Uso de mineração de dados para descoberta de regras de associação em prontuários médicos

Abstract

This study aims to extract association rules in a medical records database through data mining techniques. Association rules can become useful information for health professionals, after analysis. For professionals, the information discovery in medical records can be complex and long without specialized technology’s aid on the process. For this purpose, the application of knowledge discovery in databases shows an viable option for acquiring useful knowledge and information. The database used in this study consists of medical records about 43.879 patients and 2.296.626 health care attendances, on the year of 2015, in Pato Branco city, Paraná. The PostgreSQL software was used for the purpose of data preparation, where you can alter the database, whereas WEKA was used for data mining application of the Apriori and Hotspot algorithms. Tests where the difference between the use of the confidence metric and the lift metric in the Apriori algorithm were made. The results showed association rules that wouldn’t be seen by considering the confidence metric Apriori application only. Using the results, attributes were then chosen for the Hotspot algorithm application, were, first the difference between disease groups from the male and female genders. Profiles for the age gaps that appeared on the database were made and the application found disease groups that wouldn’t appear on the Apriori application, such as age gaps from 18 years and above. The Hotspot application was then used for profiling three disease groups which appeared between the age gaps. The groups were related to back pain, ear and mastoid diseases and hypertension. From the association rules obtained, the relation between age gaps and disease groups was made clearer, with the rules even finding some different age gaps for diseases. The application also related the diseases with other symptoms, narrowing down the profiling for the diseases. The analysis of an health professional would be necessary for the conclusion of the study results, where the rules could then become useful knowledge.