Development of a model for control the flow of patients with cardiovascular diseases using data mining methods

Introduction. Currently, the development of Big Data technologies and methods of big data mining has opened up the possibility of investigating the timeliness, availability and effectiveness of therapy when processing all available information about the treatment practice. Personalized and preventive medicine methods based on remote monitoring of patients and intelligent analysis of similar treatment practices will lead to significant cost savings and improved quality of life. One of the most effective methods of studying patient data and their electronic medical records is machine learning methods. Aim. This study is aimed at building a model for managing the flow of patients with cardiovascular diseases based on the analysis of personalized patient data maps. Materials and methods. The forecast for treatment of patients with heart diseases was determined using the method of logistic regression, the algorithm for building ID3 decision trees, and the method of training the ensemble - random forests. As part of the experimental study, the effectiveness of the methods considered for forecasting was evaluated based on the analysis of the ROC curve and the AUC metric. Results. Experiments on an array of electronic personalized data about medical services in the territorial Fund of compulsory medical insurance (TFOMS) and the medical information and analytical center of Orenburg showed that for short-term forecasting for 1 month, the ID3 algorithm for constructing decision trees showed better results, and when the period under consideration was increased to 3 months, the method of logistic regression was more effective. Conclusion. The proposed approach to predicting patient requests allows us to improve the quality of management of the clinical and organizational health care system in the provision of medical care, as well as to plan the volume and number of individual medical services.