Health professionals' and service users' interpretation of screening test results: experimental study

Abstract

Objective: To investigate the accuracy of interpretation of probabilistic screening information by different stakeholder groups and whether presentation as frequencies improves accuracy.

Design: Between participants experimental design; participants responded to screening information embedded in a scenario.

Setting: Regional maternity service and national conferences and training days.

Participants: 43 pregnant women attending their first antenatal appointment in a regional maternity service; 40 companions accompanying the women to their appointments; 42 midwives; 41 obstetricians. Participation rates were 56%, 48%, 89%, and 71% respectively.

Measures: Participants estimated the probability that a positive screening test result meant that a baby actually had Down's syndrome on the basis of all the relevant information, which was presented in a scenario. They were randomly assigned to scenarios that presented the information in percentage (n = 86) or frequency (n = 83) format. They also gave basic demographic information and rated their confidence in their estimate.

Results: Most responses (86%) were incorrect. Obstetricians gave significantly more correct answers (although still only 34% [corrected]) than either midwives (0%) or pregnant women (9%). Overall, the proportion of correct answers was higher for presentation as frequencies (24%) than for presentation as percentages (6%), but further analysis showed that this difference occurred only in responses from obstetricians. Many health professionals were confident in their incorrect responses.

Conclusions: Most stakeholders in pregnancy screening draw incorrect inferences from probabilistic information, and health professionals need to be aware of the difficulties that both they and their patients have with such information. Moreover, they should be aware that different people make different mistakes and that ways of conveying information that help some people will not help others.

Fig 1

Distribution of responses from the four participant groups. X axis is response to scenario expressed as percentage (width interval=0.5%); y axis is number of responses

Fig 2

Distribution of responses to presentation as percentages versus frequencies. X axis is response to scenario expressed as percentage (width interval=0.5%); y axis is number of responses