Application of Affinity Analysis Techniques on Diagnosis and Prescription Data

• Published in: Proceedings / IEEE International Symposium on Computer-Based Medical Systems pp. 403 - 408
• Main Authors: Sanida, Theodora, Varlamis, Iraklis
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2017
• Subjects: affinity analysis; association rules; cholesterol; Data mining; Diabetes; Diseases; dyslipidaemia; essential (primary) hypertension; Hypertension; Medical diagnostic imaging; Pharmaceuticals; pure hypercholesterolemia; sequential patterns
• ISSN: 2372-9198
• DOI: 10.1109/CBMS.2017.114

This study performs an Affinity Analysis on diagnosis and prescription data in order to discover cooccurrence relationships among diagnosis and pharmaceutical active ingredients prescribed to different patient groups. The analysis data collected during consecutive visits of 4,473 patients in a 3 years period, focused on patients suffering by hypertension and/or hypercholesterolemia and applied association rule and sequential rule mining techniques. The findings have been validated in the specific dataset using statistical analysis methods. Association rule mining shows an association between gastrooesophageal reflux and the medicines prescribed for hypertension and heart diseases, which agrees with findings in the related literature. Another interesting finding, not yet been reported in related studies is the association between heart diseases, gastroesophageal reflux and insulin-dependent diabetes mellitus for patients that have both hypertension and hypercholesterolemia. Apart from the medical findings, which must be subject of further research we propose a methodology for the analysis of data collected from a continuous screening process of a group of patients. With the use of data mining techniques we are able to extract and formulate the potential research questions, which are then validated using statistical methods and can also be validated in larger population studies.