Comparative Study of LLM Responses to 15 Pediatric Dentistry Queries

Sanjeev B. Khanagar published a research paper on figshare in 2026. The document presents a comparative cross-sectional feasibility study evaluating the accuracy, quality, reliability, and readability of responses from three large language models to 15 common pediatric dentistry questions. The study includes expert evaluations using standardized tools and statistical analysis of the results.

Use Cases

• Benchmarking LLM performance in medical domains based on the comparative study design described.
• Analyzing inter-examiner reliability in clinical evaluations based on the use of ICC and Bland-Altman analysis mentioned.
• Assessing the readability of AI-generated health content based on the application of Flesch Reading Ease and Flesch–Kincaid Grade Level scores.
• Studying methodological challenges in evaluating AI outputs based on the reported variability in accuracy and expert ratings.

Strengths

• The study is structured with a clear methodology, using 15 standardized queries and three calibrated expert evaluators.
• Evaluation employs multiple validated metrics: Global Quality Scale, modified DISCERN tool, Accuracy of Information Index, and readability scores.
• Statistical analysis is detailed, including a linear mixed-effects model, intraclass correlation coefficients, and Bland-Altman analysis.

Limitations

• The dataset is a single 80.4 KB PDF document; the underlying raw data or response texts are not provided as a structured dataset.
• Row count and column-level documentation are absent; data structure and semantics must be inferred from the paper.
• The study's scale is limited to 15 queries, which may not represent the full breadth of pediatric dentistry topics.

Provenance

Source

figshare

Collection Method

Research study involving structured searches, expert screening of questions, and standardized submission to LLMs.

Time Range

The study was published in 2026; the temporal coverage of the queries is not specified.

Freshness

Last updated 2026-04-24 12:58:40; freshness should be verified.

Geography

Geographic coverage of the queries or study context is not specified.