Innovative Algorithms Enhance GP Ability to Detect Undiagnosed Cancer

Recent advances in predictive technology are poised to significantly improve early cancer detection in primary care. Researchers from Queen Mary University of London and the University of Oxford have developed two sophisticated algorithms that utilize electronic health records, demographic data, symptoms, family history, and blood test results to estimate the likelihood of a patient having an undiagnosed cancer. These models surpass current tools like QCancer in sensitivity and accuracy, aiding general practitioners (GPs) in identifying patients who may benefit from further testing.

The algorithms incorporate seven routine blood tests, including measures of blood count and liver function, to bolster their predictive power. They have identified additional medical conditions and symptoms associated with an increased risk of various cancers, including those in the liver, pancreas, and kidneys, as well as extended associations with family history and specific symptoms like itching, bruising, back pain, and dark urine.

By embedding these models into clinical systems, GPs can conduct routine risk assessments more effectively during patient consultations. The tools are designed to be affordable and use data already available in patients’ electronic records, supporting NHS goals for earlier cancer diagnosis and management. Lead author Professor Julia Hippisley-Cox emphasized the potential for these algorithms to identify early-stage cancers more accurately, thereby enabling prompt treatment and improved patient outcomes.

This breakthrough holds promise for revolutionizing cancer detection in primary care settings, with the potential to diagnose cancers at earlier, more treatable stages across a broad age range. The research, published in Nature Communications, highlights the importance of leveraging existing data for enhanced clinical decision-making.