HIV screening: Helping clinicians make sense of test results to patients. Natural frequencies foster insight and should become part of the training of every medical students and HIV counsellor (2013)
Authors
Abstract
Does innumeracy among clinicians matter? No systematic studies of effects on patients exist—just anecdotal reports of people with false positive test results engaging in unprotected sex with other HIV positive people, believing that it would not matter any more, and of people who committed suicide or who endured harmful effects of unnecessary antiretroviral treatment.3 A US woman, newly married and pregnant, was told by her doctor to undergo HIV screening and tested positive on western blotting. The doctors told her that the false positive rate was five in 100 000, gave her handouts from the internet about living with HIV, and sent her off to tell her husband and family the news. After a bad evening, she considered her low risk lifestyle and went with her husband to a different clinic for a pinprick test; both partners have tested negative ever since.4 How can we help clinicians understand the risk of false positives? Consider a low prevalence group in which the frequency of (undiagnosed) HIV infection is about one in 10 000, as in female US blood donors.5 If the test (such as enzyme immunoassay together with western blotting) has a sensitivity of 99.95% and a specificity of 99.99%, what is the positive predictive value or P(HIV|pos)? To calculate this, medical students are taught to insert the prevalence, the sensitivity, and the false positive rate into Bayes’s rule: P(HIV|pos)=P(HIV)×P(pos|HIV)/[P(HIV)×P(pos|HIV) + P(no HIV)×P(pos|no HIV)]. In our case this gives P(HIV|pos)=0.0001×0.9995/[0.0001×0.9995+0.9999×0.0001] or about 50%.
Bibliographic entry
Gigerenzer, G. (2013). HIV screening: Helping clinicians make sense of test results to patients. Natural frequencies foster insight and should become part of the training of every medical students and HIV counsellor. BMJ, 347:f5151. doi:10.1136/bmj.f5151 (Full text)