0032/2026 - Aderência de Modelos De Inteligência Artificial às Diretrizes Brasileiras para a Obesidade na Atenção Primária à Saúde Adherence of Artificial Intelligence Models to Brazilian Guidelines for Obesity in Primary Health Care
A obesidade representa um desafio crescente para a Atenção Primária à Saúde (APS), demandando a utilização de novas tecnologias e soluções inovadoras. Este estudo avaliou a aderência de Modelos de Linguagem de Grande Escala (LLMs) às diretrizes brasileiras para o manejo da obesidade na APS. Conduziu-se estudo experimental in silico em duas fases. Inicialmente, 62 LLMs passaram por triagem de latência com 24 atingindo o critério de usabilidade (<10s) e seguindo para a avaliação de conteúdo. A aderência clínica foi mensurada por 16 questões extraídas literalmente do Protocolo Clínico (PCDT), cobrindo diagnóstico, monitoramento e tratamento, revisada por pares (Kappa 0,68-1,00). A exclusão de 38 modelos (61%) por latência evidenciou barreiras de infraestrutura. Nos 24 modelos viáveis, a aderência foi heterogênea e limitada, com o melhor desempenho atingindo apenas 61,1% de conformidade. A correlação entre tamanho do modelo e acerto foi moderada, mas a multimodalidade mostrou-se preditora de melhor desempenho. Conclui-se que os LLMs atuais não oferecem segurança para atuação autônoma na APS, exigindo curadoria profissional rigorosa, validação contínua e mecanismos de escalonamento para o cuidado humano para garantir a segurança do usuário.
Palavras-chave:
Saúde digital; inteligência artificial; sobrepeso; obesidade; atenção primária à saúde.
Abstract:
Obesity represents a growing challenge for Primary Health Care (PHC), demanding the use of new technologies and innovative solutions. This study evaluated the adherence of Large Language Models (LLMs) to Brazilian guidelines for obesity management in PHC. An in silico experimental study was conducted in two phases. Initially, 62 LLMs underwent latency screening, with 24 meeting the usability criterion (<10s) and proceeding to content evaluation. Clinical adherence was measured by 16 questions extracted verbatim from the Clinical Protocol (PCDT), covering diagnosis, monitoring, and treatment, and was peer-reviewed (Kappa 0.68–1.00). The exclusion of 38 models (61%) due to latency highlighted infrastructure barriers. Among the 24 viable models, adherence was heterogeneous and limited, with the best performance reaching only 61.1% compliance. The correlation between model size and accuracy was moderate, but multimodality proved to be a predictor of better performance. It is concluded that current LLMs do not offer safety for autonomous operation in PHC, requiring rigorous professional curation, continuous validation, and escalation mechanisms to human care to ensure user safety.
Keywords:
Digital health; artificial intelligence; overweight; obesity; primary health care.
Adherence of Artificial Intelligence Models to Brazilian Guidelines for Obesity in Primary Health Care
Resumo (abstract):
Obesity represents a growing challenge for Primary Health Care (PHC), demanding the use of new technologies and innovative solutions. This study evaluated the adherence of Large Language Models (LLMs) to Brazilian guidelines for obesity management in PHC. An in silico experimental study was conducted in two phases. Initially, 62 LLMs underwent latency screening, with 24 meeting the usability criterion (<10s) and proceeding to content evaluation. Clinical adherence was measured by 16 questions extracted verbatim from the Clinical Protocol (PCDT), covering diagnosis, monitoring, and treatment, and was peer-reviewed (Kappa 0.68–1.00). The exclusion of 38 models (61%) due to latency highlighted infrastructure barriers. Among the 24 viable models, adherence was heterogeneous and limited, with the best performance reaching only 61.1% compliance. The correlation between model size and accuracy was moderate, but multimodality proved to be a predictor of better performance. It is concluded that current LLMs do not offer safety for autonomous operation in PHC, requiring rigorous professional curation, continuous validation, and escalation mechanisms to human care to ensure user safety.
Palavras-chave (keywords):
Digital health; artificial intelligence; overweight; obesity; primary health care.
Couto, FFS, Almeida, C.P.B.. Aderência de Modelos De Inteligência Artificial às Diretrizes Brasileiras para a Obesidade na Atenção Primária à Saúde. Cien Saude Colet [periódico na internet] (2026/jan). [Citado em 14/08/2026].
Está disponível em: http://cienciaesaudecoletiva.com.br/artigos/aderencia-de-modelos-de-inteligencia-artificial-as-diretrizes-brasileiras-para-a-obesidade-na-atencao-primaria-a-saude/19930?id=19930&id=19930
