Formulation of risk prediction model for cardiovascular diseases using conventional logistic regression analysis

Abstract

Cardiovascular diseases (CVDs) are the leading cause of death, with 31% of global mortality. In this paper, we focus on the development of predictive modelling of the CVDs using Logistic Regression Analysis (LRA). The model uses self-reported information of individuals on selected variables which also called as non-laboratory-based features. We use the binary logistic regression analysis due to the binary nature of the outcome feature which is the cardiovascular diseases (CVDs) status. The methodology to develop the risk prediction models are discussed. Then, the risk prediction model of the CVDs using LRA is developed. The performance of the LRA model was evaluated through 10-fold cross-validation and the performance were presented in term of accuracy, sensitivity, specificity, kappa statistic, root mean square error (RMSE) and area under the curve (AUC). Forward stepwise LRA model highlighted the individual feature importance in the LRA model. The role of dietary habit is different from the existing risk prediction models in terms of direction or statistical significance. Mainly dietary features which significantly vary in different regions. Therefore, this finding supported the assumption that each country or region should has their own risk prediction models. It concludes that every country must have their local risk prediction models, but it is recommended to follow standard methodology and risk assessment framework which been established.