Clustering of patients based on their functional, clinical and anthropometric indicators for the construction of models for assessing bio-age

Cluster analysis has become a widely used tool for analyzing medical data to identify groups of patients. But despite the widespread use of cluster analysis, it is rare to find publications where the identification of groups of patients and the attributes by which the division into groups occurred are mathematically justified. To solve this problem, a method called clustering with a teacher can be applied, the essence of which is to apply multiclass classification methods using cluster labels as a target variable. In this paper, this method is employed to identify indicators by which groups of patients will be divided in the databases of the autonomous public health care institutions SOCP Hospital for War Veterans and Institute of Medical Cell Technologies for years 1995-2022 in volume 6440. The HDBscan method was used for clustering method, and the CatBoost method in the multiclass classification mode was used as a verification method for the obtained clusters of patients. As a result, 4 clusters were obtained divided by gender and the patient's condition. In order to identify statistical differences between the obtained clusters, an AB analysis of these clusters was carried out by means of the Kruskal-Walis criterion. The results of the AB analysis showed that the clusters have statistically significant differences in all functional parameters included in the analysis. Further, an AB analysis of the differences in the functional indicators of patients in outpatient and inpatient status for the female and male cluster was carried out. For the AB analysis, a permutation criterion and a bootstrap were used with the construction of confidence intervals of averages from samples generated in the bootstrap.

1. Krishtop V.V., Pakhrova O.A. Application of cluster and correlation analysis to assess hemorheological parameters in patients with essential arterial hypertension. Uspekhi sovremennogo estestvoznaniya = Advances in current natural sciences. 2014;9:11–16. (In Russ.).

2. Deckersbach T., Peters A.T., Sylvia L.G., Gold A.K., da Silva Magalhaes P.V., Henry D.B., Frank E., Otto M.W., Berk M., Dougherty D.D., Nierenberg A.A., Miklowitz D.J. A Cluster Analytic Approach to Identifying Predictors and Moderators of Psychosocial Treatment for Bipolar Depression: Results from STEP-BD. J Affect Disord. 2016;203:152–157. DOI: 10.1016/j.jad.2016.03.064.

3. O'Regan A., Hannigan A., Glynn L., Garcia Bengoechea E., Donnelly A., Hayes G., Murphy A.W., Clifford A.M., Gallagher S., Woods C.B. A cluster analysis of device-measured physical activity behaviours and the association with chronic conditions, multi-morbidity and healthcare utilisation in adults aged 45 years and older. Woods Preventive Medicine Reports.2021;24:101641–101651. DOI: 10.1016/j.pmedr.2021.101641.

4. Serpa Neto A., Bos L.D., Campos P.P.Z.A., Hemmes S.N.T., Bluth T., Calfee C.S., Ferner M., Güldner A., Hollmann M.W., India I., Kiss T., Laufenberg-Feldmann R., Sprung J., Sulemanji D., Unzueta C., Vidal Melo M.F., Weingarten T.N., Tuip-de Boer A.M., Pelosi P., Gama de Abreu M., Schultz M.J. Association between pre-operative biological phenotypes and postoperative pulmonary complications an unbiased cluster analysis. Eur J Anaesthesiol. 2018;35:702–709. DOI: 10.1097/EJA.0000000000000846.

5. Gagnon P., Casaburi R., Saey D., Porszasz J., Provencher S., Milot J., Bourbeau J., O'Donnell D.E., Maltais F. Cluster Analysis in Patients with GOLD 1 Chronic Obstructive Pulmonary Disease PLoS ONE. 2015;10(4):e0123626. DOI: 10.1371/journal.pone.0123626.

6. Sharma A., Zheng Y., Ezekowitz J.A., Westerhout C.M., Udell J.A., Goodman S.G., Armstrong P.W., Buse J.B., Green J.B., Josse R.G., Kaufman K.D., McGuire D.K., Ambrosio G., Chuang L.M., Lopes R.D., Peterson E.D., Holman R.R. Cluster Analysis of Cardiovascular Phenotypes in Patients With Type 2 Diabetes and Established Atherosclerotic Cardiovascular Disease: A Potential Approach to Precision Medicine. Diabetes Care. 2022;45:204–212. DOI: 10.2337/dc20-2806.

7. Demchenko M.V., Kashirina I.L., Firyulina M.A. Cluster analysis of patients’ states performed in order to develop treatment strategies for patients with atherosclerosis. Vestnik Voronezhskogo gosudarstvennogo universiteta. Serija: Sistemnyj analiz i informacionnye tehnologii = Proceedings of Voronezh State University. Series: Systems Analysis and Information Technologies. 2021;2:126–137. DOI: 10.17308/sait.2021.2/3509. (In Russ.).

8. Konno S., Taniguchi N., Makita H., Nakamaru Yu., Shimizu K., Shijubo N., Fuke S., Takeyabu K., Oguri M., Kimura H., Maeda Yu., Suzuki M., Nagai K., Yo. M. Ito, Wenzel S.E., Nishimuka M. Distinct phenotypes of Cigarette Smokers Identified by Cluster Analysis of Patients with Severe Asthma. AnnalsATS. 2015;12(12):1771–1780. DOI: 10.1513/AnnalsATS.201507-407OC.

9. O'Regan A., Hannigan A., Glynn L., Garcia Bengoechea E., Donnelly A., Hayes G., Murphy A.W., Clifford A.M., Gallagher S., Woods C.B. A cluster analysis of device-measured physical activity behaviours and the association with chronic conditions, multi-morbidity and healthcare utilisation in adults aged 45 years and older. Preventive Medicine Reports. 2021;24:101641. DOI: 10.1016/j.pmedr.2021.101641.

10. Al-Harbi S. H., Rayward-Smith V.J. Adapting k-means for supervised clustering. Journal of Applied Intelligence. 2006;24(3):219–226.

11. McInnes L., Healy J. Accelerated Hierarchical Density Based Clustering. IEEE International Conference on Data Mining Workshops (ICDMW), IEEE. 2017;1:33–42.

12. CatBoost. Available from: https://catboost.ai (accessed on 21.01.2023).

13. Meshhaninov V.N., Gavrilov I.V., Myakotny`kh V.S., Shherbakov D.L. Genderny`e demograficheskie i statisticheskie osobennosti stareniya cheloveka. Novy`e informaczionny`e tekhnologii v obrazovanii i nauke. 2022;2(6):65–72. (In Russ.).

14. Kobzar A.I. Applied mathematical statistics. M.: Fizmatlit; 2006. 466–468 p. (In Russ.).

15. Kolyadin V.L. Permutation criteria as a universal nonparametric approach to testing statistical hypotheses. Радиоэлектроника и информатика = «Radioelectronics & Informatics» Journal. 2002;3(20):7–14. (In Russ.).

16. Limanovskaya O.V. Gavrilov I.V. Meshchaninov V.N. Shcherbakov D.L. Kolos E.N. Modeling the biological age of the patients based on their functional indicators. Modeling, Optimization and Information Technology. 2021;9(2). Available from: https://moitvivt.ru/ru/journal/pdf?id=966 DOI: 10.26102/2310-6018/2021.33.2.028 (In Russ.).