Exploratory segmentation of selected cardiovascular hospital services in Chile using DRG-based normalized indicators

To assess the use of DRG-based normalized indicators in supporting an exploratory analysis of inter-hospital variation in specific cardiovascular services within the Chilean public system from 2019 to 2023, employing routinely collected administrative data and confining interpretation to descriptive comparisons within the examined sample. The study was designed as descriptive and hypothesis-generating, not to validate hospital performance measures.

An analysis was conducted of 141,893 adult discharges from the departments of Cardiology, Cardiovascular Surgery, and Peripheral Vascular Surgery in 54 hospitals within the Chilean public system. Relative throughput, severity, and mortality were summarized into three DRG-based normalized indicators as a function of coded complexity. The hospital-level means were standardized to z-scores prior to principal component analysis. Sampling adequacy for PCA was assessed using the Kaiser-Meyer-Olkin statistic (KMO = 0.6251) and Bartlett's test of sphericity (p = 1.00 × 10^-10), which suggested performing exploratory analysis with caution. K-means clustering was then performed on the first two principal components with k = 4, n_init=50, and random_state=42 for reproducibility. Differences (specialty, sex, cluster) in the relative throughput proxy were assessed using the Kruskal-Wallis test and Dunn's post hoc tests.

The first two principal components accounted for 86.26% of the variability in the hospital-indicator matrix. K-means clustering resulted in four hospital clusters: cluster 0 (8 hospitals, 14.8%), cluster 1 (8 hospitals, 14.8%), cluster 2 (26 hospitals, 48.1%), and cluster 3 (12 hospitals, 22.2%). Distinct combinations of indicators emerged from the cluster means: cluster 1 had the highest ER (0.9633) and the lowest MR (0.0016), while cluster 3 had low ER (0.2465), high SR (2.3286), and the highest MR (0.0375). Kruskal-Wallis tests indicated significant differences in ER by specialty, sex, and cluster (all p < 0.001).

The analysis suggests that DRG-based normalized indicators can be used to explore the segmentation of hospitals providing selected cardiovascular services within the Chilean public system. Their outputs should be interpreted as descriptive and hypothesis-generating rather than as validated measures of hospital efficiency or stand-alone evidence for resource-allocation decisions.