Scientists have developed a new artificial intelligence–based tool that allows for more accurate assessment of the risk of obesity-related diseases and could potentially help determine who should be prioritized for weight-loss medications. The study was published in the journal Nature Medicine.

Currently, access to such medications in the UK’s National Health Service (NHS) is mainly determined by body mass index (BMI) and the presence of related conditions. However, researchers argue that this is not sufficient: people with the same BMI can have very different risks of developing serious complications.

A team from the University of Cambridge and Queen Mary University of London analyzed data from nearly 200,000 participants in the UK Biobank project with a BMI of 27 or higher — meaning individuals with overweight or obesity. Using an “interpretable machine learning” approach, the researchers identified 20 key factors influencing the likelihood of developing diseases over a 10-year period.

These factors include age, sex, cholesterol levels, kidney function indicators, lifestyle, and other health parameters. Based on this data, the system — named Obscore — estimates the risk of 18 obesity-related diseases, including type 2 diabetes, stroke, gout, and cardiovascular disorders.

One of the most important findings was that some people who were only overweight, rather than obese, were placed in a very high-risk group. According to the researchers, such patients are often overlooked when doctors rely solely on BMI.

The authors believe the tool could help allocate limited healthcare resources more efficiently and identify who truly needs expensive medications such as tirzepatide or Wegovy. At the same time, they emphasize that the goal is not to expand drug use, but to enable a more precise and personalized approach.

Independent experts called the study promising, but noted that further validation and adaptation to real-world healthcare settings will be needed before the system can be implemented in routine medical practice.