Understanding eHealth Acceptance Among COPD Patients: An Innovative Machine Learning Approach to Identify Interaction Patterns and Prototypes

Background: eHealth can support patients and healthcare providers in the management of chronic illness. However, acceptance and adoption of eHealth among COPD patients is a challenge, which limits the effectiveness of these solutions in clinical practice. It is difficult for healthcare providers to know what type of health is feasible for which patient. A personalized approach is needed to support patients in using eHealth. Aim: We aim to define groups of COPD patients based on their barriers to use eHealth. These groups can be used to develop more personalised solutions. Secondary aim is to examine the possible impact of cultural differences and to look for facilitators that help patients to use eHealth, even if their digital skills are poor. Possible methodology: A cross-sectional study has been conducted in several European countries, using the survey drawn from our previous qualitative study and grounded in the Unified Theory of Acceptance and Use of Technology (UTAUT) framework. The survey was distributed both online and offline, depending on the patients' preferences. In total, there are 545 responses in 08 countries. Four commonly used tree-based models (Classification and regression tree (CART), random forest, LightGBM and XGBoost) are employed to predict the acceptance of eHealth utilization. The efficacy of models were evaluated using 10-fold cross-validation. Finally, the SHapley Additive exPlanations (SHAP) tool was used for the best model to determine the important features that influence eHealth acceptance. To that end, the prototypes that are drawn from this model can be applied to personalize eHealth solutions. This study is funded by NWO (Nederlandse Organisatie voor Wetenschappelijk Onderzoek). Questions to discuss: What are the advantages and disadvantages of using machine learning-based than traditional statistical models? How can the results of this study be useful for daily clinical practice?