Development and internal validation of a model for predicting cefoperazone/sulbactam-associated coagulation disorders in Chinese inpatients

Abstract

Abstract Background and aim The use of cefoperazone/sulbactam (CPZ/SAM) could commonly cause vitamin K-dependent coagulation disorders and even hemorrhage sometimes. However, there is a lack of prediction tools estimating the risk for this. This study aimed at developing and internally validating a model for predicting CPZ/SAM-associated coagulation disorders in Chinese inpatients. Methods A case-control study was conducted in 11,092 adult inpatients admitted to a Chinese general hospital between 2020 and 2021 and treated with CPZ/SAM. Patients with CPZ/SAM-associated coagulation disorders were identified through the Adverse Drug Events Active Surveillance and Assessment System-II and subsequent manual evaluation. Controls were selected from eligible patients who didn’t develop coagulation disorders after CPZ/SAM therapy, with a 1:1 propensity score matching. The final predictors were obtained by univariable and multivariable logistic regression analyses. Internal validation and calibration for the model were performed using 1000 bootstrap resamplings. Results 258 patients were identified as CPZ/SAM-associated coagulation disorders in 2184 patients eligible for inclusions and exclusions and the incidence was 11.8%. A final population of 252 cases and 252 controls was included for model development and validation. Malnutrition (OR = 2.41 (1.56–3.77)), history of recent bleeding (OR = 1.95 (1.32–2.90)), treatment duration (OR = 1.10 (1.07–1.14)), combination with carbapenems (OR = 4.43 (1.85–11.88)), and serum creatinine (OR = 1.01 (1.00–1.01)) were identified as final predictors. The model showed good discrimination, calibration, and clinical practicality, with the validated area under the receiver operating characteristic curve being 0.723 (0.683–0.770). Conclusions The model with good performance quantifies the risk for CPZ/SAM-associated coagulation disorders, and may support individual assessment and interventions to mitigate the risk after external validation.