Can artificial intelligence manage malnutrition? A critical look at ChatGPT's performance in geriatric nutrition

Cigiloglu, Ahmet; Efendioglu, Eyyup Murat; Ozturk, Zeynel Abidin

doi:10.1590/1806-9282.20252286

Sumário Anterior Atual Seguinte

Sumário

Original Article • Rev. Assoc. Med. Bras. 72 (05) • 2026 • https://doi.org/10.1590/1806-9282.20252286 linkcopy

Can artificial intelligence manage malnutrition? A critical look at ChatGPT's performance in geriatric nutrition

Authorship SCIMAGO INSTITUTIONS RANKINGS

BACKGROUND: Artificial intelligence represents a rapidly advancing innovation in healthcare with the potential to revolutionize the field of clinical nutrition.

OBJECTIVE: The aim of this study was to evaluate ChatGPT's potential to support the clinical decision-making process regarding nutrition in older adults.

METHODS: Twelve questions and three clinical vignettes addressing fundamental concepts of malnutrition, including general information, diagnosis, follow-up, and treatment, were created and asked to ChatGPT. Three geriatricians independently examined ChatGPT's responses. The quality of the responses was assessed using the Quality Analysis of Medical Artificial Intelligence tool.

RESULTS: The inter-rater reliability among the authors was calculated, and an excellent intraclass correlation coefficient of 0.84 (95%CI 0.77–0.89; p<0.001) was found. The total mean Quality Analysis of Medical Artificial Intelligence score for the ChatGPT-generated responses to questions related to malnutrition was 26.60, indicating very good quality. In evaluating the clinical scenarios, the lowest scores were observed in source use. The total Quality Analysis of Medical Artificial Intelligence, accuracy, relevance, and use of sources scores for the clinical scenario involving the patient with a hip fracture were statistically significantly lower compared to other scenarios.

CONCLUSION: Our study highlighted that ChatGPT has the potential to generate correct answers related to complex clinical scenarios about malnutrition. ChatGPT can help clinicians make more informed decisions regarding patients’ nutritional requirements and management by utilizing more up-to-date medical resources and guidelines.

KEYWORDS:
Artificial intelligence; ChatGPT; Malnutrition; Older adults; Accuracy

		QAMAI score	p
Accuracy			0.717
	General information	4.70±0.17
	Diagnosis	4.50±0.50
	Treatment/follow-up	4.53±0.12
	Total	4.58±0.29
Clarity			0.863
	General information	4.63±0.15
	Diagnosis	4.50±0.50
	Treatment/follow-up	4.53±0.12
	Total	4.56±0.27
Relevance			0.604
	General information	4.63±0.15
	Diagnosis	4.33±0.58
	Treatment/follow-up	4.47±0.12
	Total	4.48±0.33
Completeness			0.709
	General information	4.50±0.10
	Diagnosis	4.33±0.58
	Treatment/follow-up	4.27±0.12
	Total	4.37±0.32
Sources			0.012*
	General information	4.05±0.26
	Diagnosis	3.83±0.29
	Treatment/follow-up	4.67±0.12*
	Total	4.18±0.43
Usefulness			0.520
	General information	4.63±0.15
	Diagnosis	4.33±0.58
	Treatment/follow-up	4.33±0.12
	Total	4.43±0.34
Total QAMAI score		26.60

	Scenario 1	Scenario 2	Scenario 3	p
Accuracy	4.27±0.52	4.42±0.50	4.00±0.76*	0.048*
Clarity	4.24±0.56	4.45±0.51	4.09±0.74	0.081
Relevance	4.30±0.58	4.48±0.51	4.09±0.70*	0.050*
Completeness	4.21±0.55	4.25±0.44	4.03±0.72	0.379
Sources	4.18±0.53	4.15±0.52	3.70±0.70*	0.008*
Usefulness	4.27±0.52	4.42±0.50	4.03±0.78	0.054
Total QAMAI score	25.48	26.18	23.94*	0.000*

location_on

Associação Médica Brasileira R. São Carlos do Pinhal, 324, 01333-903 São Paulo SP - Brazil, Tel: +55 11 3178-6800, Fax: +55 11 3178-6816 - São Paulo - SP - Brazil
E-mail: ramb@amb.org.br

rss_feed Acompanhe os números deste periódico no seu leitor de RSS

[TFN1] ^*
p<0.05;

[TFN1a] the QAMAI scores are presented as mean±SD. QAMAI: Quality Analysis of Medical Artificial Intelligence.