Laboratory data clustering in defining population cohorts: Case study on metabolic indicators Scientific paper

Abstract

The knowledge on the general population health is important for creating public policies and organization of medical services. However, per­sonal data are often limited, and mathematical models are employed to achieve a general overview. Cluster analysis was used in this study to assess general trends in po­pulation health based on laboratory data. Metabolic indicators were chosen to test the model and define population cohorts. Data on blood analysis of 33,049 persons, namely the concentrations of glucose, total cholesterol and triglycerides, were col­lected in a public health laboratory and used to define metabolic cohorts employing computational data clustering (CLARA method). The population was shown to be distributed in 3 clusters: persons with hyper­cholesterolemia with or without changes in the concentration of triglycerides or glucose, persons with reference or close to reference concentrations of all three analytes and persons with predo­mi­nantly elevated all three parameters. Clus­tering of biochemical data, thus, is a use­ful statistical tool in defining popul­ation groups in respect to certain health aspect.