Geospatial patterns and a two-year comparison of outpatient hospital utilization in Indonesia: An ecological analysis of 38 provinces, 2021–2024

This ecological study examined provincial variation, geospatial distribution and ecological correlates of outpatient hospital utilization across 38 Indonesian provinces, together with a two-year comparison of utilization between 2023 and 2024. Aggregate secondary data were compiled from official provincial statistical and health datasets for 2021–2024. The cross-sectional analysis used 2024 data from all 38 provinces; the two-year comparison used the 34 provinces with comparable 2023 and 2024 public- and private-hospital outpatient values. The primary outcome was total outpatient hospital utilization, defined as the sum of the percentage of residents reporting outpatient care at a public hospital and the percentage reporting outpatient care at a private hospital in the preceding month. Analyses comprised descriptive statistics, utilization tertiles compared with Kruskal–Wallis tests and epsilon-squared effect sizes, a paired 2023–2024 comparison, bivariate and multivariable ecological regression with heteroskedasticity-consistent standard errors, Bayesian linear regression estimated by Markov chain Monte Carlo (MCMC), thematic maps, global and local Moran's I, Getis–Ord Gi*, a spatial-weights sensitivity analysis, a readiness–utilization visualization and an exploratory k-means typology with quantitative internal validation. In 2024, median total outpatient hospital utilization was 19.01% (interquartile range (IQR) 15.36–23.37; range 7.85–31.19). Among the 34 comparable provinces, the mean difference between 2024 and 2023 was +0.91 percentage points (95% confidence interval (CI) −0.22 to +2.04; paired t(33)=1.64, p=0.110; Cohen's dz=0.28) and the median difference was +1.25 percentage points (Wilcoxon p=0.026); utilization was higher in 24 provinces and lower in 10. Previous-year utilization was the strongest bivariate correlate of 2024 utilization (Pearson r=0.872, R²=0.761), followed by prior hospital facility preference (r=0.708, R²=0.501), non-contribution-assistance insurance coverage and medical workforce density, whereas total insurance coverage was not associated with utilization. After mutual adjustment, only prior hospital facility preference remained associated with utilization (b=1.28 percentage points per percentage point, 95% CI 0.70 to 1.86), and Bayesian estimation reproduced the same ordering. Total utilization showed no significant global spatial autocorrelation, whereas public-hospital share and poverty were strongly clustered; local statistics located a high–high poverty and public-dependence cluster in the Papua region and a low–low public-share cluster in Jawa. Provincial outpatient hospital use therefore appears to be linked less to insurance coverage totals than to historical care-seeking patterns, provider readiness, socioeconomic position and public-sector dependence. Because the design is ecological and only two comparable years were available, these associations describe provinces rather than individuals and are not evidence of causal effects.