JMIR Publications

ABSTRACT

Background:

Asthma is a complex condition marked by persistent airway inflammation, prompting numerous investigations into the potential applications of artificial intelligence within the realm of asthma management. ChatGPT, a robust language model, has demonstrated utility across various domains such as medical education, physician consultation, medical content creation, and recommendations for lung cancer screening.

Objective:

This study seeks to evaluate the viability of ChatGPT-4.0 as a tool for supporting asthmatic patients and caregivers in the medical setting.

Methods:

A total of 33 questions regarding six aspects of asthma were gathered for analysis using ChatGPT-4.0. The responses were assessed by three respiratory specialists and categorized as "appropriate", "appropriate but insufficient", "helpful", or "inconsistent". Percentages were computed to determine the distribution of evaluations across the six aspects.

Results:

The study revealed that 60% of basic information and diagnosis were deemed "appropriate," while 40% were considered "appropriate but insufficient." In terms of clinical features, 75% were classified as "appropriate" and 25% as "helpful." Regarding treatment, 66.7% were deemed "appropriate," with 33.3% falling under the category of "appropriate but insufficient." In the assessment and management section, 81.8% were labeled as "appropriate," and 18.2% as "appropriate but insufficient." The differential diagnosis part scored 100% in appropriateness. COVID-19 related sections were rated as 75% "appropriate" and 25% "appropriate but insufficient".

Conclusions:

The results suggest that ChatGPT-4.0 could serve as a valuable supplementary tool in various aspects of asthma identification, treatment, assessment, and management beyond conventional medical practices.