Operational efficiency of clinical departments under DRG-based payment: a super-SBM model with undesirable outputs and quantile regression analysis

China is now transitioning to Diagnosis‑Related Groups (DRG)‑based prospective payment systems. Under this transition, hospitals urgently need tools that are scientifically grounded. These tools are needed for baseline efficiency diagnosis. In this study, we develop and pilot‑test a department‑level analytical framework. We do this within a single tertiary hospital in Foshan. The framework was tested using data from 2024. We show that it is feasible to combine super‑efficiency SBM modeling with quantile regression. This combination can assess operational efficiency. It can also identify heterogeneous correlates of inefficiency. In addition, it can generate actionable hypotheses for tiered resource optimization. We used a super‑efficiency Slack‑Based Measure (SBM) model. This model incorporates undesirable outputs. These undesirable outputs are average length of stay and the 30‑day readmission rate. The model includes three inputs. They are the number of doctors, the number of nurses, and the number of beds. It also includes three desirable outputs. These are bed turnover frequency, total DRG weight, and the Case Mix Index (CMI). There are two undesirable outputs as well. Quantile regression was employed as the second-stage analytical approach. This method allows us to examine whether the correlates of efficiency differ across the performance distribution. We also used OLS regression and BCC‑based robustness checks. These checks provide benchmark comparisons. They also give indicative support for ranking stability. The mean overall technical efficiency (OTE) was 0.73. The mean pure technical efficiency (PTE) was 0.83. The mean scale efficiency (SE) was 0.91. These values indicate that there is substantial room for improvement. The highest‑performing departments were Pediatric Surgery (PTE = 2.56), Neonatology (PTE = 1.75), and Cardiothoracic Surgery II (PTE = 1.58). The quantile regression results showed that the associations varied by performance level. The composite size index had a significant negative association with efficiency. This negative association was most pronounced in low‑performing departments ( τ = 0.25: β = −0.122). Average length of stay showed a consistently negative association. This negative association grew stronger as we moved across quantiles. The coefficients went from − 0.351 to − 0.452. CMI showed a positive association. This positive association also grew stronger across quantiles. The coefficients went from 0.336 to 0.705. However, this positive association was only significant among high‑performing departments ( τ = 0.75: β = 0.705, p = 0.029). The 30‑day readmission rate did not show any independent association with efficiency. The OLS regression results showed that none of the individual input variables—doctors, nurses, or beds—had a statistically significant independent effect on efficiency. This study shows that a closed‑loop diagnostic framework is feasible. This framework integrates efficiency measurement, slack analysis, and quantile regression. It is designed for department‑level assessment during the DRG implementation period. Our findings suggest that indiscriminate resource expansion is not uniformly associated with higher efficiency. It may even make diseconomies of scale worse. Instead, we propose a stratified strategy. For low performers, downsizing may be appropriate. For medium performers, process streamlining may help. For high performers, concentrating on complex cases may be beneficial. However, these hypotheses need validation across multiple hospitals and over longer time periods. We also acknowledge that the single‑hospital, cross‑sectional design limits our ability to draw broad policy conclusions. Nevertheless, our integrated methodology offers a transferable analytical template. Other tertiary hospitals can replicate and refine this template. They can use it as they navigate their own DRG‑based payment transitions.