Regresar

Cluster analysis to identify groups of diabetic patients based on symptoms

Abstract:

Currently, health professionals use computer systems to save patient information. This information can have two formats: structured, for example, medical codes to identify diagnoses; unstructured, for example, a text field to write more detail of the patients’ health. On the other hand, machine learning techniques, such as cluster analysis, identifying groups of objects with similar characteristics. The present work, has taken information from diabetic patients to perform a cluster analysis. Symptoms were extracted from text boxs of a medical computer system of a health institution in Ecuador. DBSCAN algorithm was applied at two levels to find specific groups of diabetic patients based on these symptoms. With the help of two physicians, the patient groups were classified as belonging to three phases: in treatment, control and diagnosis.