Intuitive Understanding of Base Rates: Physician and lay Estimates of Predictive Values in Diagnostic versus Screening Mammography

Purpose: This study examines whether physicians and other adults intuitively understand that the probability a positive test result is a true positive (positive predictive value, PPV) depends on the base rate of disease in the population tested. In particular, this research seeks to examine perceptions of positive test results and PPV when mammography is used in a symptomatic population versus an asymptomatic mass screening population. 
Methods: 59 attendees at a general practitioner (GP) conference in Scandinavia and 45 multi-national MBA students enrolled in a Danish University were surveyed. Respondents read two scenarios, each about a 52-year-old woman. The women differed only in that one had a mammogram following a breast cancer symptom while the other took part in mass screening and was asymptomatic. Both received a positive test result. Surveys elicited estimates of the probability of breast cancer prior to mammography and estimates of the probability of breast cancer given the positive result (the PPV) for each woman. Open-ended questions followed.
Results: GPs and MBAs overestimated the base rates of cancer. Among women with symptoms, breast cancer incidence (base rate) is about .15; on average GPs estimated .36 and MBAs .42. In mass screening populations breast cancer incidence is estimated as < .005; on average GPs estimated .08 and MBAs .21. Both groups underestimated the diagnostic PPV and overestimated the screening PPV, meaning they underestimated how differently cancer is predicted by a screening versus a diagnostic test. Actual predictive values differ by about 65 points, with PPV in screening at about .25 versus over .90 in diagnostic testing. On average, GPs estimated an 18 point difference, and MBAs a 10 point difference. Fully 40% of MBAs and 10% of GPs estimated the two as equal. Written responses confirm that MBAs who estimated the two PPVs as equal believe that a positive mammogram is a positive mammogram, and interpretation is not affected by prior beliefs of how likely cancer is.
Conclusions: These results suggest fundamental misunderstandings, especially among lay decision makers, about how to interpret positive medical tests. In particular, they misunderstand that a positive mammogram in mass screening is substantially less predictive of cancer than a positive diagnostic mammogram. The fact that many people do not understand this fundamental difference in screening versus diagnostic testing may have implications for how people, including medical doctors, understand and contribute to debates about the benefits and harms of mass screening programs.