Abstract: OBJECTIVE To observe the etiological characteristics of multidrug-resistant organisms (MDROs) infections in the patients with severe pneumonia, construct and validate the risk prediction model for MDROs infections so as to provide bases for early identification of high-risk patients and implementation of precise prevention and control. METHODS A total of 452 patients with severe pneumonia who were treated in critical care medicine department of Longhua Hospital, Shanghai University of Traditional Chinese Medicine from Jan. 2024 to Jul. 2025 were retrospectively enrolled in the study and were divided into the training group with 315 cases and the validation group with 137 cases. The distribution of pathogens isolated from respiratory tract specimens of the patients in the training group was observed. The predictive factors were screened out through univariate analysis, LASSO regression and multivariate logistic regression, the nomogram model was established. The internal validation was carried out by Bootstrap, the model was validated by temporal validation set. The discrimination, calibration and clinical utility of the model were assessed by the areas under the curves (AUCs), calibration curves and decision curve analysis (DCA). RESULTS Totally 173 strains of pathogens were isolated from the training group, of which 61.85% were gram-negative bacteria, and 38.15% were gram-positive bacteria. The multivariate analysis showed that there were 4 predictive factors( all P<0.05) as follows: history of cerebrovascular diseases(OR=1.764), use of no less than 3 types of antibiotics(OR=3.071), no less than 3 days of invasive mechanical ventilation(OR=2.772), gastric tube indwelling duration no less than 7 days(OR=2.436). The AUC of the model was 0.773 in the training model, 0.790 in the validation group. The calibration curve demonstrated good agreement between the predicted and observed risks, and DCA indicated that the model provided clinical net benefit across a wide range of threshold probabilities. CONCLUSIONS The gram-negative bacteria are dominant among the pathogens isolated from the severe pneumonia patients with MDROs infections. The prediction model exhibits satisfactory discrimination, calibration, and clinical applicability, and it can be served as a visual tool for early identification and targeted intervention to the MDROs infection in the severe pneumonia patients.