ChatGPT in healthcare: perceptions, ethical considerations, and practice implications among healthcare professionals in Ecuador and other countries in the Americas: a cross-sectional survey study

Résumé fourni par la source

Background Generative artificial intelligence tools, such as ChatGPT, are increasingly discussed in healthcare; however, evidence from Latin American professional settings is limited and must be interpreted in light of regional digital inequities and ethical concerns. Objective To examine awareness, ethical perceptions, usage patterns, and attitude determinants related to ChatGPT among HCPs practicing mainly in Ecuador and other countries in the Americas. Methods We conducted a cross-sectional online survey between May and September 2023. Recruitment was conducted using professional networks, institutions, conferences, and social media. Descriptive analyses were performed for the entire sample. Ethical and usefulness analyses were restricted to respondents with at least minimal knowledge of ChatGPT ( n = 424), and models involving job facilitation, perceived risk, and future intention were restricted to prior users with complete data ( n = 218–230, depending on the model). Adjusted linear and binary logistic regression models controlled for age, gender, professional group, years of experience, and country (Ecuador vs. other). Results Of the 747 respondents, 610 (81.66%) practiced in Ecuador, 281 (37.62%) were physicians, 203 (27.18%) were dentists, and 42 (5.62%) were nurses or allied health professionals. Overall, 424 (56.76%) had heard of ChatGPT, 239 (31.99%) had ever used it, and 323 (43.24%) reported no knowledge of it. Among respondents with at least minimal familiarity, 311/394 (78.93%) agreed that ChatGPT could be beneficial in healthcare settings, 256/389 (65.81%) considered it trustworthy for healthcare information, and 286/391 (73.15%) considered it useful for specific medical questions. Ethical judgments were mixed: 179/395 (45.32%) were neutral about language revision in a scientific manuscript, 189/395 (47.85%) considered AI-generated manuscript text unethical, and 268/395 (67.85%) considered sole-source clinical use unethical. Among prior users, the most common uses were homework or academic support (134/239, 56.07%), research writing support (117/239, 48.95%), and medical or healthcare education and training (55/239, 23.01%). Higher knowledge and ethical acceptance scores were independently associated with more favorable attitudes. Ethical acceptance predicted stronger agreement that ChatGPT could be beneficial in healthcare settings ( β = 0.30, 95% CI 0.19 to 0.41) and trustworthy ( β = 0.32, 95% CI 0.20 to 0.44). Future weekly-or-more use was more likely among respondents who viewed ChatGPT as beneficial in healthcare (adjusted OR 1.95, 95% CI 1.39 to 2.76) and useful for specific medical questions (adjusted OR 1.97, 95% CI 1.46 to 2.65). Conclusions Healthcare professionals in this sample showed cautious optimism toward ChatGPT, but awareness remained moderate, use was limited, and ethical reservations were substantial. Structured, context-specific training and more regionally representative studies are needed before large-scale integration can be recommended.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.