Background

Digitalization has made significant progress in many areas, and healthcare is no exception. The sector is seeing promising new approaches in diagnostics and therapy. Specifically, methods that use Artificial Intelligence (AI) carry high expectations. They hold great potential to significantly improve clinical reasoning (i.e., reflective decision-making processes involved in patient care) and offer new ways of analysing patient data, uncovering patterns that traditional methods may overlook. This, in turn, can lead to more informed decisions and targeted therapeutic interventions.The present research project contributes to this effort by investigating the extent to which existing AI technologies can support clinical reasoning in physiotherapy.

Project Objectives

Language models such as ChatGPT have already transformed workflows in many domains. The development is still at an early stage, and it is difficult to predict what else can be achieved with AI applications. In clinical settings, so-called reasoning models show particular promise for therapy and treatment. Whether these models are reliable and valid enough to support clinical reasoning is an open question—especially in hypothesis-driven fields such as physiotherapy. Here we try to give an answer by examining how reasoning models can improve outcomes in physiotherapy practice.

Methodology

We use a mixed-methods approach to examine the potential benefits of AI-powered reasoning models in physiotherapy. The study compiles five representative cases from clinical practice and asks three widely used reasoning models for advice. Physiotherapy experts then evaluate the models’ responses, focusing on professional quality and practical applicability.

The analysis is guided by three key questions:

• On what basis do AI-based reasoning models develop working hypotheses?
• Which diagnostic tests do they select, and why?
• How do they plan treatments, what therapies do they suggest, and how do they justify their choices?

Outcome

This project investigates the ways in which artificial intelligence can assist with key tasks in physiotherapy. It also sets the stage for future research that goes beyond the scope of this study and further explores the usefulness of AI language models in the healthcare sector.

Funding