Combining the Clinical and Operational Perspectives in Heterogeneous Treatment Effect Inference in Healthcare Processes

Abstract

Recent developments in causal machine learning open perspectives for new approaches that support decision-making in healthcare processes using causal models. In particular, Heterogeneous Treatment Effect (HTE) inference enables the estimation of causal treatment effects for individual cases, offering great potential in a process mining context. At the same time, HTE literature typically focuses on clinical outcome measures, disregarding process efficiency. This paper shows the potential of jointly considering the clinical and operational effects of treatments in the context of healthcare processes. Moreover, we present a simple pipeline that makes existing HTE machine learning techniques directly applicable to event logs. Besides these conceptual contributions, a proof-of-concept application starting from the publicly available sepsis event log is outlined, forming the basis for a critical reflection regarding HTE estimation in a process mining context.

Mining Statistical Relations for Better Decision Making in Healthcare Processes

“…What these approaches have in common is that they only identify potential causal relations. More recent approaches explicitly account for causation, for example, by combining action-rule mining with uplift trees [6] or using neural networks [7]. Both Verboven and Martin [7] and Bozorgi et al [6] account for confounding as long as confounders are included in the data.…”

“…We adopt the same data generating process as [7], [25] to generate the PFA functions. This means that the variables and response assignment are the same as in the original data set, so the original response propensities (P(W i = 1 | X i = x)) are retained and selection bias persists.…”

“…In process mining literature, various techniques have been proposed that aim to discover such relations. They can be subdivided into three main groups: machine-learning-based approaches [5]- [7], statistics-based approaches [8]- [10], and context-based approaches [11], [12]. However, none of the existing techniques can deal with confounding variables at the event level.…”