Abstract: OBJECTIVE To explore the related factors for intracranial infections in intraventricular chemotherapy patients after they were treated with Ommaya reservoir (OmR) and construct the risk prediction model. METHODS The meningeal metastases patients who were treated with OmR in Affiliated Brain Hospital of Nanjing Medical University from Jan. 2023 to May 2025 were recruited as the research subjects. Multivariate logistic regression analysis was performed to screen out the related factors for the intracranial infections after the use of OmR, and the prediction model was established. The internal validation was carried out by Bootstrap method, and the model's discrimination was assessed by means of the area under the receiver operating characteristic (ROC) curve; the calibration was evaluated by Hosmer-Lemeshow test and calibration curves. The clinical net profit was assessed through decision curve analysis (DCA). RESULTS The 300 patients involved in the study accumulatively received intraventricular chemotherapy with OmR for 1627 times, and the incidence of intracranial infections was 14.29%. The multivariate analysis showed that the number of times of OmR punctures (OR = 1.102, 95%CI: 0.965 to 1.259), cerebrospinal fluid (CSF) leakage at the OmR puncture site (OR = 14.185, 95%CI: 5.495 to 36.617), local skin infection around the OmR (OR = 12.202, 95%CI: 4.481 to 33.229), concurrent cranial radiotherapy (OR = 4.739, 95%CI: 2.033 to 11.045), and CSF glucose level (OR = 0.651, 95%CI: 0.436 to 0.972) were the predictive factors for the OmR-associated intracranial infections in the intraventricular chemotherapy patients(P＜0.05). The area under the ROC curve of the model was 0.879(95%CI: 0.816 to 0.942), and the area under the curve was 0.877(95%CI:0.806 to 0.933) after the internal validation with Bootstrap, indicating that the model had excellent discrimination ability. The calibration curves showed that the predicated probability was highly consistent with the actual risk, and Hosmer-Lemeshow test indicated that the model had high goodness of fit(χ²=4.098, P=0.848). DCA suggested that the model could bring net profit to the clinical decision when the threshold probability ranged between 6% and 89%. CONCLUSIONS The established risk prediction model for OmR-associated intracranial infections demonstrates high discrimination and calibration, and it may provide references for clinical identification of high-risk patients.