{
  "abstract": "Background Artificial intelligence (AI) is reshaping healthcare, with Large Language Models (LLMs), such as ChatGPT and Google Gemini, gaining traction. LLMs offer opportunities for enhanced decision-making and patient engagement but their integration into healthcare remains unexplored. This includes cardiovascular rehabilitation, where patient communication, education, and behaviour change are central. There is significant confusion around what the various AI-related terms mean, and rehabilitation professionals must be equipped to critically appraise and contribute to the development of AI tools to ensure they are safe, equitable, and appropriate for use in practice.Aim To systematically map the existing literature on how LLM-generated data is evaluated in healthcare (including cardiovascular rehabilitation) contexts, highlighting best practices and identifying gaps in current evidence and approaches.Methods This scoping review is being conducted in accordance with Joanna Briggs Institute methodology. The search strategy includes peer-reviewed and grey literature. Eligible studies will focus on the evaluation of outputs generated by LLMs in healthcare settings and in particular cardiovascular rehabilitation. The review will also map out the key terms and concepts being used in AI-based healthcare.Results The review is currently in progress. Findings will summarise the how LLMs are being used in healthcare settings, the range of evaluation strategies used, highlight inconsistencies in reporting and methodological robustness, and identify key domains that require standardisation. Particular attention will be paid to how these insights might inform the safe future use of LLMs in developing or supporting individualised AI-based cardiovascular rehabilitation programmes.Conclusion This review will identify best practices and current limitations in the evaluation of LLM-generated healthcare content. The study will highlight the critical role of rehabilitation professionals in ensuring the safe integration of AI tools in cardiovascular rehabilitation. Results will be presented at the conference.",
  "authors": [
    {
      "affiliations": [
        "Edinburgh Napier University. Centre for Cardiovascular Health. 9 Sighthill Court, Edinburgh, EH11 4BN"
      ],
      "name": "Natalie Elliott"
    },
    {
      "affiliations": [
        "Edinburgh Napier University. School of Computing, Engineering and The Built Environment, 10 Colinton Road, Edinburgh EH10 5DT"
      ],
      "name": "Alistair Lawson"
    },
    {
      "affiliations": [
        "Edinburgh Napier University. Centre for Cardiovascular Health. 9 Sighthill Court, Edinburgh, EH11 4BN"
      ],
      "name": "Coral L Hanson"
    },
    {
      "affiliations": [
        "Edinburgh Napier University. Centre for Cardiovascular Health. 9 Sighthill Court, Edinburgh, EH11 4BN"
      ],
      "name": "Sheona McHale"
    }
  ],
  "title": "P15 Artificial Intelligence (AI) in cardiovascular rehabilitation: a scoping review of methods for assessing AI-generated healthcare content",
  "uid": "3cb2662e-e446-5c79-bc9d-ad30a3ce3be2"
}