OBJECTIVE To investigate the risk factors for prolonged somnolence of patients caused by diazepam and to establish and evaluate the risk prediction model for the safe use of diazepam.

METHODS The Prescription Automatic Screening System combined with the Hospital Information System were used to retrospectively collect the clinical data of patients who used diazepam injection followed by flumazenil injection in Liuzhou People’s Hospital from July 2020 to December 2021. Logistic stepwise regression analysis was performed on the included relevant variables to screen the independent influencing factors of diazepam causing somnolence in patients. Independent influencing factors were included to establish a nomogram prediction model. The area under the receiver operating characteristic curve and the calibration curve were used to evaluate the discrimination and calibration of the model, respectively. Decision curve analysis was employed to assess the clinical utility of the model, and the bootstrap method was utilized for internal validation of the model.

RESULTS A total of 244 patients were included, of which 118 patients had drowsy adverse reactions. Binary logistic regression analysis revealed that patients with albumin less than 30 g∙L⁻¹(OR=3.241, 95%CI 1.364−7.701), aspartate aminotransferase greater than 40 U∙L⁻¹(OR=2.589, 95%CI 1.062−6.307), and prothrombin time greater than 14 s(OR=2.180, 95%CI 1.138−4.175) were independent risk factors for diazepam-induced somnolence in patients, and a nomogram model was developed to predict the risk of diazepam adverse reactions. The AUC value of the model was 0.688, the specificity of the model was 0.651, the sensitivity of the model was 0.653, and the accuracy of Bootstrap method was 63.5%, the Kappa value was 0.26. The calibration curve showed good consistency between the prediction probability and the observation probability of nomogram. The decision curve analysis indicated that the nomogram could be applied clinically if the risk threshold was between 38% and 83%.

CONCLUSION This model has good fitting, discrimination, calibration and certain predictive ability to help medical professionals predict the risk of possible adverse reactions with diazepam.