ABSTRACT
Unplanned events such as epidemic outbreaks, natural disasters, or major scandals are usually accompanied by supply chain disruption and highly volatile demand. Besides, authors have recently outlined the need for new applications of artificial intelligence to provide decision support in times of crisis. In particular, natural language processing allows for deriving an understanding from unstructured data in human languages, such as online news content, which can provide valuable information during disruptive events. This article contributes to this research strand as it aims to leverage textual data from news through sentiment analysis and predict demand volatility of pharmaceutical products in times of crisis. As a result, (1) a deep-learning-based sentiment analysis model was developed to extract and structure information from medicines-related news;(2) a framework allowing for combining extracted information from unstructured data with structured data of medicines demand was defined;and (3) an approach combining efficient artificial intelligence techniques with existing forecasting models was proposed to enhance demand forecasting in times of disruption. Additionally, the framework was applied to two examples of disruptive events in France: a pharmaceutical scandal and the COVID-19 pandemic. Findings outlined that using sentiment analysis allowed for enhancing demand forecasting accuracy.
ABSTRACT
OBJECTIVE: To define the factors associated with overprescription of antibiotics by general practitioners (GPs) for patients diagnosed with COVID-19 during the first wave of the pandemic. METHODS: Anonymised electronic prescribing records of 1370 GPs were analysed. Diagnosis and prescriptions were retrieved. The initiation rate by GP for 2020 was compared with 2017-2019. Prescribing habits of GPs who initiated antibiotics for > 10% of COVID-19 patients were compared with those who did not. Regional differences in prescribing habits of GPs who had consulted at least one COVID-19 patient were also analysed. RESULTS: For the March-April 2020 period, GPs who initiated antibiotics for > 10% of COVID-19 patients had more consultations than those who did not. They also more frequently prescribed antibiotics for non-COVID-19 patients consulting with rhinitis and broad-spectrum antibiotics for treating cystitis. Finally, GPs in the Île-de-France region saw more COVID-19 patients and more frequently initiated antibiotics. General practitioners in southern France had a higher but non-significant ratio of azithromycin initiation rate over total antibiotic initiation rate. CONCLUSION: This study identified a subset of GPs with overprescribing profiles for COVID-19 and other viral infections; they also tended to prescribe broad-spectrum antibiotics for a long duration. There were also regional differences concerning antibiotic initiation rates and the ratio of azithromycin prescribed. It will be necessary to evaluate the evolution of prescribing practices during subsequent waves.