A Model to Patient Abandonment Prediction in the Pulmonary Rehabilitation

• Wesllei Felipe Heckler Universidade Feevale https://orcid.org/0000-0002-2108-3752
• Juliano Varella de Carvalho Universidade Feevale https://orcid.org/0000-0001-7795-0475
• Cássia Cinara da Costa Universidade Feevale
• Jorge Luis Victória Barbosa UNISINOS https://orcid.org/0000-0002-0358-2056

DOI: https://doi.org/10.5753/sbsi.2020.13776

Resumo

Respiratory diseases reach one of the main systems of the human body and affect a large part of the Brazilian population. In severe cases, they can limit the functionalities and the muscular strength of the disease carriers, decreasing their quality of life, because of the impact on simple daily activities. Pulmonary Rehabilitation Programs help to treat these diseases. The Pulmonary Rehabilitation Program of Feevale University helps patients with chronic respiratory diseases in the community and aims to improve their quality of life through the development of educational and assistance actions. The information about the patients and the treatment results are stored in a database. The amount of data disturbs the analysis of results. Through of a systematic review, it was possible to see that does not exist a model to predict abandonment trends of patients in Pulmonary Rehabilitation Programs. Therefore, this paper proposes a model to apply machine learning techniques in the database of the Pulmonary Rehabilitation Program of Feevale University. The model aims to identify the abandonment trends of patients entering the treatment and to extract knowledge about this database to contribute to the application of pulmonary rehabilitation treatment. In addition, this paper presents the development of a tool to apply the model proposed and provides visualizations to assist reading the data and results by healthcare professionals. The Support Vector Machine, Decision Tree and Random Forest techniques were compared. A predictive model created with the Random Forest technique demonstrated the best performance in the abandonment predictions, reaching 69.82% of accuracy. Thus, it is possible to use machine learning to predict abandonment trends of patients in Pulmonary Rehabilitation Programs.

Palavras-chave: Machine Learning, Predictive Analysis, Respiratory Diseases, Pulmonary Rehabilitation Programs