Using of machine learning methods in prescribing hypertension therapy

Despite the emergence of new modern medicines, mortality rates from essential hypertension remain high. This problem is since a combination of several groups of medicines is required to effectively treat this disease. The aim of this study is to develop models for the automated selection of medicines for the treatment of hypertension based on the individual characteristics of the patient, as well as to assess the effectiveness of the prescribed treatment based on the available clinical indicators of patients and the proposed combination of drugs. The original dataset contains depersonalized information on 262 patients of the cardiological hospital for 66 clinical parameters. Six groups of drugs were considered: BAB, I-ACE\ARA, CCB of the nifedipine group, CCC of the verapamil group, diuretics, centrally acting medicines. Machine learning techniques have been used to identify determinants that contribute to the success of drug treatment for hypertension in each sample of patients. During the study, to achieve this goal, several machine learning models were built to solve classification and regression problems. The highest accuracy was shown by the gradient boosting models XGBOOST for the classification problem and CATBOOST for the regression problem. Based on the results of the study, it can be concluded which clinical indicators are most significant for effective treatment with each of the medicines under consideration.

1. N. Ikeda. Control of hypertension with medication: a comparative analysis of national surveys in 20 countries. Bulletin of the World Health Organization. 2014;92(1):10-19.

2. S. Kjeldsen Updated national and international hypertension guidelines: a review of current recommendations. Drugs. 2014;6:2033-2051.

3. FeatureSelector: feature selection for machine learning in Python. Available at: https://proglib.io/p/feature-selector/ (accessed 10.11.2020).

4. Will Koehrsen. A Feature Selection Tool for Machine Learning in Python : Towards Data Science. Available at: https://towardsdatascience.com/a-feature-selection-tool-formachine-learning-in-python-b64dd23710f0/ (accessed 01.11.2020).

5. Leo Breiman. Random Forests. Machine Learning. 2001:5-32.

6. The Ultimate Guide to Random Forest Regression. Available at: https://www.keboola.com/blog/random-forest-regression/ (accessed 10.11.2020).

7. Shitikov V.K., Mastitsky S.E. Classification, regression, Data Mining algorithms using R. Available at: https://github.com/ranalytics/data-mining/ (accessed 09.09.2020)

8. Firyulina, M.A., Kashirina, I.L. Classification of cardiac arrhythmia using machine learning techniques. Journal of Physics: Conference Series, 2020;1479(1),012086. DOI: 10.1088/1742-6596/1479/1/012086

9. Kashirina I. L., Firyulina M. A., Gafanovich E. Y. Analysis of the significance of predictors of survival after myocardial infarction using the Kaplan-Meyer method. Modeling, optimization and information technology. 2019;1(24):7-20. DOI: 10.26102/2310- 6018/2019.24.1.007

10. Gafanovich E.YA., Kashirina I.L. Model-oriented approach to the choice of antihypertensive therapy. Vrach-aspirant. 2015;(3.1):183-191.