Abstract
Background: Effective project and programme planning is essential in healthcare to ensure that interventions are grounded in evidence and meet patient needs, but healthcare teams often face time constraints to work on projects aimed at tackling healthcare and integrated care challenges and issues - our custom Generative Pre-Trained Transformer (GPT) model aims to address this.
Approach: We developed a custom GPT model, on the OpenAI platform, to facilitate the creation and refinement of essential planning tools such as the Theory of Change (ToC), Driver Diagrams (DD), and Logic Models (LM). These tools are critical in defining project objectives, identifying drivers, and mapping out interventions for tackling healthcare challenges.
The model was developed with a focus on ease of use, ensuring it could be implemented by multidisciplinary teams with varying levels of experience in project design. The GPT model simplifies the planning process by asking a series of directed questions, multidisciplinary teams in designing the ToC, DD and LM tailored to the specific issues they are addressing or the projects they are planning. We engaged healthcare teams with an interest in tackling specific healthcare issues to use the GPT model to develop the ToC, DD and LM for their projects.
Results: The use of our custom GPT model significantly improved healthcare teams' ability to develop the essential planning tools needed for integrated care programme design. Participants reported increased confidence in using and developing the ToC, DD, and LM, as well as greater clarity in articulating their integrated care project goals. The model’s ability to ask guided questions and suggest refinements enabled teams to focus on the content of their projects rather than the technicalities of the planning tools themselves.
Furthermore, the use of the GPT model fostered improved collaboration among multidisciplinary team members. By providing a structured, step-by-step process, the model facilitated clearer communication on the project’s desired outputs and outcomes, leading to more actionable and effective project plans. Teams reported that the structured output of the GPT model made it easier to align their integrated care projects with broader organisational objectives and meet tight deadlines without sacrificing the quality of their plans.
Implications: The potential of AI tools, like our custom GPT model, to transform healthcare project planning is significant. By automating and guiding the development of ToC, DD, and LM, the GPT model reduces the time and cognitive load required for project design. This, in turn, allows healthcare teams to focus on the strategic elements of their projects, ensuring well-structured, data-driven interventions that are more likely to achieve meaningful integrated care outcomes.
We plan to further integrate the GPT model into ongoing and future projects across our organisation, using the insights gained from initial feedback to refine its functionality. Additionally, we aim to explore new applications of AI to strengthen programme planning and evaluation, particularly in areas like quality improvement and value based healthcare.