What the study found
Electronic health record-based predictive models identified more true cancer cases among people labeled high risk than traditional screening risk factors alone. The authors report that even a baseline approach produced a 3- to 6-fold higher enrichment of true cancer cases, and a foundation model improved predictive performance across 26 cancer types.
Why the authors say this matters
The study suggests that electronic health records, which contain longitudinal patient health information, may help support more precise and scalable early cancer detection than screening criteria based on narrow factors such as age, smoking history, gene mutations, or family history. The authors conclude that this could expand the ability to find high-risk individuals beyond current screening guidelines.
What the researchers tested
The researchers systematically evaluated the clinical utility of EHR-based predictive models against traditional risk factors, including gene mutations and family history of cancer. They used data from the All of Us Research Program, which includes electronic health record, genomic, and survey data from more than 865,000 participants, and examined eight major cancers: breast, lung, colorectal, prostate, ovarian, liver, pancreatic, and stomach.
What worked and what didn't
EHR-based models outperformed traditional risk factors alone, both as a standalone tool and as a complementary tool. The abstract states that the baseline modeling approach achieved a 3- to 6-fold higher enrichment of true cancer cases among those identified as high risk, and that the EHR foundation model further improved predictive performance across 26 cancer types.
What to keep in mind
The abstract does not provide detailed limitations, such as model calibration, subgroup performance, or how the models would work in actual clinical screening programs. It also notes that evidence has been limited so far on the usefulness of these models compared with current screening criteria.
Key points
- EHR-based predictive models found more true cancer cases among people classified as high risk than traditional risk factors alone.
- A baseline approach produced a 3- to 6-fold higher enrichment of true cancer cases.
- The study examined eight major cancers using data from more than 865,000 All of Us participants.
- A foundation model trained on comprehensive patient trajectories improved predictive performance across 26 cancer types.
- The abstract does not describe detailed limitations or clinical implementation results.
Disclosure
- Research title:
- EHR-based models outperformed traditional screening criteria
- Image credit:
- Photo by National Cancer Institute on Unsplash
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.