Emprego de técnicas de data mining na identificação de padrões relacionados às doenças oculares em pacientes pediátricos atendidos em um Hospital Universitário de uma cidade do sudoeste do Brasil

Abstract

Objective: To employ data mining techniques to treat data collected from patient records, identifying the most relevant attributes and the most appropriate data mining tools for analyzing health data. Methods: Cross-sectional study was conducted with secondary data from pediatric patients seen in a University Hospital's ophthalmology sector located in southeastern Brazil from January 2018 to December 2019. The IBM SPSS Statistics v.25 was used to characterize the sample regarding demographic characteristics and ophthalmological consultation of patients. The tool R version 4.0.0 was used to create a classification model by the Naves Bayes algorithm, whose function was to evaluate the variables to predict each patient's diagnosis. Results: The sample consisted of 196 eyes. The mean age of the patients was 10.04 years. 53% were male, and 66.7% were brown. The main complaints that led patients to seek eye care were: Crooked eye (26.9%) and Low Visual Acuity (25.3%). The classification model created obtained a diagnosis accuracy rate of 73%. Conclusions: The importance of computerization of the hospital system and the training of health professionals in data science were identified.