Objective To analyze the antimicrobial resistance characteristics of intensive care unit (ICU)-acquired multidrugresistant Klebsiella pneumoniae (MDR-KP) infection and construct a risk prediction model. Methods A retrospective analysis was conducted on the clinical data of 146 patients admitted to the ICU of the 990th Hospital of the Joint Support Force of the People’s Liberation Army from January 2024 to December 2025. All patients had Klebsiella pneumoniae isolated from sterile fluid specimens more than 48 hours after ICU admission. The patients were divided into the MDR-KP group (n＝66) and the non-MDR-KP group (n＝80) based on the drug sensitivity results. The baseline characteristics, drug resistance profile features and clinical outcomes between the two groups were compared. Through multivariate Logistic regression analysis, independent risk factors were screened out. A nomogram prediction model was constructed and internal validation was conducted using the Bootstrap resampling method (1,000 times). Results Univariate analysis showed that the acute physiology and chronic health evaluation Ⅱ(APACHE II) score, the proportion of patients with a history of hospitalization within 3 months prior to ICU admission, the proportion of carbapenem exposure history, the proportion of mechanical ventilation≥7 days, the proportion of continuous renal replacement therapy (CRRT), and the proportion of serum albumin (ALB)＜28 g/L in the MDR-KP group were higher than those in the non-MDR-KP group (all P＜0.05). The resistance rates of cephalosporin, ceftazidime, imipenem, meropenem, etc. in the MDR-KP group were higher than those in the non-MDR-KP group, and they remained highly sensitive to tigecycline. Multivariate analysis showed that carbapenem exposure history (OR＝4.91), CCRT (OR＝ 3.48), mechanical ventilation≥7 days (OR＝2.30), serum ALB＜28 g/L (OR＝2.07), and a history of hospitalization within 3 months prior to ICU admission (OR＝1.98) were independent risk factors for ICU-acquired MDR-KP infection (all P＜0.05). The proportion of combined use of≥3 antibiotics in the MDR-KP group and the length of ICU stay were significantly higher than those in the non-MDR-KP group (all P＜0.001). The constructed nomogram model was internally validated using the Bootstrap resampling method, and the corrected area under the curve (AUC) was 0.86 (95%CI: 0.81～0.92). The calibration curve showed that the predicted probability was highly consistent with the actual observed probability (Hosmer-Lemeshow test P＝0.412). Conclusion Carbapenem exposure history, CRRT, mechanical ventilation≥7 days, serum ALB＜28 g/L and a history of hospitalization within 3 months prior to ICU admission are independent risk factors for ICU-acquired MDR-KP infection. The nomogram prediction model constructed accordingly can identify highrisk patients at an early stage prior to the return of antimicrobial susceptibility testing results, supporting precise risk alerting and targeted intervention.