Prevention of frailty in the elderly: an expert-based evaluation of measures proposed by artificial intelligence

INTRODUCTION: The use of artificial intelligence (AI), particularly conversational agents such as ChatGPT, is expanding in the healthcare industry. We aimed to assess the responses provided by ChatGPT regarding the prevention of frailty in the elderly. METHODS: The responses provided by ChatGPT 3.5 were evaluated by a panel of 29 experts in the field of geriatric care, comprising geriatricians, geriatric nurses, preventionists, and educators specializing in adapted physical activity. Each proposal was rated on a numerical scale from 0 (not at all appropriate) to 10 (very appropriate) in terms of items relevance, formulation, exhaustiveness and with an overall rating. RESULTS: ChatGPT created 17 themes for a total of 24 items. The 17 themes proposed by the conversational agent are very similar to those proposed in the literature on the prevention of frailty in the elderly. For each theme, the overall rating was consistently higher than the exhaustivity (8, IQR 8 to 10 vs 7, 6 to 8; p < 0.05). The relevance of the items was systematically and significantly higher than their wording (p < 0.05). There were no significant differences in the global evaluation or item exhaustivity based on expert characteristics (gender, age, profession, and seniority). CONCLUSION: Despite ChatGPT exhibits some shortcomings and inconsistencies, AI could be a useful tool for healthcare professionals in the prevention of frailty in the elderly, but it should be used with caution. Further research is required to gain a deeper understanding of the role and potential applications of this tool in the healthcare sector.