What question did this study set out to answer?

June 28, 2026Journal of Experimental OrthopaedicsOpen Access

Promising performance of locally deployed large language models for postoperative orthopaedic patient questions: An In Silico analysis

View Full Paper

Ask AI

Bookmark

Authors

LLLea LanterUniversitätsklinik Balgrist SMSophie MaselUniversitätsklinik Balgrist BHBettina HochreiterUniversity of Zurich

Discussion

Member takes

Implication

Randomized trial evaluates the performance of local and commercial large language models for postoperative patient queries, suggesting implications for patient communication.

Key Points

This study aims to evaluate and compare the performance of local and commercial large language models (LLMs) on postoperative patient questions.
Twenty postoperative patient questions were posed to four different LLMs: GPT‐5, Claude 4.5 Sonnet, GPT‐OSS, and Apertus.
Responses were assessed using the QUEST framework, evaluated by four blinded reviewers including two board-certified orthopaedic surgeons.
Overall model performance was compared using the Friedman test with Wilcoxon signed‐rank tests for pairwise comparisons.
Claude 4.5 Sonnet performed best with a mean score of 0.949, followed by GPT‐5 with 0.937, GPT‐OSS with 0.873, and Apertus with 0.693.
Apertus had significant safety issues, with 22.5% of ratings showing harmful content or fabrication.
Inter-rater agreement among reviewers was moderate, with an overall Fleiss' κ of approximately 0.50.

Cite This Study

Lanter et al. (2026) studied this question.

synapsesocial.com/papers/6a40bb4561bb0a67205c6da2 https://doi.org/10.1002/jeo2.70813

View Full Paper

Ask AI

Bookmark

Implication

Randomized trial evaluates the performance of local and commercial large language models for postoperative patient queries, suggesting implications for patient communication.

Promising performance of locally deployed large language models for postoperative orthopaedic patient questions: An In Silico analysis

Discussion

Member takes

Implication

Key Points

Cite This Study

Promising performance of locally deployed large language models for postoperative orthopaedic patient questions: An In Silico analysis

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider