Patient-centric approach to maximise CPAP therapy acceptance: AI-driven design, delivery, and monitoring of interventions

Obstructive Sleep Apnea (OSA) is a sleeping disorder that manifests itself in various ways. In the absence of treatment, it has a detrimental effect on the cardiovascular system, causes excessive daytime sleepiness, and increases motor vehicle accidents. It is characterised by repeated narrowing (hypopnea) and closure (apnea) of the upper airway during sleep. In adults with OSA, continuous positive airway pressure (CPAP) therapy is the first-line medical treatment. The primary requirement for CPAP therapy to be practical is that the patient adheres to the therapy. The patient must utilise the CPAP equipment for at least 4 hours per night to see clinical improvement. However, according to research assessing this level of adherence, CPAP therapy has the lowest amount of compliance when compared to 17 other therapies such as HIV, cancer...With the advent of big data and highly accurate monitoring data in-homecare, new opportunities are open to tackle these ongoing issues. One of these opportunities is a complete understanding of the patient to give targeted interventions that empower the patient in the CPAP therapy. This research project, which is being conducted in conjunction with the homecare provider Linde Homecare France, falls within the scope of these new opportunities. In this research project, we have four core objectives, namely: 1) patient characterisation, 2) personalised intervention design and delivery, 3) verification, validation and monitoring of the interventions, and 4) automatic models readjustment based on experts feedback. To fully comply with these four objectives, we address the following question: "How to leverage CPAP monitoring data with telemedicine to deliver patient-centric interventions to increase the CPAP therapy adherence level in the long-term and how should an expert-assisted system be developed and deployed on a wide range of patients for a personalised patient management ?" . To address this research question, we developed the novel framework Homecare Intervention as a Service (HIaaS). The HIaaS framework defines the patient by modelling multiple patient profiles and introducing a patient pathway concept tailored to homecare. We determine the most effective interventions based on the patients' compliance barriers. To deepen the patient-centric approach, we add a level of personalisation to each intervention. The system keeps track of both intervention use and patient perceptions of the interventions. Finally, throughout the entire process beforehand, the HIaaS framework proposes a verification mechanism and a validation mechanism for the intermediate results. Three types of experts are responsible for these mechanisms: sleep experts, homecare managers, and homecare providers. As a result, each expert is bounded to their field of expertise. We solicit expert feedback throughout the verification and validation process. We then propagate this feedback to the various models in the appropriate manner.