Use of a deep learning application to classify recommendations made by hospital pharmacists during medication prescription review

Abstract

Abstract Background Recommendations are proposals made by hospital pharmacists to address the sub-optimal use of medications during prescription review. Objective To perform a large-scale descriptive analysis of recommendations formulated during prescription review using a deep neural network classifier in a hospital Setting This retrospective study was conducted at the University Hospital of Strasbourg. Main outcome measures Recommendations were automatically classified according to the coding of the French Society of Clinical Pharmacy. Method Data from 2018 to 2020 were collected from prescription support software. Results 2,930,656 prescription lines were analysed for a total of 119,689 patients. Among these prescription lines, 153,335 resulted in recommendations (n = 48,202 patients). Recommendations were predominantly observed in patients aged 65 years or older (n=26,141 patients) and in patients taking 5 or more medications (44,702). The most frequently identified types of Drug-related problems associated with recommendations were “Non conformity to guidelines or contra-indication” (n =88,523; 57.7%), “Overdosage” (16,975; 11.1%) and “Improper administration” (13, 898; 9.1%). The most frequently encountered drugs were: Paracetamol (n= 10,585; 6.9%), Esomeprazole (6,031; 3.9%), Hydrochlorothiazide (2,951; 1.9%), Enoxaparin (2,191; 1.4%), Tramadol (1,879; 1.2%), Calcium (2, 073; 1.3%), Perindopril (1,950; 1.2%), Amlodipine (1,716; 1.1%), Simvastatin (1,560; 1.0%) and Insulin (1,019; 0.7%). Conclusion The deep neural network classifier used met the challenge of automatically classifying recommendations from a large database without mobilizing significant human resources. The use of such a classifier can lead to alerting caregivers about certain risky attitudes in prescription and administration, and triggering actions to improve practices.