Establishing practical guidelines for microalbuminuria screening using J48 algorithm decision tree in data mining

Abstract

This study aimed to develop practical, cost-effective screening guidelines using Robert's Test and J48 decision tree algorithm to optimize microalbuminuria detection in health screening programs. Data were collected from random spot urine samples during daytime from 709 participants who received routine health checks at the Center of Medical Laboratory Service, Medical Technology Program, Nakhonratchasima College. The samples were screened with protein and microalbumin dipsticks, confirmed by Robert's test, and analyzed for the microalbumin-to-creatinine ratio (MA/CR). Kruskal-Wallis and Pearson correlation tests assessed statistical significance at p < 0.05 and 95% confidence intervals. Several significant correlations were found, including the relationship between increasing age and higher blood pressure, fasting plasma glucose, creatinine, and urea nitrogen levels. The reduced estimated glomerular filtration rate (eGFR) was significantly correlated with microalbuminuria (p < 0.01) among participants younger than 83 years. The J48 analysis showed that Robert's test and fasting plasma glucose (FPG) were effective for screening. Participants with a negative Robert's test and FPG < 96 mg/dL had 97.7% normal microalbumin levels. However, participants with trace protein and FPG > 84 mg/dL had a 35% chance of microalbuminuria. This algorithm can develop practical guidelines for detecting microalbuminuria, potentially reducing costs and laboratory burden and identifying high-risk groups in population health.