|
Abstract:
|
Digitalization of patient records and increasing computational power have led to a paradigm shift in the field of medical decision-making from one-size-fits-all interventions to data-driven treatment strategies optimised for particular sub-populations or individuals. In this research, we make use of statistical techniques for counterfactual prediction to support patient-centred treatment decisions on the initiation of renal replacement therapy. For this, we contrast existing techniques for counterfactual prediction and for the estimation of heterogeneous treatment effects, such as the DR-learner, the X-learner and an inverse probability weighted (IPW) approach, with a novel proposal in settings with and without runtime confounding. This proposal combines the strengths of the IPW-approach with those of the DR-learner, in the sense that it is generic, is guaranteed to deliver predictions in the range of the counterfactual outcome mean, and delivers oracle behaviour as a result of being based on an orthogonal loss function (where orthogonality is relative to the infinite-dimensional propensity score and conditional outcome mean).
|
