Actionable companion · Assessment and evidence of learning

From evidence to district action: Performance of large language models in mitral valve surgery patient education: a comparative analysis

AI-assisted editorial implementation guide. The district actions are proposed applications, not findings or recommendations attributed to the original authors. No additional peer review is asserted.

Evidence and context

Health-professions education context: test transfer to K–12 rather than assuming it.

Original authors’ abstract, reproduced under the linked source license:

Background Large language models (LLMs), a form of artificial intelligence, are increasingly being utilized in healthcare to support patient education and information delivery. The aim of this study was to perform a comparative analysis of five different LLMs (i.e., ChatGPT-4o, Claude 3.7 Sonnet, Gemini 2.5 Pro Preview, DeepSeek-V3, and Microsoft Copilot) in terms of accuracy, completeness, and readability, based on their responses to frequently asked questions in preoperative patient education for mitral valve surgery (MVS). Methods A standardized questionnaire comprising seven frequently asked questions by patients prior to MVS was developed. Prompting procedures and model parameters were fully reported to support reproducibility. These questions were presented to each LLM in an identical manner. The responses were evaluated by two academic experts in cardiac surgery using structured assessment criteria across three main dimensions: accuracy, completeness, and readability. For the readability analysis, the Simplified Measure of Gobbledygook (SMOG) Index and the Flesch-Kincaid Grade Level (FKGL) scale were utilized. Results The ChatGPT-4o and Gemini 2.5 Pro Preview models received statistically significantly higher scores than Claude 3.7 Sonnet and Microsoft Copilot for both accuracy (median 5 for ChatGPT-4o and Gemini 2.5 Pro Preview vs. 4 for Claude 3.7 Sonnet and Microsoft Copilot, p < 0.001) and completeness (median 5 for Gemini 2.5 Pro Preview vs. 3 for Claude 3.7 Sonnet, p < 0.001). Claude 3.7 Sonnet achieved the highest readability scores, with significantly lower SMOG (10.90 for Claude 3.7 Sonnet vs. 12.24 for ChatGPT-4o, p = 0.006) and FKGL (8.0 for Claude 3.7 Sonnet vs. 9.04 for ChatGPT-4o, p = 0.004) scores, indicating simpler and more comprehensible sentence structures. Significant differences were observed among the evaluated models across all three assessment dimensions (p < 0.001 for all comparisons). Conclusions The LLMs represent valuable supplementary tools in patient education processes. However, their implementation in clinical practice must be carefully evaluated, particularly with regard to accuracy and completeness. This study highlights the potential applicability of ChatGPT-4o and Claude 3.7 Sonnet models for preoperative patient education in MVS, while emphasizing that all LLMs should be used under the supervision and guidance of healthcare professionals. For LLMs to be reliably utilized in the medical field, improvement in medical accuracy and standardization are essential. Supplementary Information The online version contains supplementary material available at https://doi.org/10.1186/s12911-026-03662-3.

A 90-day district pilot

1. Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
2. Days 16–30 · Design. Redesign one low-stakes assessment: require a first attempt, an AI-assisted revision, a source-check log and a short oral explanation. Use the same rubric for the comparison task. Name an accountable owner and set the baseline, success threshold and budget before implementation.
3. Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
4. Days 61–90 · Decide. Blind-score reasoning, factual accuracy and unaided transfer on a fresh task. Record suspected misconduct separately from demonstrated evidence. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Banu Bahriye Akdag; Mehmet Senel Bademci; Ihsan Peker; Guven Okay Karaca; Cagrı Kandemir; Barcin Özcem; Huseyin Durmaz; Meryem Cakır; Irem Ozcetin; Hidayet Onur Selcuk. Performance of large language models in mitral valve surgery patient education: a comparative analysis. BMC Medical Informatics and Decision Making (2026-06-29). https://doi.org/10.1186/s12911-026-03662-3

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.