Prescriptive medicine : clinical recommendation tools : towards automated ideal treatment pathways in healthcare

Details

Supervisors

Absil, Pierre-Antoine

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en science des données, orientation technologies de l'information, à finalité spécialisée

Abstract

During the follow-up of patients, MoveUp, through a team of physiotherapists, makes recommendations to improve the rehabilitation of patients after knee and hip surgery. These recommendations aim to encourage patients to promote physical activity by setting maximum and minimum numbers of steps to be achieved on a daily basis. To get feedback from their experience, patients are asked to fill out daily questionnaires and are provided with an activity tracker that records their daily number of steps. The aim of this thesis is to propose an artificial intelligence solution capable of assisting physiotherapists in the process of recommending the steps to be taken on a daily basis for patients who have undergone mainly knee and hip arthroplasties with a horizon set to a one-week period. In addition to recall bias, the data made available has many deficiencies in terms of its composition and its heavy reliance on people (physiotherapists as well as patients) who are responsible for remembering to fill in daily questionnaires and update recommendations. However, the use of physiotherapists' recommendations remains the only clinical approach available. Before proceeding to the detailed exploration of the machine learning algorithms, we will verify the hypothesis of the existence of groups in the set of patients during the evolution of their treatment. The objective of this procedure is to discover, if possible, the existence of a pattern that would allow the implementation of techniques specific to each group and to make recommendations targeted to specific patient profiles. Next, regression models are tested using data from patients and physiotherapists in order to make recommendations in the first instance. Secondly, an alternative way of recommending numbers of steps is proposed that is based solely on patient data and completely ignores physiotherapist recommendations.