JMIR Publications

ABSTRACT

Background:

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder that affects functioning and quality of life across the lifespan. Despite extensive online information, patients and caregivers continue to report unmet needs, particularly regarding diagnosis, treatment, medication effects, comorbidities, and long-term management. Provider-generated FAQs are widely used but may not reflect concerns expressed in online communities.

Objective:

To assess (1) semantic coverage of provider-generated ADHD FAQs for online community questions, (2) topics among low-similarity questions, and (3) differences in response style between provider-generated and community responses using LLM-assisted analysis.

Methods:

ADHD-related questions from a Korean online community were compared with provider-generated FAQs. Sentence embedding–based semantic similarity analysis estimated coverage and identified matched and unmatched questions. Unmatched questions were analyzed using the LimTopic framework (BERTopic with LLM-assisted summarization). High-similarity FAQ–community pairs were examined using LLM-assisted content classification.

Results:

At the optimal threshold, 88.8% (4,427/4,988) of questions were matched, with coverage concentrated in 18 FAQs. Topic modeling identified 18 clusters consolidated into five unmet need categories. Provider-generated responses were predominantly informational and neutral, whereas community responses more often included emotional and experiential elements.

Conclusions:

Despite high overall coverage, structural concentration and residual gaps indicate persistent unmet needs. Differences in response style suggest informational adequacy and communicative resonance represent distinct dimensions of alignment, supporting the need for more responsive, consumer-centered ADHD information strategies.