Predicting Chronic Phase Progression in Chikungunya Patients Using Machine Learning Models

Context: This research is set within the domain of neglected tropical diseases, specifically focusing on Chikungunya, a mosquito-borne viral disease. The study is motivated by the prevalence of Chikungunya in Brazil and the challenges associated with its chronic symptoms. Problem: The primary issue addressed is the difficulty in predicting which patients with acute Chikungunya will progress to the chronic phase. This progression leads to prolonged joint pain and other severe symptoms, affecting quality of life. Solution: We evaluate machine learning models, trained to predict the likelihood of acute Chikungunya cases progressing to the chronic phase. Theory of IS: The Information Processing Theory serves as the theoretical foundation, explaining how data is transformed into actionable insights. In this context, machine learning models act as information processors, learning from a real dataset how to produce predictions that can aid clinical decisions. Method: The methodology involves a quantitative analysis of different machine learning models, using documented data from Chikungunya diagnosed patients. The goal is to assess the models’ performance in predicting disease progression. Summary of Results: Preliminary results show that the XGBoost model demonstrates 79.31% of recall and 74.19% of f1-score in distinguishing patients likely to develop chronic conditions. Contributions and impact on the field of IS: This study contributes to both information systems and healthcare by enhancing the process of decision-making for health professionals. The implications are significant for academic research and practical application (including SUS applicability), providing information that facilitate earlier and more targeted treatments, focusing on patient quality of life.