Medication names that look alike and sound alike (LASA) are one of the common causes contributing to medication errors. This study is built upon the foundation of previous probing study of statistical classifiers for identifying medication incidents due to look-alike, sound-alike (LASA) mix-ups published in Health Informatics Journal in 2014. We retrieved two hundred and twenty-seven patient safety incident advisories from the Canadian Patient Safety Institute’s GPSA system with the indication of LASA case binary marker. The study aims to assess the performance of deep neural network (DNN) models in classifying LASA incident reports, and compared the model’s performance with that of other state of the art classifiers (including logistic regression, support vector machines and the decision-tree method). The base model setting is a 2 hidden layers DNN model with 200 neurons each via rectifier activation function. Ten folds cross-validation result shows that the mean accuracy is at 0.833 at the base DNN model setting. The average values obtained for under the curve (AUC) is reported as 0.921, indicating that the DNN model offered a superior classification approach. The study explores the applicability of using deep learning method for patient safety information retrieval from free-text medical incidence reports.