MACHINE LEARNING ANALYTICS TO IMPROVE VERIFICATION CAPACITY IN SOCIAL HEALTH INSURANCE

• Main Author: Virlandra, Revien
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75985
• Institution: Institut Teknologi Bandung

In 2018, the Social Security Administrator (SSA) in Indonesia faced a significant mismatch problem between premium collection and healthcare service purchasing, resulting in a negative difference of IDR 12 trillion in the second half of the year. This issue caused financial unrest among healthcare service providers, leading to bill payment delays, and affecting health service delivery to Social Health Insurance (SHI) consumers in Indonesia. Although Social insurance actuaries expect premium mismatch, the potential for fraud in healthcare billing is a critical issue that needs immediate attention. Previous research indicates that fraudulent billing for health services is common across all healthcare financing schemes, including SHI. The SSA faces a significant challenge with limited resources to identify fraudulent health service bills and implement appropriate measures in its operations. In terms of addressing the limited resources challenge, this study uses Machine Learning to assist verifiers in identifying and predicting fraud in health service billing. The study uses the Nearest Neighbour and Random Forest algorithms and adjusted healthcare service billing data for 2020-2021, involving the professional judgment SSA verifiers. The findings of this study demonstrate that social health insurance administration institutions can use predictions based on the Nearest Neighbour and Random Forest algorithms to identify fraud. Testing the two algorithms in this study resulted in valid results reaching 86% (compared to the verifier assessment), which significantly contributes to identifying fraud in the social insurance industry.