| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 957.96 KB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
No Setor Farmacêutico existe uma integração crescente dos Large Language Models (LLMs) em diversas áreas, pela sua capacidade extraordinária de resumo científico, análise de dados e produção de texto contextualizado. No âmbito da deteção de interações medicamentosas, os LLMs podem solucionar problemas relacionados com o uso de medicamentos (PRMs) nomeadamente com a diminuição da ocorrência de reações adversas aos medicamentos (RAMs).
O presente trabalho tem como objetivo estudar a aplicabilidade dos LLMs na área Farmacêutica. No seguimento pretendeu-se avaliar algumas limitações dos mesmos, nomeadamente possíveis armadilhas, riscos e efeitos colaterais da sua utilização. Procurou-se ainda definir novas estratégias para assegurar uma utilização segura dos LLMs nesta área.
Foi realizado um estudo comparativo no qual, foi avaliado a exatidão, sensibilidade e especificidade do ChatGPT - 4. ChatGPT - 3.5 e Pharmacy GPT na deteção e previsão de interações medicamentosas em doentes fictícios criados através de inteligência artificial (IA). No sentido de conseguir uma maior aproximação à realidade, o perfil dos doentes foi definido de acordo com as patologias mais prevalentes na população portuguesa, de diferentes especialidades médicas e sendo polimedicados com 10 fármacos concomitantes. Nesta análise contendo na totalidade 135 interações fármacos-fármaco, foram demonstrados resultados que comprovam a necessidade de melhorias para futura implementação na prática clínica.
Como principais resultados verificou-se que nenhum dos modelos estudados apresenta os parâmetros necessários para uma previsão e deteção de interações medicamentosas adequada para implementação na prática clínica, embora tenham sido identificadas algumas vantagens nesta utilização. Existe assim, a necessidade de vigilância contínua e de avaliação de risco, com melhorias na gestão, correção e atualização destas plataformas, para garantir a melhor prestação de cuidados de saúde.
Em conclusão, estando a área Farmacêutica em constante evolução, acompanhar a inovação tecnológica bem como continuar com um papel ativo e interventivo é imperativo para a continuação da formação de excelentes profissionais que tanto caracterizam a Profissão Farmacêutica.
In the pharmaceutical sector, Large Language Models (LLMs) are becoming increasingly integrated in various areas, due to their extraordinary capacity for scientific summarization, data analysis and the production of contextualized text. In the context of detecting drug interactions, LLMs can solve problems related to the use of medicines (PRMs), by reducing the occurrence of adverse drug reactions (ADRs). The aim oh this paper is to study the applicability of LLMs in the pharmaceutical field. The aim was to assess some of their limitations, including possible pitfalls, risks and side effects of their use. It also sought to define new strategies to ensure the safe use of LLMs in this area. A comparative study was carried out to assess the accuracy, sensitivity and specificity of ChatGPT - 4, ChatGPT - 3.5 and Pharmacy GPT in detecting and predicting drug interactions in fictitious patients created using AI. To get closer to reality, the patient profile was defined according to the most prevalent pathologies in the Portuguese population, from different medical specialties and being polymedicated with 10 concomitant drugs. In this analysis of 135 drug-drug interactions, results demonstrate the need for improvements for future implementation in clinical practice. The main results were that none of models studied have the necessary parameters for predicting and detecting drug interactions suitable for implementation in clinical practice, although some advantages have been identified in this use. There is therefore a need for continuous surveillance and risk assessment, with improvements in the management, correction and updating of these platforms, to ensure the best provision of healthcare. In conclusion, as the pharmaceutical field is constantly evolving, keeping up with technological innovation and continuing to play an active and intervening role is imperative for the continued training of the excellent professionals who so characterize the pharmaceutical profession.
In the pharmaceutical sector, Large Language Models (LLMs) are becoming increasingly integrated in various areas, due to their extraordinary capacity for scientific summarization, data analysis and the production of contextualized text. In the context of detecting drug interactions, LLMs can solve problems related to the use of medicines (PRMs), by reducing the occurrence of adverse drug reactions (ADRs). The aim oh this paper is to study the applicability of LLMs in the pharmaceutical field. The aim was to assess some of their limitations, including possible pitfalls, risks and side effects of their use. It also sought to define new strategies to ensure the safe use of LLMs in this area. A comparative study was carried out to assess the accuracy, sensitivity and specificity of ChatGPT - 4, ChatGPT - 3.5 and Pharmacy GPT in detecting and predicting drug interactions in fictitious patients created using AI. To get closer to reality, the patient profile was defined according to the most prevalent pathologies in the Portuguese population, from different medical specialties and being polymedicated with 10 concomitant drugs. In this analysis of 135 drug-drug interactions, results demonstrate the need for improvements for future implementation in clinical practice. The main results were that none of models studied have the necessary parameters for predicting and detecting drug interactions suitable for implementation in clinical practice, although some advantages have been identified in this use. There is therefore a need for continuous surveillance and risk assessment, with improvements in the management, correction and updating of these platforms, to ensure the best provision of healthcare. In conclusion, as the pharmaceutical field is constantly evolving, keeping up with technological innovation and continuing to play an active and intervening role is imperative for the continued training of the excellent professionals who so characterize the pharmaceutical profession.
Description
Trabalho Final de Mestrado Integrado, Ciências Farmacêuticas, 2024, Universidade de Lisboa, Faculdade de Farmácia.
Keywords
ChatGPT Interações medicamentosas Large language models Polimedicação Riscos Mestrado Integrado - 2024
