Temporal Patterns in Hospital Performance Indicators in the Chilean Public Hospital System: A Longitudinal Analysis...

Objectives: To examine temporal patterns in hospital performance indicators within the Chilean public system between 2019 and 2023 using dimensionless coefficients, which allow comparisons between hospitals of different sizes and case-mix. Methods: Data from 3,803,275 discharges across 68 Chilean public hospitals were evaluated in this study. Three dimensionless indicators (relative efficiency, severity, and mortality) were constructed using DRG FONASA fields. The analysis approach is based on descriptive time trends, within-hospital coefficient of variation (CV), IQR-based outlier identification, and RF predictive modeling with time-series cross-validation. Results: Most hospitals (95.6%) showed low within-hospital variability (CV < 0.5). The mean efficiency proxy increased from 0.3051 (95% CI: 0.3044-0.3057) in 2019 to 0.3134 (95% CI: 0.3127-0.3142) in 2023. Lagged REP was the strongest predictor in the model (85.1% variable importance). The Spearman correlation between the proposed REP and a VRS input-oriented DEA model (using total length of stay as input and total DRG weight as output) was 0.7857 (p < 0.001), indicating a strong convergent association. However, the DEA model employed is limited to a single input and a single output, as data constraints precluded more complex specifications. Conclusions: The proposed dimensionless indicators showed relatively stable temporal patterns across most hospitals. The framework may support descriptive longitudinal comparisons across heterogeneous institutions, though external validation is required.