BAYESIAN SPATIO-TEMPORAL MODELING FOR LONG-TERM HEALTHCARE COST PROJECTIONS (CASE STUDY: BPJS KESEHATAN IN MAKASSAR CITY)

• Main Author: Nisa SH, Khaerun
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81649
• Institution: Institut Teknologi Bandung

This study employs Bayesian Spatio-Temporal (BST) modeling to project long-term healthcare costs based on BPJS Kesehatan claim data in Makassar City for the year 2022. BST modeling is a statistical approach that integrates spatial and temporal aspects into data modeling. The objective of this study is to plan policies and allocate healthcare costs efficiently and effectively to ensure the sustainability of the BPJS Kesehatan insurance system. The results of this study include the distribution characteristics of RJTL claim costs following a Gamma-Exponential distribution and RITL claim costs following an Exponential distribution, as well as the identification of significant factors affecting claim costs such as the number of cases and the number of rooms. The linear model with spatial error (M2) proved optimal in modeling both types of claim costs with the lowest values for evaluation criteria such as DIC, WAIC, and CRPS. The visualization results of hospital clusters show that the majority (25-30 hospitals for RJTL and 18-24 hospitals for RITL) have low claim costs (Cluster 1), followed by Cluster 2 comprising 7-12 hospitals for RJTL and 12-17 hospitals for RITL with moderate claim costs, while Cluster 3 with high claim costs includes only 1-3 hospitals each month. Additionally, Cluster 0 includes 1-4 hospitals that do not submit claim costs. Long-term projections using the Markov chain model indicate the dominance of hospitals in the low-cost cluster for both RJTL and RITL. For RJTL, the steady-state probabilities are 1.1% of hospitals with no healthcare costs, 69.2% with low costs, 25.8% with moderate costs, and 3.8% with high costs. For RITL, the steady-state probabilities are 5.8% of hospitals with no healthcare costs, 53.9% with low costs, 34.8% with moderate costs, and 5.5% with high costs. By utilizing the information obtained from BST modeling, BPJS Kesehatan can more effectively manage healthcare costs, improve healthcare service efficiency, and ensure the sustainability of the health insurance program.