Objective To investigate the dynamic impact of diagnosis-related group (DRG) payments reform and multidisciplinary team (MDT) management on the antimicrobial usage density (AUD) in intensive care unit (ICU), and construct an interrupted time series prediction model calibrated by the case-mix index (CMI), so as to break through the static limitations of traditional cross-sectional studies. Methods Data from ICU of a tertiary hospital from January 2021 to December 2024 were analyzed by the double interrupted time series (DITS) approach combined with an autoregressive integrated moving average (ARIMA) model. The implementation of DRG in October 2022 and the implementation of MDT management in August 2023 were identified as the key intervention points. Residual-calibrated sequences were constructed via CMI linear regression to control case complexity confounding, and model performance and predictive capability were assessed. Results The AUD exhibited a downward trend (β₁=-1.70) after the implementation of DRG, while the trend reversed to an upward direction (γ₁= 3.38) after the implementation of MDT management, though with no statistical significance. After adjusting case complexity confounders via the CMI linear regression residual method, MDT management demonstrated a significant positive impact on the trend in antimicrobial usage. The ARIMA constructed based on the calibrated sequence demonstrated robust predictive performance. Conclusion The CMI-calibrated time-series model can effectively control confounding and analyze the dynamic heterogeneity of policy interventions. The "confounding control-dynamic prediction" integrated framework constructed in this study provides a data-driven decision support tool for the refined management of antimicrobial agents.