Analysis of Drug Product Dispensing Error Characteristics, Construction and Evaluation of a Risk Warning Model in Outpatient Pharmacy: A Retrospective Study in China.

Background/Objectives: Dispensing errors have the potential to cause significant and preventable patient harm, including adverse drug events, hospitalization, or even death. This study aims to analyze the characteristics of drug product dispensing errors in the outpatient pharmacy, identify risk factors, and develop a risk warning model for error prediction. Methods: A retrospective study analyzed 930 prescriptions with product dispensing errors and 1860 control prescriptions without errors in an outpatient pharmacy of a tertiary hospital from April 2021 to March 2023. Univariate and multivariable logistic regression were used to identify risk factors. A risk warning model with a cutoff value was constructed and its reliability evaluated using Receiver Operating Characteristics(ROC) curve analysis. The cutoff value was then used to assess the model's test effectiveness with validation dataset. Results: Logistic regression analysis identified six independent risk factors for product dispensing errors in outpatient pharmacies: work experience, professional title, education level, similar drug names, similar drug appearances, and multiple specifications. A risk warning model (p=ex/(1+ex), x=3.721-2.133×X1-0.424×X2-0.382×X3+0.736×X4+0.890×X5+0.701×X6) was established. ROC curve analysis showed an AUC of 0.921 (95% CI: 0.908, 0.933), cutoff value of 0.508, sensitivity of 86.0%, specificity of 91.7%, and Youden index of 0.777 for the training dataset. For the validation dataset, results revealed an AUC of 0.928 (95% CI: 0.901, 0.956), sensitivity of 85.90%, specificity of 83.10%, and Youden index of 0.69. Conclusion: The risk warning model demonstrated high accuracy in predicting product dispensing errors in outpatient pharmacies. Validated externally, it provides a practical reference for preventing such errors.