1. A retrospective analysis was performed to determine the prevalence of multidrug- resistant organisms infection in ICU from October 2017 to October 2019.
2. Non-MDRO patients were selected by random sampling in a ratio of 1:1 to the final MDRO group during the same period , and select the risk factors of infection with multi-drug resistant bacteria by comparing the two groups.
3. Randomly select 30% of the sample size as the validation set, and the remaining 70% for the training set to establish a model. Using multi-factor Logistic regression, decision tree classification, artificial neural network, support vector machine, Bayesian network Method to establish risk assessment system for multidrug-resistant organisms infection respectively.Using validation set data to calculate the area under the ROC curve (AUC) and sensitivity, specificity of models and comparing the prediction accuracy of several models. Finally, choose a more suitable risk assessment system for multidrug-resistant organisms infection.
4. Predict the patient's infection risk level according to the best risk assessment system and develop a low-to-high intervention plan.

Inclusion Criteria：

• Duration of ICU admission ≥ 48h
• no less than 18 years old

Exclusion Criteria:

• Patients with MDRO infection at the time of admission
• Lack of case data
• MDRO colonized patients

①Length of stay in ICU is more than 48h; ② no less than 18 years old.