Effect of the Digital Livsstilsverktyget in Conjunction With a Large Language Model on the Prevention of Type 2 Diabetes

New scalable means to prevent type 2 diabetes is needed. AI-based large language models enable new means to support people in lifestyle changes. It is, however, important that these are coupled with serious health information.

The investigators have previously shown that the digital Livsstilsverktyget ('Lifestyle tool') developed at the University of Gothenburg leads to improved glycemic controls in patients using it regularly. To make the intervention more personal the investigators have coupled it with a large language model (LLM) to promote reflection on the expert-written content of the Lifestyle tool. Thus, participants interact with the LLM to explore a current area of interest. They then get recommendations from the LLM on relevant texts from the Lifestyle tool. These texts are written by experts (physicians, researchers). The participants go through the text and are then stimulate by the LLM to reflect on how to use the content in daily life.

This study will recruit people without type 2 diabetes, aged 35 years or older, from the general population. They will be randomly assigned to the intervention or to a control group without access to the intervention. The primary outcome is the incidence type 2 diabetes during three years. The primary variable, diagnosis of type 2 diabetes, is obtained via national health registers. Thus, no physical visits or blood sampling are required. The participants will complete the Findrisc questionnaire online at baseline to assess the risk for type 2 diabetes and enable effect analysis of individuals with different risk.

The study is an investigator-initiated single-center study conducted over three years. Technical problems are referred to a study coordinator, who also responded to requests to clarify content in a general manner without providing personal advice. Study participants are managed by their ordinary healthcare providers throughout the study.