Evaluación cualitativa de modelos de inteligencia artificial generativa para resolución de preguntas clínicas de rehabilitación infantil

Introduction: The use of artificial intelligence in the healthcare field has shown multiple potentials. Among them, using generative artificial intelligence models, such as ChatGPT, to support the creation of clinical documents like reviews, clinical guidelines, or protocols. The objective of this study is to evaluate and compare the responses to clinical questions defined by experts on two rehabilitation topics, provided by the most widely used generative AI models on the market, in order to analyze their role in the creation of clinical documents, such as clinical guidelines or protocols. Material or Patients and Methods: Qualitative descriptive study. Through prompts of various types of clinical questions, an expert evaluates and rates, using a scale design for this study, the responses provided by 5 generative artificial intelligence models: ChatGPT, Gemini, Claude, Perplexity, and a customized GPT. Results: All generative artificial intelligence models are capable of delivering well-structured and coherent responses, but with some shortcomings in their technical content and updates according to evidence-based medicine. ChatGPT received the highest ratings, using the study rating scale. Discussion: Generative artificial intelligence models can play a role in the creation of clinical documents, such as reviews, guidelines, and protocols, providing a quick and effective tool. However, each step in the creation of these documents should be supervised by a human expert to identify possible errors, hallucinations and protection of ethic and security issues.