Clinical reasoning using ChatGPT: Is it beyond credibility for physiotherapists use?

Area of Science:

• Healthcare technology
• Clinical decision support systems
• Artificial Intelligence in Medicine

Background:

• Artificial Intelligence (AI) tools are increasingly adopted in healthcare.
• OpenAI's ChatGPT, a language model, offers data analysis and recommendations.
• AI integration in physiotherapy has the potential to improve healthcare efficiency.

Purpose of the Study:

• Evaluate ChatGPT-3.5's consistency and accuracy in clinical responses.
• Assess ChatGPT's ability to mimic human clinical reasoning in varied scenarios.
• Determine ChatGPT's capacity for generating differential diagnoses.

Main Methods:

• Two studies utilized ChatGPT-3.5.
• Study 1: Ten participants queried 'physiotherapy assessment steps' to assess response consistency and accuracy.
• Study 2: Two independent evaluators used published case studies (simple and complex) to assess differential diagnosis capabilities.

Main Results:

• Study 1 revealed response variability, with 30% omitting re-assessment and 40% omitting subjective examination.
• Study 2 showed ChatGPT could generate evidence-based clinical reasoning, especially in simple cases.
• The phrasing of user queries significantly influenced the quality of ChatGPT's output.

Conclusions:

• ChatGPT offers potential benefits for physiotherapy, but limitations exist.
• A balanced view of AI's strengths and weaknesses is essential.
• Responsible and informed use of AI tools in clinical practice is paramount.