ORCID Identifier(s)


Graduation Semester and Year




Document Type


Degree Name

Master of Science in Computer Science


Computer Science and Engineering

First Advisor

Manfred Huber


Patient safety is a key aspect for good consumer care. When an individual is hospitalized or receives medication the family wants the patient safety to be above all factors. For instance, a drug can do both either cure the disease or perhaps, give rise to an adverse event. A drug administered for an indicated condition has substantial power to reduce or cure a disease and further to prevent it from happening again in the future but at the risk of side effects. At present, there are several methods in patient safety and in particular in the area of signal detection and off-label drug usage identification that are incorporated in patient’s safety. Even though these methods have relatively high accuracy, they have to be executed manually by a health professional who has to perform a case review which consumes a significant amount of time. To address the above issue, this thesis explores a new method that can help identify whether the treatment has a lack of effect or not from medical text. This classification is based on the model’s prediction probability where a convolutional neural network algorithm is used as a systems classifier. The input to the classifier are raw text (case narratives), reported by patients, physicians, or any other reporter in the form of text or phone calls to local safety offices. Based on this, the classifier outputs a class label indicating predicted effectiveness. Currently, the model has three layers consisting of 1-convolution, 1-relu and 1-max pooling layers. The above-mentioned model is trained and tested on 4 different medications; The average size of data is approximately between 60-120 cases for each medicine gathered from electronic health records. This thesis is a proof-of-concept which demonstrates the automated version of an existing manual process which is carried out in many pharma companies for patient safety. Pharma companies have recently realized and begun to address the need for this transformation which will increase efficiency in patient’s safety reporting to the Center of Complaint Vigilance. Results: In this thesis, the CNN model was trained and tested on 320 and 160 clinical texts achieving an average accuracy (4 medications) of 87% and 85.76% on training and test data, respectively. Furthermore, precision of 90.3%, recall of 86.8% and F1-measure of 84.5% were achieved. In addition, this thesis also depicts a comparison between GPU (GTX-1080 hybrid) and CPU training time after running the model for 1000 epochs.


Classification, Convolutional neural networks, Random forest, Medical text, Case narratives


Computer Sciences | Physical Sciences and Mathematics


Degree granted by The University of Texas at Arlington