<p>Using the Causal Inference Framework to Support Individualized Drug Treatment Decisions Based on Observational Healthcare Data</p>

Abstract

When healthcare professionals have the choice between several drug treatments for their patients, they often experience considerable decision uncertainty because many decisions simply have no single “best” choice. The challenges are manifold and include that guideline recommendations focus on randomized controlled trials whose populations do not necessarily correspond to specific patients in everyday treatment. Further reasons may be insufficient evidence on outcomes, lack of direct comparison of distinct options, and the need to individually balance benefits and risks. All these situations will occur in routine care, its outcomes will be mirrored in routine data, and could thus be used to guide decisions. We propose a concept to facilitate decision-making by exploiting this wealth of information. Our working example for illustration assumes that the response to a particular (drug) treatment can substantially differ between individual patients depending on their characteristics (heterogeneous treatment effects, HTE), and that decisions will be more precise if they are based on real-world evidence of HTE considering this information. However, such methods must account for confounding by indication and effect measure modification, eg, by adequately using machine learning methods or parametric regressions to estimate individual responses to pharmacological treatments. The better a model assesses the underlying HTE, the more accurate are predicted probabilities of treatment response. After probabilities for treatment-related benefit and harm have been calculated, decision rules can be applied and patient preferences can be considered to provide individual recommendations. Emulated trials in observational data are a straightforward technique to predict the effects of such decision rules when applied in routine care. Prediction-based decision rules from routine data have the potential to efficiently supplement clinical guidelines and support healthcare professionals in creating personalized treatment plans using decision support tools.

Machine learning for tumor growth inhibition: Interpretable predictive models for transparency and reproducibility

“…This means that there are treatment effect modulators that determine how individual responses to available treatment options T may differ. 7 Consequently, we are not only interested in the average treatment effect (population mean), but in the conditional average treatment effect (CATE) in the individual patient given his or her covariates Z . Because we can only observe the outcome under one treatment in a patient, we consider the predicted potential outcomes Y * to estimate CATEs.…”

Can Machine Learning from Real-World Data Support Drug Treatment Decisions? A Prediction Modeling Case for Direct Oral Anticoagulants

“…Model development: Estimation of ITEs for treatment recommendations. Using the causal inference framework, 2 we estimated the potential outcomes of each patient under the potential treatment with apixaban and rivaroxaban. The ITE is thus the difference between the individual outcome probabilities under apixaban and rivaroxaban.…”

“…When choosing between several drug treatments for their patients, physicians and health care professionals are often unsure which option is “best.” 1 , 2 Generalization of guidelines or evidence from randomized controlled trials (RCTs) cannot overcome this uncertainty because they validly estimate average treatment effects between groups in a study population but cannot explain potentially heterogeneous treatment effects of single individuals. 3 , 4 Thus, patient characteristics can modulate average effects and lead to different individual treatment effects (ITEs) at the patient level.…”

“… 3 , 4 Thus, patient characteristics can modulate average effects and lead to different individual treatment effects (ITEs) at the patient level. 5 Good assessment of ITEs is critical when physicians seek to provide individualized care to their patients, 2 , 6 especially in situations in which significant benefits must be weighed against serious harms of treatment options, such as the prevention of thromboembolic events in patients with nonvalvular atrial fibrillation (AF) with direct oral anticoagulants (DOACs) and the associated bleeding risk. 7 , 8 Here, treatment decisions could become more precise if real-world evidence was processed using machine learning to provide personalized recommendations for decision making with DOACs.…”

Causal Machine Learning for Healthcare and Precision Medicine

“…This results in confounding, which is best mitigated by the use of an RCT 2.3. Performing an RCT as detailed in Section 2.3, however, is not always feasible, and causal inference techniques can be used to estimate the causal effect of treatment from observational data [Meid et al, 2020]. A number of assumptions need to hold in order for the treatment effect to be identifiable from observational data [Lesko et al, 2017, Imbens andRubin, 2015].…”