Assessing the effectiveness of sequences of treatments using sequential patterns

Abstract

In this paper, we tackle the issue of assessing the effectiveness of sequences of treatments by introducing the concept of state-changing sequential patterns. Our proposal aims at identifying sequential patterns in an environment where certain actions are taken for patients (medical procedures, administration of pharmaceuticals, etc.) while simultaneously measuring some indicator of their health (e.g., blood pressure). We propose to combine the information about the events with the information about the states of the patients targeted by these events when mining for sequential patterns. To be able to properly interpret the changes in states as outcomes of sequences of events, we rely on the concept of a control group known from clinical trials. We illustrate the usefulness of our proposal with a proof-of-concept experiment.