Overview of modern approaches for identifying and evaluating heterogeneous treatment effects from clinical data.

BackgroundRetrospective exploratory analyses of randomised controlled trials (RCTs) seeking to identify treatment effect heterogeneity (TEH) are prone to bias and false positives. Yet the desire to learn all we can from exhaustive data measurements on trial participants motivates the inclusion of such analyses within RCTs. Moreover, widespread advances in machine learning (ML) methods hold potential to utilise such data to identify subjects exhibiting heterogeneous treatment response.MethodsWe present a novel analysis strategy for detecting TEH in randomised data using ML methods, whilst ensuring proper control of the false positive discovery rate. Our approach uses random data partitioning with statistical or ML-based prediction on held-out data. This method can test for both crossover TEH (switch in optimal treatment) and non-crossover TEH (systematic variation in benefit across patients). The former is done via a two-sample hypothesis test measuring overall predictive performance. The latter is done via ‘stacking’ the ML predictors alongside a classical statistical model to formally test the added benefit of the ML algorithm. An adaptation of recent statistical theory allows for the construction of a valid aggregate p value. This testing strategy is independent of the choice of ML method.ResultsWe demonstrate our approach with a re-analysis of the SEAQUAMAT trial, which compared quinine to artesunate for the treatment of severe malaria in Asian adults. We find no evidence for any subgroup who would benefit from a change in treatment from the current standard of care, artesunate, but strong evidence for significant TEH within the artesunate treatment group. In particular, we find that artesunate provides a differential benefit to patients with high numbers of circulating ring stage parasites.ConclusionsML analysis plans using computational notebooks (documents linked to a programming language that capture the model parameter settings, data processing choices, and evaluation criteria) along with version control can improve the robustness and transparency of RCT exploratory analyses. A data-partitioning algorithm allows researchers to apply the latest ML techniques safe in the knowledge that any declared associations are statistically significant at a user-defined level.

Estimating heterogeneous survival treatment effects of lung cancer screening approaches: A causal machine learning analysis

NM2 Using Randomized and Observational DATA to Predict Heterogeneous Treatment Effects

Developing Theory Using Machine Learning Methods

This dissertation addresses two significant issues in the effective application of machine learning algorithms and models for the physical and engineering sciences. The first is the broad challenge of automated modeling of data across different processes in a physical system. The second is the dilemma of obtaining insightful interpretations on the relationships between the inputs and outcome of a system as inferred from complex, black box machine learning models.Automated Geometric Shape Deviation Modeling for Additive Manufacturing SystemsAdditive manufacturing systems possess an intrinsic capability for one-of-a-kind manufacturing of a vast variety of shapes across a wide spectrum of processes. One major issue in AM systems is geometric accuracy control for the inevitable shape deviations that arise in AM processes. Current effective approaches for shape deviation control in AM involve the specification of statistical or machine learning deviation models for additively manufactured products. However, this task is challenging due to the constraints on the number of test shapes that can be manufactured in practice, and limitations on user efforts that can be devoted for learning deviation models across different shape classes and processes in an AM system. We develop an automated, Bayesian neural network methodology for comprehensive shape deviation modeling in an AM system. A fundamental innovation in this machine learning method is our new and connectable neural network structures that facilitate the transfer of prior knowledge and models on deviations across different shape classes and AM processes. Several case studies on in-plane and out-of-plane deviations, regular and free-form shapes, and different settings of lurking variables serve to validate the power and broad scope of our methodology, and its potential to advance high-quality manufacturing in an AM system.Interpretable Machine LearningMachine learning algorithms and models constitute the dominant set of predictive methods for a wide range of complex, real-world processes. However, interpreting what such methods effectively infer from data is difficult in general. This is because their typical black box natures possess a limited ability to directly yield insights on the underlying relationships between inputs and the outcome for a process. We develop methodologies based on new predictive comparison estimands that effectively enable one to ``mine’’ machine learning models, in the sense of (a) interpreting their inferred associations between inputs and/or functional forms of inputs with the outcome, (b) identifying the inputs that they effectively consider relevant, and (c) interpreting the inferred conditional and two-way associations of the inputs with the outcome. We establish Fisher consistent estimators, and their corresponding standard errors, for our new estimands under a condition on the inputs' distributions. The significance of our predictive comparison methodology is demonstrated with a wide range of simulation and case studies that involve Bayesian additive regression trees, neural networks, and support vector machines. Our extended study of interpretable machine learning for AM systems demonstrates how our method can contribute to smarter advanced manufacturing systems, especially as current machine learning methods for AM are lacking in their ability to yield meaningful engineering knowledge on AM processes.

Biomarkers are known to be the key driver behind targeted cancer therapies by either stratifying the patients into risk categories or identifying patient subgroups most likely to benefit. However, the ability of a biomarker to stratify patients relies heavily on the type of clinical endpoint data being collected. Of particular interest is the scenario when the biomarker involved is a continuous one where the challenge is often to identify cut-offs or thresholds that would stratify the population according to the level of clinical outcome or treatment benefit. On the other hand, there are well-established Machine Learning (ML) methods such as the Support Vector Machines (SVM) that classify data, both linear as well as non-linear, into subgroups in an optimal way. SVMs have proven to be immensely useful in data-centric engineering and recently researchers have also sought its applications in healthcare. Despite their wide applicability, SVMs are not yet in the mainstream of toolkits to be utilised in observational clinical studies or in clinical trials. This research investigates the very role of SVMs in stratifying the patient population based on a continuous biomarker across a variety of datasets. Based on the mathematical framework underlying SVMs, we formulate and fit algorithms in the context of biomarker stratified cancer datasets to evaluate their merits. The analysis reveals their superior performance for certain data-types when compared to other ML methods suggesting that SVMs may have the potential to provide a robust yet simplistic solution to stratify real cancer patients based on continuous biomarkers, and hence accelerate the identification of subgroups for improved clinical outcomes or guide targeted cancer therapies.

Optimizing operational decisions, routine actions within some business or operational process, is a key challenge across a variety of domains and application areas. The increasing availability of data, computational power, and advanced machine learning (ML) algorithms offers exciting opportunities for data-driven decision support. To advance the potential of ML for optimizing operational decision-making, we explore two research directions, aiming to develop ML models that are decision-focused and causal. This dissertation presents several developments in machine learning in these two areas. ML is effective at making predictions from historical data: for example, estimating a transaction's fraud probability by comparing it to past cases. However, decision-makers not only need to consider these predictions, but also the operational context. For example, the decision-maker uses predicted fraud probabilities to determine which transactions to investigate, while aiming to minimize monetary losses due to fraud and considering the available capacity of the fraud investigations team. Predictions can help reduce uncertainty (e.g., by predicting the fraud probability), but standard ML models are prediction-focused, instead of decision-focused. This distinction involves two challenges for data-driven decision-making. First, prediction-focused models prioritize predictive accuracy instead of the resulting decision quality (e.g., fraud losses recovered by the bank). Second, these models fail to account for operational constraints, such as the available investigation capacity. Decision-focused learning aims to improve data-driven decision-making by addressing these issues and incorporating the operational context into the optimization of ML models. In this dissertation, we analyze cost-sensitive learning within this prediction-optimization framework and evaluate general strategies for making cost-optimal decisions with ML. Additionally, we propose a novel ML method for optimal decision-making under capacity constraints based on learning to rank. To make effective decisions, a decision-maker has to estimate the causal effect of possible interventions in order to choose actions that achieve the desired outcome. Unfortunately, standard ML models identify correlations in the data instead of causal relationships. Because of this, these models cannot guarantee the effectiveness of decisions made based on their predictions. Causal inference provides a formal framework for reasoning about causality and identifying causal effects from data. This dissertation explores the intersection of causality and ML. First, we illustrate the potential of causal ML for optimizing preventive maintenance. Next, we propose novel causal ML methods for predicting causal effect distributions and for addressing informative sampling when predicting treatment outcomes over time. We also argue for a practical, end-to-end perspective for building ML pipelines for causal inference and propose an automated framework doing so. Finally, we combine decision-focused learning with causal inference by introducing ranking metalearners to optimize treatment decisions under capacity constraints.

In this paper, we provide efficient estimators and honest confidence bands for a variety of treatment effects including local average (LATE) and local quantile treatment effects (LQTE) in data-rich environments. We can handle very many control variables, endogenous receipt of treatment, heterogeneous treatment effects, and function-valued outcomes. Our framework covers the special case of exogenous receipt of treatment, either conditional on controls or unconditionally as in randomized control trials. In the latter case, our approach produces efficient estimators and honest bands for (functional) average treatment effects (ATE) and quantile treatment effects (QTE). To make informative inference possible, we assume that key reduced form predictive relationships are approximately sparse. This assumption allows the use of regularization and selection methods to estimate those relations, and we provide methods for post-regularization and post-selection inference that are uniformly valid (honest) across a wide-range of models. We show that a key ingredient enabling honest inference is the use of orthogonal or doubly robust moment conditions in estimating certain reduced form functional parameters. We illustrate the use of the proposed methods with an application to estimating the effect of 401(k) eligibility and participation on accumulated assets. The results on program evaluation are obtained as a consequence of more general results on honest inference in a general moment condition framework, which arises from structural equation models in econometrics. Here too the crucial ingredient is the use of orthogonal moment conditions, which can be constructed from the initial moment conditions. We provide results on honest inference for (function-valued) parameters within this general framework where any high-quality, modern machine learning methods can be used to learn the nonparametric/high-dimensional components of the model. These include a number of supporting auxilliary results that are of major independent interest: namely, we (1) prove uniform validity of a multiplier bootstrap, (2) offer a uniformly valid functional delta method, and (3) provide results for sparsity-based estimation of regression functions for function-valued outcomes.

In causal inference, estimating heterogeneous treatment effects (HTE) is critical for identifying how different subgroups respond to interventions, with broad applications in fields such as precision medicine and personalized advertising. Although HTE estimation methods aim to improve accuracy, how to provide explicit subgroup descriptions remains unclear, hindering data interpretation and strategic intervention management. In this paper, we propose CURLS, a novel rule learning method leveraging HTE, which can effectively describe subgroups with significant treatment effects. Specifically, we frame causal rule learning as a discrete optimization problem, finely balancing treatment effect with variance and considering the rule interpretability. We design an iterative procedure based on the minorize-maximization algorithm and solve a submodular lower bound as an approximation for the original. Quantitative experiments and qualitative case studies verify that compared with state-of-the-art methods, CURLS can find subgroups where the estimated and true effects are 16.1% and 13.8% higher and the variance is 12.0% smaller, while maintaining similar or better estimation accuracy and rule interpretability. Code is available at https://osf.io/zwp2k/.

Ultra-high-performance concrete (UHPC) bonded to normal concrete (NC) can significantly enhance the mechanical performance of UHPC–NC composite structures, and the interface shear strength is a crucial indicator for assessing the bonding performance. In this study, interpretable machine learning (ML) methods were used to analyse the effects of different parameters on interface shear strength. A database consisting of 305 UHPC–NC shear tests was created, and the isolation forest algorithm was applied to filter outliers. Subsequently, four ML models were trained to predict the interface shear strength of UHPC–NC composite structures. Among them, the extreme gradient boosting (XGBoost) model demonstrated the highest prediction accuracy, achieving an R2 value of 0.95. Shapley additive explanations (SHAP), partial dependence plots (PDP) and individual conditional expectation (ICE) were used for feature importance analysis, aiding in the interpretation of the ‘black box’ nature of the ML models. The results demonstrate that the normal compressive stress at the interface is the most influential factor affecting interfacial shear strength. Finally, a physically meaningful predictive equation for the interface shear strength of UHPC–NC composite structures was proposed based on the XGBoost model combined with curve fitting. This equation enhances the prediction accuracy of interface shear strength for UHPC–NC structures and offers deeper insights into the model’s decision making process.

Treatment effect estimation from observational data relies on auxiliary prediction exercises. This chapter presents recent developments in the econometrics literature showing that machine learning methods can be fruitfully applied for this purpose. The double machine learning (DML) approach is concerned primarily with selecting the relevant control variables and functional forms necessary for the consistent estimation of an average treatment effect. We explain why the use of orthogonal moment conditions is crucial in this setting. Another, somewhat distinct, strand of the literature focuses on treatment effect heterogeneity through the discovery of the conditional average treatment effect (CATE) function. Here we distinguish between methods aimed at estimating the entire function and those that project it on a pre-specified coordinate. We also present an empirical application that illustrates some of the methods.

Artificial intelligence has been applied in wildfire science and management since the 1990s, with early applications including neural networks and expert systems. Since then, the field has rapidly progressed congruently with the wide adoption of machine learning (ML) methods in the environmental sciences. Here, we present a scoping review of ML applications in wildfire science and management. Our overall objective is to improve awareness of ML methods among wildfire researchers and managers, as well as illustrate the diverse and challenging range of problems in wildfire science available to ML data scientists. To that end, we first present an overview of popular ML approaches used in wildfire science to date and then review the use of ML in wildfire science as broadly categorized into six problem domains, including (i) fuels characterization, fire detection, and mapping; (ii) fire weather and climate change; (iii) fire occurrence, susceptibility, and risk; (iv) fire behavior prediction; (v) fire effects; and (vi) fire management. Furthermore, we discuss the advantages and limitations of various ML approaches relating to data size, computational requirements, generalizability, and interpretability, as well as identify opportunities for future advances in the science and management of wildfires within a data science context. In total, to the end of 2019, we identified 300 relevant publications in which the most frequently used ML methods across problem domains included random forests, MaxEnt, artificial neural networks, decision trees, support vector machines, and genetic algorithms. As such, there exists opportunities to apply more current ML methods — including deep learning and agent-based learning — in the wildfire sciences, especially in instances involving very large multivariate datasets. We must recognize, however, that despite the ability of ML models to learn on their own, expertise in wildfire science is necessary to ensure realistic modelling of fire processes across multiple scales, while the complexity of some ML methods such as deep learning requires a dedicated and sophisticated knowledge of their application. Finally, we stress that the wildfire research and management communities play an active role in providing relevant, high-quality, and freely available wildfire data for use by practitioners of ML methods.

Recently, machine learning (ML) methods have been used in causal inference to estimate treatment effects in order to reduce concerns for model mis-specification. However, many ML methods require that all confounders are measured to consistently estimate treatment effects. In this paper, we propose a family of ML methods that estimate treatment effects in the presence of cluster-level unmeasured confounders, a type of unmeasured confounders that are shared within each cluster and are common in multilevel observational studies. We show through simulation studies that our proposed methods are robust from biases from unmeasured cluster-level confounders in a variety of multilevel observational studies. We also examine the effect of taking an algebra course on math achievement scores from the Early Childhood Longitudinal Study, a multilevel observational educational study, using our methods. The proposed methods are available in the CURobustML R package.

A dose deposition matrix (DDM) prediction method using several voxel features and a machine learning (ML) approach is proposed for plan optimization in radiation therapy. Head and lung cases with the inhomogeneous medium are used as training and testing data. The prediction model is a cascade forward backprop neural network where the input is the features of the voxel, including 1) voxel to body surface distance along the beamlet axis, 2) voxel to beamlet axis distance, 3) voxel density, 4) heterogeneity corrected voxel to body surface distance, 5) heterogeneity corrected voxel to beamlet axis, and (6) the dose of voxel obtained from the pencil beam (PB) algorithm. The output is the predicted voxel dose corresponding to a beamlet. The predicted DDM was used for plan optimization (ML method) and compared with the dose of MC-based plan optimization (MC method) and the dose of pencil beam-based plan optimization (PB method). The mean absolute error (MAE) value was calculated for full volume relative to the dose of the MC method to evaluate the overall dose performance of the final plan. For patient with head tumor, the ML method achieves MAE value 0.49 × 10-4 and PB has MAE 1.86 × 10-4. For patient with lung tumor, the ML method has MAE 1.42 × 10-4 and PB has MAE 3.72 × 10-4. The maximum percentage difference in PTV dose coverage (D98) between ML and MC methods is no more than 1.2% for patient with head tumor, while the difference is larger than 10% using the PB method. For patient with lung tumor, the maximum percentage difference in PTV dose coverage (D98) between ML and MC methods is no more than 2.1%, while the difference is larger than 16% using the PB method. In this work, a reliable DDM prediction method is established for plan optimization by applying several voxel features and the ML approach. The results show that the ML method based on voxel features can obtain plans comparable to the MC method and is better than the PB method in achieving accurate dose to the patient, which is helpful for rapid plan optimization and accurate dose calculation. Establishment of a new machine learning method based on the relationship between the voxel and beamlet features for dose deposition matrix prediction in radiation therapy.

Observational studies provide the only evidence on the effectiveness of interventions when randomized controlled trials (RCTs) - apart from the initial RCT that establishes the efficacy of a treatment compared to a placebo - are impractical due to cost, ethical concerns, or time constraints. While many methodologies aim to draw causal inferences from observational data, there is a growing trend to model observational study designs after hypothetical or existing RCTs, a strategy known as "target trial emulation." Despite its potential, causal inference through target trial emulation is challenging because it cannot fully address the confounding bias inherent in real-world data due to the lack of randomization. In this work, we present a novel framework for target trial emulation that aims to overcome several key limitations, including confounding bias. The framework proceeds as follows: First, we apply the eligibility criteria of a specific trial to an observational cohort derived from real-world data. We then "correct" this cohort by extracting a subset that, through optimization techniques, matches both the distribution of covariates and baseline prognoses (i.e., the prognosis in the trial's control group) of the target RCT. Next, we address unmeasured confounding by adjusting the prognosis estimates of the treated group to align with those observed in the trial, using cost-sensitive counterfactual models. Following trial emulation, we go a step further by leveraging the emulated cohort to train optimal decision trees, developed by our team, to identify subgroups of patients exhibiting heterogeneity in treatment effects (HTE). The absence of confounding is verified using two external models, and the validity of the treatment effects estimated by our framework is independently confirmed by the team responsible for the original trial we emulate. To our knowledge, this is the first framework to successfully address both observed and unobserved confounding, a challenge that has historically limited the use of randomized trial emulation and causal inference in general since the 1950s. Additionally, our framework holds promise in advancing precision or personalized medicine by identifying patient subgroups that benefit most from specific treatments.