Identification of Referral Pattern in Indonesian Primary Healthcare Facilities Using Data Mining Techniques

The implementation of national health insurance has significantly increased Indonesian people's access to health services, particularly in rural and remote areas. However, the rise in BPJS Kesehatan participants has led to a budget deficit due to the increasing number of referrals from primary healthcare facilities. This research aims to identify referral patterns in primary healthcare facilities using data mining techniques, specifically Naive Bayes, Decision Tree, and Random Forest algorithms. Our findings highlight a concentration of referral cases in the Java-Bali region, necessitating targeted resource allocation. Addressing this could reduce unnecessary referrals and improve service efficiency. The prominence of non-communicable diseases among referrals underscores the need for preventive action at the primary care level. Model evaluations reveal that Naive Bayes offers fast computation, Decision Tree performs well on training data, and Random Forest balances accuracy with consistency. These insights can inform healthcare management and policy decisions, ultimately enhancing the effectiveness of healthcare delivery.