Computers can spot young-onset type 2 diabetes years before diagnosis using just routine medical records.

Other research shows that computer programs can indeed predict who will develop type 2 diabetes at a young age by analyzing everyday medical data like blood sugar levels and weight, sometimes years before a diagnosis is made. These models are especially good at identifying the highest-risk individuals and outperform older prediction methods. While the post mentions deep learning, the studies supporting this don’t all use that specific technique—but they still prove the core idea works: your doctor’s notes and lab results hold hidden clues. The accuracy is impressive, and the fact that it works with routine data makes it feel surprisingly practical.