This survey study found that 34.5% of women who had previous been invited to screening reported being non-regular attenders. The strongest predictors of regular attendance were prioritising cervical screening, and memory. Other predictors were environmental context and resources, intention, and having a household income of £50,000 or over. No other demographic variables were found to significantly predict regular attendance, except that women who were married or in a civil partnership were more likely to be regular attenders than single women, but this was only found in the univariate comparison. Having ever been sexually active, and having had a STI in the past were related to higher rates of reported regular attendance in univariate comparisons, but likewise, these were not significant when included with individual barriers and facilitators.
The finding that prioritising cervical screening was the strongest predictor of reported past attendance is consistent with previous research. A UK survey found that an independent predictor of being overdue for screening was not getting round to it right away [13]. The importance of priority has also been highlighted in qualitative work [21], and another theory-based survey, which found that the extent to which women felt incapable of getting screened due to competing commitments (analagous to ‘cervical screening priority’ in this study) was associated with lower attendance [15]. Similarly, we found that memory was a significant predictor of regular attendance. This is consistent with other research, where forgetting to make an appointment was a main reason for non-attendance in a survey study of Dutch women [22], and not getting round to making an appointment was a commonly reported barrier in a study of Black women in London [23].
The importance of environmental context and resources has been commonly reported in other studies, with factors such as a lack of time or childcare being reported as a barrier to attendance [11, 24, 25], and being associated with lower attendance levels [13, 15]. Struggling to arrange a convenient appointment time was found to be an independent predictor of being overdue for screening [13].
Interestingly, emotional factors were not significant predictors of reported regular attendance. This is consistent with the findings by Waller et al., whereby embarrassment was cited as a barrier, but was not predictive of attendance [13]. However, in contrast, a survey study in Australia found that embarrassment and anxiety were related to past screening [26].
Social norms variables were not independent predictors of reported regular attendance. While univariate comparisons found higher scores for social norms (descriptive, injunctive peers, and from healthcare professionals) in regular attenders, in the regression analysis the descriptive norms variable, and the social norms or expectations from healthcare professionals tended towards being predictors of non-regular attendance (with p < 0.1). Social norms do not appear to be a commonly reported predictor of attendance in the literature. However, Knops-Dullens et al., found attenders had more positive role models, and social support to attend [15]. Yet this study only reported univariate comparisons between attenders and non-attenders, so the findings are comparable to the positive univariate relationship observed in the present study between social norm variables and being a regular attender. It could be that smear tests are something that are not commonly discussed, and therefore the impact of social norms on behaviour is inconsistent, or not as expected.
Household income level was the only demographic factor which was a significant predictor within the regression model with barriers and facilitators. This is consistent with a recent systematic review which found that household income is associated with CS uptake [27]. However, that review also identified that all included studies except one found a positive relationship between education and screening uptake, yet this was not observed in the present study. The finding from univariate comparisons that women who were married/in a civil partnership were more likely to report being regular attenders than single women is consistent with several other studies [7, 8, 13]. These demographic findings can be used to support the targeting of an intervention to those who are least likely to attend (e.g. lower household income, and potentially single women or those who have not had an STI), and the inclusion of relevant subgroups when co-designing any intervention.
In the present study, while the younger age group did have lower absolute levels of regular attendance than older groups, this difference was not significant. Previously, younger age has commonly been associated with lower attendance [8, 24], however, other studies report that those in an older age category (55–64) are more likely to be overdue for screening than a younger category (35–44) [3, 6, 13]. Therefore, there are conflicting findings regarding the relationship between age and attendance. In contract to expectations, we did not observe higher rates of non-regular attendance in ethnic minority groups [13]. This may indicate that our sample, particularly those from ethnic minority backgrounds, may not be fully representative. It is interesting that we did not observe differences in regular attendance in women who had experienced physical, sexual or psychological abuse.
The study findings can help inform interventions which target individual barriers most closely associated with attendance. For example, reminder-based interventions, which target the priority for making an appointment are likely to be effective. In addition, addressing environmental barriers, such as availability of transport and childcare, or facilitating booking appointments through provision of online scheduling or out-of-hours appointments, is likely to support increased screening attendance. The finding in univariate comparisons that those who have ever been sexually active, and have had a STI in the past had higher rates of reported regular attendance may suggest an intervention which emphasises the importance of attendance even if one has not previously had an STI, or many sexual partners, or which is targeted to people who may be in that category. However, it is interesting that these variables were no longer significant when included in a model with individual barriers and motivators. Future research could consider the differences in barriers between higher and lower income groups, to ensure that interventions are likely to reduce screening inequalities.
Study limitations and strengths
This was a large survey of a representative population in London. The income levels (45% of the sample with household income under £35,000) are representative of the London population [28], as are education levels (45% having a vocational degree or lower) [29]. The proportion classed as regular attenders (65.5%) is similar to London cervical screening coverage rates (64.7%) [3]. The survey was based on a comprehensive framework of behavioural predictors [16], and a review of barriers reported in the literature. The study quantitatively shows the predictors of attendance of a large representative sample, and the findings can be used to inform the design of interventions to improve CS uptake.
This study has some limitations. Attendance measurements were self-reported, rather than from objective attendance data. As the survey was conducted online and in English, it is only representative of the online, English-speaking population. This may explain why the percentage of participants of non-White ethnicity (26%) was lower than that found in London from the 2011 census, where 40.2% identified with being from an Asian, Black, Mixed, or Other ethnic group [30]. Therefore, certain segments of the population may not be well represented. In addition, there may be selection bias, in that women who have previously attended cervical screening may have been more likely to participate in a survey about cervical screening than non-attenders. However, the proportion classed as regular attenders (65.5%) is in line with London coverage rates for cervical screening (64.7%) [3]. Another key limitation is that the findings may have limited generalisability outside the London online population, or at least outside a UK urban setting. The use of backwards stepwise regression also has limitations, including a bias towards regression coefficients appearing larger, and p values appearing smaller. However, this approach was selected given that this was an exploratory analysis using a large yet comprehensive set of potential predictors of behaviour.