Are self-reports of cancer screening participation accurate?
By Jo Waller, on 26 September 2015
By Siu Hing Lo
Health behaviour research often relies on surveys to collect data of people’s lifestyle and views about health. However, there is concern about the reliability of self-report measures. Common reasons for inaccurate self-report include the desire to give socially desirable answers and issues with recall. When people report their participation in cancer screening their answers are likely to be affected by both. Surveys consistently show that most people agree that screening is a good idea, so it is reasonable to assume they might be tempted to give a socially desirable answer when asked about their own behaviour.
Previous evidence from the United States suggests that social desirability is not the (main) explanation (Vernon et al., 2012). Nevertheless, recall could still be a significant issue. The most common screening tests used in the UK require participation at two- to five-yearly intervals. This means that accurate self-report requires people to recall what they have done over a long period of time. In our latest survey study, we asked respondents permission to check their NHS screening records, so we could compare their self-reported participation in bowel cancer screening with their NHS records.
Unfortunately only around 40% of the total survey sample agreed to this ‘record check’. People who agreed were also more likely to be more affluent and have participated in bowel cancer screening.
On the positive side, we showed that those who consented to the record check could accurately report whether they have ever take part in bowel cancer screening. A large majority also accurately reported whether they had taken part at least twice (81%) and whether they had taken part every time they had been invited (77%).
Interestingly, mismatches between self-report and records were due to both ‘over-reporting’ and ‘under-reporting’ of screening participation. On the one hand, one-fifth of respondents who said they had taken part every time, had in fact failed to respond to at least one invitation. On the other hand, roughly one-sixth reported having taken part once, even though – in reality – they had taken part at least twice.
Although we could only examine the accuracy of self-reported bowel cancer screening among survey respondents who gave permission for the record check, it allowed us to explore what type of biases are likely to result from different survey questions. The biggest obstacle to accurate self-report of bowel cancer screening seemed to be recall of the number of screening tests received and completed. Survey measures which rely less on recall of each screening episode are therefore more likely to yield reliable data.
References