The Study
Detection of young-onset type 2 diabetes using deep learning across primary and secondary care: a nationwide, retrospective cohort study.
This study looked at lots of medical records to find patterns that might help guess who will get diabetes soon. It doesn't prove that those patterns cause diabetes — it just shows that people with certain records are more likely to get it later.
Analysis score
Maximum 0 for a computational/algorithm study.
Where the score came from
Scientists used a computer program to look at people's medical records and guess who might get type 2 diabetes before they're diagnosed.
Where does this study sit?
Reviews of RCTs (Meta-analyses)
Max 100Randomized Trials
Max 90Reviews of Cohort Studies
Max 85Cohort Studies
Max 72Reviews of Case-Control Studies
Max 63Case-Control Studies
Max 58Cross-Sectional & Case Series
Max 50Expert Opinion
Max 50 / 100
Quality score
Based on clinical experience or non-systematic literature reviews. The lowest level of evidence as they are most susceptible to bias and personal perspective.
Key takeaways
Summary
Based on the study abstract and findings.
- 1Even though it only flags a tiny fraction of people, this could help doctors screen those most likely to get diabetes early — before serious damage happens.
- 2The computer was right 118 times more often for the highest-risk 0.1% of people compared to the average person, and caught 0.23% of all future cases with very few false alarms.
Score breakdown, methodology, conflicts of interest, evidence analysis & raw study data
Publication
Journal
The Lancet. Digital health
Year
2026
Authors
C. H. Johansen, J. Hjaltelin, D. Placido, Samuel Cadell, L. Mortensen, F. Waldorff, A. D. Haue, S. Brunak
Related Content
Claims (6)
A machine learning model using routine medical records can identify people who are in the highest 0.1% of risk for developing type 2 diabetes before age 40, with 118.1 times higher risk than the average person, when predicting onset 3 to 15 months ahead.
Using both primary and secondary healthcare records together predicts young-onset type 2 diabetes more accurately than using either set of records alone, and the most accurate model identifies people at 97.2 times higher risk of developing diabetes within 3 to 12 months.
People prescribed cardiovascular medications such as blood pressure or cholesterol drugs are more likely to be diagnosed with type 2 diabetes within one to two years, suggesting these prescriptions coincide with early metabolic changes.
A deep learning model used for medical predictions performs consistently across five different regions in Denmark, even though those regions differ in how healthcare is delivered, how data is collected, and the characteristics of their populations.
A computer model identifies 0.23% of people who will develop type 2 diabetes before age 40, using data to predict their diagnosis 3 to 15 months ahead, with only 5% of its predictions being false positives.
Deep learning models can identify young-onset type 2 diabetes using routine medical records from primary and secondary care.
Not medical advice. For informational purposes only. Always consult a qualified healthcare professional before making health decisions.