Cost Efficiency Management of Early Heart Disease Detection in the Era of Digital Health Service Transformation, A Data Mining Classification Model Analysis Using Naive Bayes and Decision Tree C4.5

Abstract

Cardiovascular disease remains the leading cause of death worldwide and imposes the largest catastrophic financial burden on Indonesia's national health insurance program, yet conventional diagnostic pathways remain costly and time consuming. This study evaluates Naive Bayes and Decision Tree C4.5 for early heart disease detection and quantifies the cost and time efficiency achieved through a data mining based approach compared with conventional examination. Using 300 anonymized patient records from multiple types of Indonesian health facilities, thirteen clinical attributes were used to train and test both classifiers on an identical eighty to twenty stratified split. Naive Bayes achieved an accuracy of 63.33 percent and an F1 score of 62.07 percent, outperforming Decision Tree C4.5, which achieved an accuracy of 55.00 percent and an F1 score of 50.91 percent, although a McNemar test indicated that this difference was not statistically significant. In contrast, the data mining based pathway produced a mean cost efficiency of 84.82 percent per patient and reduced diagnostic turnaround from an average of 8.26 days to approximately 15 minutes, with efficiency remaining consistently high, between 83.28 and 85.86 percent, across all facility types examined. These findings suggest that the primary value of data mining based early detection lies not in the superiority of one classifier over another but in the substantial and broadly distributed reduction in cost and time it offers, supporting its relevance for Indonesia's digital health transformation, particularly at primary care facilities with limited resources.