Artificial intelligence insights into osteoporosis: assessing ChatGPT's information quality and readability.

Abstract

Accessible, accurate information, and readability play crucial role in empowering individuals managing osteoporosis. This study showed that the responses generated by ChatGPT regarding osteoporosis had serious problems with quality and were at a level of complexity that that necessitates an educational background of approximately 17 years.

Purpose: The use of artificial intelligence (AI) applications as a source of information in the field of health is increasing. Readable and accurate information plays a critical role in empowering patients to make decisions about their disease. The aim was to examine the quality and readability of responses provided by ChatGPT, an AI chatbot, to commonly asked questions regarding osteoporosis, representing a major public health problem.

Methods: "Osteoporosis," "female osteoporosis," and "male osteoporosis" were identified by using Google trends for the 25 most frequently searched keywords on Google. A selected set of 38 keywords was sequentially inputted into the chat interface of the ChatGPT. The responses were evaluated with tools of the Ensuring Quality Information for Patients (EQIP), the Flesch-Kincaid Grade Level (FKGL), and the Flesch-Kincaid Reading Ease (FKRE).

Results: The EQIP score of the texts ranged from a minimum of 36.36 to a maximum of 61.76 with a mean value of 48.71 as having "serious problems with quality." The FKRE scores spanned from 13.71 to 56.06 with a mean value of 28.71 and the FKGL varied between 8.48 and 17.63, with a mean value of 13.25. There were no statistically significant correlations between the EQIP score and the FKGL or FKRE scores.

Conclusions: Although ChatGPT is easily accessible for patients to obtain information about osteoporosis, its current quality and readability fall short of meeting comprehensive healthcare standards.