Using cancer risk algorithms to improve risk estimates and referral decisions

Abstract

Plain language summary Cancer risk algorithms are statistical formulae that calculate the probability that a patient presenting with certain symptoms has cancer. Their aim is to aid primary care physicians when deciding whether a patient should be seen by an oncologist urgently. We presented 157 UK primary care physicians with 20 descriptions of hypothetical patients at varying degrees of risk and asked how likely they would be to refer them to an oncologist for suspected cancer urgently. We then showed them the risk (probability of these patients having cancer), as calculated by an algorithm. The algorithm changed the physicians’ inclination to refer the patients 26% of the time. Decisions improved overall. We propose such algorithms be used to aid cancer referral decisions, and to train doctors making decisions about which patients should be seen by oncologists urgently.