Deep learning application to automatic classification of recommendations made by hospital pharmacists during drug prescription review

Abstract

Abstract Purpose: Recommendations to improve therapeutics (Recos) are proposals made by pharmacists during the prescription review process to address sub-optimal use of medicines. In hospitals, Recos are generated daily as text documents that are sent to prescribers. If collected Recos data were easier and less time-consuming to summarize, they could be used retrospectively to improve safeguards for better prescribing. The objective of this work was to train a deep learning algorithm for automatic Recos classification in order to value the large amount of Recos data. Methods: The study was conducted at the University Hospital of Strasbourg. Recos data were collected throughout 2017. Data from the first six months of 2017 were labeled by two pharmacists who assigned to each of the Recos one of the 29 possible classes of the French Society of Clinical Pharmacy classification. A deep neural network classifier was trained to predict the class of Recos from the raw text data. Results: 27,699 labeled Recos from the first half of 2017 were used to train and evaluate a classifier. The prediction accuracy calculated on a validation data set was 78.0%. We predicted classes for the unlabeled Recos collected during the second half of 2017. Of the 4,460 predictions reviewed, 67 required corrections. After these additional labeled data were concatenated with the original data set and the neural network re-trained, accuracy reached 81.0 %. Conclusions: We report an efficient automatic classification of Recos. Making retrospective prescription review data easier to understand should enable better anticipation of prescription-related problems in future prescriptions, thereby improving patient safety.