Artificial Intelligence and Clinical Pharmacy

Artificial intelligence can help clinical pharmacy services better manage their prescription analysis, therapeutic patient education and drug dispensing activities. In order to assess the place and role of these tools in this field, researchers from the Translational Research and Innovation in Medicine and Complexity laboratory (TIMC - CNRS/Université Grenoble Alpes) carried out a literature review. This work is published in the International Journal of Medical Informatics and was the subject of a presentation in the CNRS INS2I News by Pierrick Bedouch, professor of clinical pharmacy at the Grenoble university hospital and researcher at TIMC:

“As the last line of security before dispensing a medication to the patient, the pharmacist is faced with increasingly complex therapeutic cases. At issue: the gradual shift towards personalized medicine for an aging and sometimes multi-pathological population. Identifying the right medications, their risks of interactions or even drug iatrogenics (all the undesirable effects), turns into a real headache. However, the digitization of patient records paves the way for artificial intelligence (AI) tools tailored to their needs. But to what extent are they deployed? For what uses?

“In order to answer these questions, researchers at TIMC carried out a review of the state of the art on AI methods and applications in clinical pharmacy services. “Pharmacy has always been one of the most digitalized disciplines in health. However, our study shows that between 2000 and 2021, only 19 scientific publications relate to AI tools applied to this field,” presents Pierrick Bedouch. This trend is recent since 63% of this work was published in 2020 and 2021. In this context, two tools are emerging: automatic natural language processing and AI methods based on deep learning.

“In practice, prescribing decision support is the main application targeted by AI tools in clinical pharmacy - or how to help the pharmacist ensure the best therapeutic choice for a given patient. A little less than a quarter of the cases concerned the dispensing of medicines. For example, models make it possible to anticipate the preparation of prescriptions, others aim to avoid errors during this preparation, or to strengthen patients' therapeutic adherence. Finally, two studies focused on the clinical activity of the pharmacist with the deployment of tools to support therapeutic education or the carrying out of pharmaceutical interviews, with a view to strengthening the advice to be given to patients.

“This study also highlights many perspectives regarding the future of these tools. Indeed, structuring these algorithms requires training them on reliable data from electronic medical records. “To increase the power of these models in France, health data warehouses must be more exhaustive across all stages of therapeutic patient care at the scale of a territory and not just on the data from a hospital,” notes Pierrick Bedouch.

AI tools can help pharmacists manage increasingly complex prescriptions and treatments with patients with multiple pathologies.

Furthermore, new AI methods must be evaluated, such as large language models in automatic language processing – ChatGPT type models. Other larger-scale issues, such as the management of medication management during periods of drug shortage, should in turn benefit from decision support tools.

“The published developments mainly concern the hospital sector (12 studies out of 19). Research must be further developed in outpatient settings, because AI can provide significant assistance in the piloting and complex management of patients with chronic illnesses and in community medicine in general,” concludes the researcher.