Cite

View full text Add to dashboard Cite

We recently developed a new method random‐intercept accelerated failure time model with Bayesian additive regression trees (riAFT‐BART) to draw causal inferences about population treatment effect on patient survival from clustered and censored survival data while accounting for the multilevel data structure. The practical utility of this method goes beyond the estimation of population average treatment effect. In this work, we exposit how riAFT‐BART can be used to solve two important statistical questions with clustered survival data: estimating the treatment effect heterogeneity and variable selection. Leveraging the likelihood‐based machine learning, we describe a way in which we can draw posterior samples of the individual survival treatment effect from riAFT‐BART model runs, and use the drawn posterior samples to perform an exploratory treatment effect heterogeneity analysis to identify subpopulations who may experience differential treatment effects than population average effects. There is sparse literature on methods for variable selection among clustered and censored survival data, particularly ones using flexible modeling techniques. We propose a permutation‐based approach using the predictor's variable inclusion proportion supplied by the riAFT‐BART model for variable selection. To address the missing data issue frequently encountered in health databases, we propose a strategy to combine bootstrap imputation and riAFT‐BART for variable selection among incomplete clustered survival data. We conduct an expansive simulation study to examine the practical operating characteristics of our proposed methods, and provide empirical evidence that our proposed methods perform better than several existing methods across a wide range of data scenarios. Finally, we demonstrate the methods via a case study of predictors for in‐hospital mortality among severe COVID‐19 patients and estimating the heterogeneous treatment effects of three COVID‐specific medications. The methods developed in this work are readily available in the package .

show abstract

Estimating heterogeneous survival treatment effect in observational data using machine learning

Cited by 45 publications

References 64 publications

A flexible approach for causal inference with multiple treatments and clustered survival outcomes

A flexible approach for causal inference with multiple treatments and clustered survival outcomes

Propensity Score Weighting With Survey Weighted Data When Outcomes Are Binary

A new method for clustered survival data: Estimation of treatment effect heterogeneity and variable selection

Contact Info

Product

Resources

About