A carregar...
Publicação
Erros de medicação: a exatidão e os riscos do aconselhamento gerado por LLM na utilização de medicamentos não sujeitos a receita médica
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.81 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A evolução da Inteligência Artificial (IA) impulsionou o avanço dos Large Language Models (LLMs), modelos treinados a partir de grandes bases de dados para gerar texto em linguagem humana e com a capacidade de executar diversas tarefas, como a resposta a perguntas num dado contexto. Estes modelos têm vindo a demonstrar ter aplicabilidade nas mais diversas áreas, entre as quais a área da Saúde, onde está incluído o Aconselhamento Farmacêutico.
Os medicamentos não sujeitos a receita médica (MNSRM) estão ao dispor de todos para a prática da automedicação visando o tratamento de queixas ligeiras. Do enanto, a intervenção farmacêutica é fundamental para que haja uma avaliação correta da situação clínica e seleção da terapêutica mais adequada. É essencial que seja fornecida a informação necessária para que haja uma utilização correta e segura dos medicamentos de modo a evitar a ocorrência de erros de medicação que ponham em causa a saúde dos doentes. A adoção de LLMs como ferramenta de auxílio para o aconselhamento de MNSRM poderá ser bastante útil aos farmacêuticos comunitários, apesar de apresentar alguns riscos e limitações.
Esta monografia tem como objetivo estudar o impacto do uso de LLMs no aconselhamento de MNSRM e quais as vantagens que a sua integração nas farmácias poderá vir a trazer tanto aos profissionais como aos doentes. Para tal, é imperativo abordar não só os seus benefícios, mas também os riscos e limitações inerentes à utilização destes modelos na prática farmacêutica, como, por exemplo, os erros de medicação.
Com este objetivo, realizou-se uma análise bibliográfica e foi elaborado um caso de estudo de um caso clínico no qual se utilizaram três LLMs diferentes (GhatGPT, Copilot e Gemini) de modo analisar os resultados obtidos relativamente ao aconselhamento farmacêutico e perceber se são coerentes com a informação de guidlines terapêuticas. Podemos concluir que a integração de LLMs no aconselhamento poderá vir a ter um importante papel enquanto ferramentas de auxílio, mas não devem substituir a intervenção de um profissional de saúde especializado.
The evolution of Artificial Intelligence (AI) has driven the advance of Large Language Models (LLMs), models trained from large databases to generate human-like text and with the ability to perform various tasks, such as answering questions in a given context. These models have proven to be applicable in a wide variety of areas, including healthcare, which includes pharmaceutical counselling. Non-prescription medicines are available to everyone for self-medication to treat minor complaints. However, pharmaceutical intervention is essential in order to correctly assess the clinical situation and select the most appropriate therapy. It is essential that the necessary information is provided for the correct and safe use of medicines in order to avoid medication errors that jeopardise patients' health. The adoption of LLMs as a tool to help with pharmaceutical counselling could be very useful for community pharmacists, although it does present some risks and limitations. The aim of this monograph is to study the impact of using LLMs in MNSRM counselling and what advantages their integration into pharmacies could bring to both pharmacists and patients. To this end, it is imperative to address not only their benefits, but also the risks and limitations inherent in the use of these models in pharmacy practice, such as medication errors. To this end, a literature review was conducted, and a case study of a clinical case was developed in which three different LLMs were used (GhatGPT, Copilot and Gemini) in order to analyse the results obtained regarding pharmaceutical counselling and whether they are consistent with the information in therapeutic guidelines. We can conclude that the integration of LLMs into counselling may have an important role as tools of assistance but should not replace the intervention of a specialised health professional.
The evolution of Artificial Intelligence (AI) has driven the advance of Large Language Models (LLMs), models trained from large databases to generate human-like text and with the ability to perform various tasks, such as answering questions in a given context. These models have proven to be applicable in a wide variety of areas, including healthcare, which includes pharmaceutical counselling. Non-prescription medicines are available to everyone for self-medication to treat minor complaints. However, pharmaceutical intervention is essential in order to correctly assess the clinical situation and select the most appropriate therapy. It is essential that the necessary information is provided for the correct and safe use of medicines in order to avoid medication errors that jeopardise patients' health. The adoption of LLMs as a tool to help with pharmaceutical counselling could be very useful for community pharmacists, although it does present some risks and limitations. The aim of this monograph is to study the impact of using LLMs in MNSRM counselling and what advantages their integration into pharmacies could bring to both pharmacists and patients. To this end, it is imperative to address not only their benefits, but also the risks and limitations inherent in the use of these models in pharmacy practice, such as medication errors. To this end, a literature review was conducted, and a case study of a clinical case was developed in which three different LLMs were used (GhatGPT, Copilot and Gemini) in order to analyse the results obtained regarding pharmaceutical counselling and whether they are consistent with the information in therapeutic guidelines. We can conclude that the integration of LLMs into counselling may have an important role as tools of assistance but should not replace the intervention of a specialised health professional.
Descrição
Trabalho Final de Mestrado Integrado, Ciências Farmacêuticas, 2024, Universidade de Lisboa, Faculdade de Farmácia.
Palavras-chave
Erros de medicação Inteligência artificial LLMs Aconselhamento farmacêutico Mestrado integrado - 2024