Oxford University retrospective data study of 2.5M patients

This peer-reviewed publication on a retrospective data cohort was conducted by Oxford University researchers and published in Cancer Medicine. The independent academic study validates Medial EarlySign’s ColonFlag^™ model on a large UK-based adult population, providing validation of this solution’s ability to identify individuals at risk for having colorectal cancer (CRC). The Oxford study also supports other approaches to early CRC detection, including screening and active case finding.

Oxford University’s peer-reviewed publication gives healthcare systems, physicians, CMIOs, Chief Quality Officers, population health managers, and payors valuable insights on valuable ways to identify patients at risk for CRC who were non-compliant with their colonoscopy screenings.

Readers will come away with a better understanding of how machine learning-based healthcare models can analyze ordinary, routine EHR data to identify patients at risk:

Highlights of the Oxofrd include:

• 2.5 million patient records
• Model performance via AUC was similar to a real-world, live implementation study
• The model was sensitive to flag patients with elevated risk for harboring CRC up to 24 months prior to diagnosis
• 8.8% PPV at 99.5% specificity
• It was concluded that the model can support the unscreened population:
  • “At the less extreme cut off values, where sensitivity is higher, the algorithm could also help target those who do not take up the invitations for FOBt or who refuse colonoscopy when offered through the screening program.”

Note:  In the U.S. Medial EarlySign commercializes LGI Flag, which identifies individuals at high risk of having lower GI disorders. The algorithm analyzed in this study, ColonFlag™, bears CE mark and is commercialized for use in the EU, UK, Israel.