An intelligent system for prognosis of noncommunicable diseases’ risk factors

•Real time model for finding the impact on risk factors due to actions taken by users.•ChronicPrediction uses data generated by patients for improve its predictions models.•We have built a Bayesian Network for predicting coronary artery disease.•Training data was statistically validated showing similarity with real conditions.•Four scenarios were designed for functionally testing the prototype of our model. Noncommunicable diseases are the main reason to the rise of diseases incidence in the developed world. The management and prevention of these diseases can be done by controlling the behavioral and biological risk factors which are related to them. ChronicPrediction is an intelligent system for noncommunicable diseases care which determines in real time the impact on risk factors due to actions taken by users. Based on impact information, the system presents on users’ smartphones strategic messages to help in their treatment. ChronicPrediction applies Bayesian Networks (BNs) which use risk factors for mapping the causes of noncommunicable diseases worsening. The support to multiple chronic diseases and the integrated use of multiple BNs based on risk factors are the main contributions of this work and differentiate the proposed system from related work. We have built a functional prototype that allowed us to conduct two experiments. The first one successfully tested the main functionalities provided by ChronicPrediction to support BNs based on risk factors and the sending of messages to users’ smartphones. The evaluation involved the building of a BN for predicting coronary artery disease made with real world data obtained in a prospective cohort study. The study involved 302 patients from a hospital localized in southern Brazil. The second experiment assessed the ChronicPrediction support to multiple BNs at same time. The test involved the previous BN and another from a thirty part research work to map risk factors of diabetes. The results were encouraging and show potential for implementing ChronicPrediction in real-life situations.