Article Text
Abstract
Background Despite the decline in the incidence of cervical cancer in Australia as a result of population screening, a substantial proportion of women in NSW screen less regularly than the recommended two-yearly interval or do not screen. With higher rates of cervical cancer in unscreened and underscreened women, and despite the introduction of the human papillomavirus vaccine, there remains a need to continue to remind women to screen. The mass media has been shown to be effective at improving participation in cervical screening. A 2007 television advertising campaign to promote cervical screening in New South Wales (NSW) was examined.
Methods Data from the NSW Papanicolaou (Pap) Test Register were used to compare weekly numbers of Pap tests for NSW overall and in metropolitan local government areas with low screening rates by age group and by time since the last Pap test. Time series regression analysis incorporating seasonal effects was used to estimate the strength of the association between screening and the media campaign.
Results Overall during the advertising campaign, 15% more screens (16 700) occurred than expected for 2007 without the advertising campaign. Increases were evident among unscreened and underscreened women, with little overscreening occurring. Women living in low screening areas also showed a significant increase in mean weekly screens of 21% (388) over that expected in the absence of the media campaign.
Conclusions Despite the ecological nature of this study, the mass media campaign appears to have been successful in increasing screening in unscreened and underscreened women in NSW.
- Pap test
- cervical
- mass media
- social marketing
- cancer screening
- cancer: cervix
- health education SA
- health related behav
- prevention PR
Statistics from Altmetric.com
- Pap test
- cervical
- mass media
- social marketing
- cancer screening
- cancer: cervix
- health education SA
- health related behav
- prevention PR
Introduction
Cervical cancer is largely preventable if precancerous cell changes are detected early enough and treated appropriately. Since the introduction of organised cervical screening in Australia by means of the Papanicolaou (Pap) test in 1991, the incidence of cervical cancer in New South Wales (NSW), the most populous Australian state, has declined at an accelerated rate.1 2 The recommended cervical screening interval in Australia is 2 years for well women who have ever had sex, and the target age group is 20–69 years. However, getting women to screen every 2 years remains a challenge. In NSW, the biennial cervical screening rate for women in the target age group in 2004–5 was 57.1%, indicating that a substantial proportion of women are screening outside the recommended 2-yearly interval (underscreened) or are not screening at all (unscreened).3
Recent studies conducted in NSW and Sweden report that approximately two-thirds of cases of cervical cancer occur in women who are unscreened or underscreened.4 5 Higher cervical cancer incidence and mortality rates occur among women of lower socioeconomic status,6 migrant women from some country-of-birth groups7 8 and indigenous women.9 10 Women in these groups are also more likely to be unscreened or underscreened.3 10–15
The cervical cancer or human papillomavirus (HPV) vaccine was introduced in Australia in April 2007. It is available free of charge under the National HPV Vaccination Program to girls aged 12–18 years in the school-based program, and to women aged 18–26 years from GPs and community-based immunisation providers (until July 2009).16 There is an underlying concern that the introduction of the program may have a negative impact on already declining cervical screening rates.
The NSW Cancer Plan 2007–10 places an emphasis on raising awareness of and increasing participation in cervical screening, particularly among women who are unscreened or underscreened.17 The design and implementation of a mass media campaign forms part of the plan to increase cervical screening rates in NSW.
Mass media campaigns have been shown to be effective in improving screening participation. A 2004 review of seven studies using mass media to promote either cervical screening or mammography showed all but one to impact positively on screening behaviour.18 19 Additionally, these reviews found multicomponent interventions to be effective in increasing initial and repeat screening across a wide variety of settings and target populations, but the contribution of individual components is less clear. A recent study on the effectiveness of a targeted television campaign in Victoria (Australia) reported an 18% increase in the number of Pap tests observed during the campaign period compared with the equivalent period in previous years.20 Moreover, the highest increases were observed among women due or overdue for a Pap test, and screening numbers were observed to decline to background levels approximately 3 weeks after the campaign.
An adapted version of the Victorian Don't Just Sit There advertisement was subsequently implemented in NSW as part of the Cervical Screening Campaign. The purpose of the present paper is to examine the cervical screening outcomes of the NSW mass media campaign implemented in 2007. The paper also aims to report on the effectiveness of this strategy in reaching unscreened and underscreened women for NSW overall and for women in metropolitan areas of NSW with historically low cervical screening rates.
Methods
Mass media campaign
In 2007 the Cancer Institute NSW commissioned a state-wide mass media campaign using an adaptation of the successful campaign advertisement Don't Just Sit There developed by the Cancer Council Victoria incorporating PapScreen Victoria. The advertisement featured a series of women's legs and torsos in a variety of situations and was designed to address discomfort and the benefits of early detection. The campaign primarily targeted women aged ≥30 years who were underscreening. The secondary target audience was unscreened women.
The campaign advertisement aired on free-to-air television over 9 weeks from 1 May to 30 June 2007. The communication strategy also included radio, press (with translations into Arabic, Chinese (Mandarin and Cantonese) and Vietnamese), plus convenience advertising (placement in women's toilets) and posters (placement in waiting rooms of general practitioners, specialists, women's health clinics, etc). Further details regarding the development and implementation of the campaign can be found at http://www.cancerinstitute.org.au.21 However, this paper focuses on the television campaign only.
Television audience rating points (TARPs)
The exposure measure was weekly Target Audience Rating Points (TARPs) for women aged 25–54 years. TARPs are a standard composite measure of television advertising weight and are a multiple of the percentage of the target audience reached by the campaign and the average number of times they are exposed to the advertising. For example, one TARP represents 1% of the target audience who have had the opportunity to see a campaign advertisement on television once. One hundred and twenty TARPS can mean that 40% of the target audience had the opportunity to see the advertisement on television three times in a given period, or 30% had the opportunity to see it four times, and so on. Over the campaign period the target group was exposed to medium to high TARP levels (TARPS ranged from 136 to 190 per week).21
NSW Pap Test Register (PTR) data
The NSW Pap Test Register is a secure and confidential database of women's Pap test and related follow-up test results. The Register was established by the Public Health Act 1991 and commenced operation in July 1996. The Act stipulates that pathology laboratories inform the NSW Pap Test Register of the results of all cervical cancer tests, Pap tests and cervical histology for NSW women. Demographic data for all these women except those who choose not to participate are forwarded to the Register. The term ‘opt-off’ applies to women who withhold personal details from the NSW PTR which prevents allocation of address or recording of other demographic details including age. During the 24-month reporting period to 31 December 2007, 0.84% of Pap tests performed in NSW involved women who opted off the NSW PTR.22
De-identified unit screening records from the PTR were used to produce numbers of Pap tests conducted weekly for 2006–7 to allow comparison between the campaign period in 2007 and the same time in 2006, and between the campaign period and pre- and post-campaign. Age, local government area (LGA) and screening interval (defined as time since last Pap test) were also used to assess campaign reach. Only Pap tests of the cervix were included in the analysis. All vaginal vault smears were excluded.
Weekly Pap screens as recorded on the NSW PTR were plotted for 1 January to 31 December 2006 and 2007, and numbers of screens during the campaign period (1 May to 30 June 2007) were compared with the same period in 2006. Screening interval groups were defined as women whose last screen was: (1) 0–21 months before their screen occurring during the comparison periods January–December 2006 and January–December 2007 (potential overscreeners); (2) 21–27 months before (women due for a screen); (3) 27–36 months (overdue women); (4) 36–48 months before; (5) 48+ months before; and (6) women with no previous cervical screen recorded on the PTR. Pap screening in areas of NSW with historically low Pap screening rates and large populations, primarily characterised by indices of low socioeconomic status and/or high non-English speaking population compositions, were also examined in relation to the mass media campaign. These high priority areas comprised the metropolitan LGAs of Fairfield, Liverpool, Bankstown, Blacktown and Penrith (concentrated in south-western and western Sydney) and Wollongong. For the present paper, screening in these high priority areas was analysed in aggregate by the screening interval groups defined above.
Analysis of data
Differences in mean weekly screens occurring during the advertising period of 2007 compared with the corresponding period in 2006 were tested by a t-test. These differences were similarly tested after adjusting for underlying changes in screening between 2006 and 2007 outside the advertising period to produce estimates of screening expected in 2007 during the comparison period assuming no advertising. This was by pro-rating the weekly screens during May/June 2006 by the mean screening increase in 2007 over 2006 outside May and June.
Time series regression analysis incorporating seasonal effects using Seasonal Autoregressive Integrated Moving Average (SARIMA) was then used to test the strength of the association and statistical significance of weekly cervical screens against the mass media campaign. This analysis was conducted for screening overall and separately for each screening interval group, age group and for the screening high priority areas.
In the time series modelling, the mass media campaign was modelled using TARPs as the exposure variable of interest. The seasonality of the screening series was treated as week by week, with 52 weeks taken in each year as the differencing factor—that is, the series of weekly screening number differences between the nth week of 2007 and the nth week of 2006 was analysed as the outcome variable of interest. The time series of these differences were first ‘pre-whitened’, where the ‘autoregressive’ and ‘moving average’ effects are factored out of the series to produce residuals not significantly different from white noise. The effect of the campaign was then assessed with regard to this ‘pre-whitened’ residual series, standard practice in time series regression analysis. Modelled regression estimates of TARPs against the pre-whitened screening series were then used to estimate the extra mean Pap tests associated with the television advertising against those expected for 2007 from the ARIMA modelling, based on a total exposure for the May/June advertising period of 1344 TARPs for NSW overall and 963 TARPs for the high priority areas. All statistical analyses were conducted using SAS Version 9.1, PROC ARIMA (SAS Institute Inc, Cary, North Carolina, USA).
Results
During the period of the campaign of May–June 2007, the number of Pap tests exceeded those for the same 2006 period by approximately 1500–3500 per week, which represented an unadjusted mean rise of 22% (figure 1). The mean weekly number of screens following the campaign also appears to have remained higher than in 2006, whereas before the campaign overall screening numbers in 2006 and 2007 were broadly similar. When the increase in screening was examined according to age groups 20–29 years, 30–49 years and 50–69 years, the increase during the campaign was reasonably uniform but a sustained increase following the campaign was observed in women aged 20–29 years (figure 2).
The increase in screening during the campaign also occurred in the appropriate screening groups. There was no significant advertising-related increase in screening in women not due for a screen (figure 3, 0–21 months), but significant increases occurred in women due or overdue for a screen or those in whom no previous Pap test was recorded on the PTR (figure 3, remaining graphs). However, as is also evident from the plots and particularly in the 48+ month and unscreened groups, substantially higher screening was occurring in 2007 compared with 2006 in the weeks outside the campaign period, and there was also a sustained increase in screening after the campaign. A closer inspection of these trends showed that the increased screening following the campaign was more pronounced with increasing age in women whose last screen was ≥48 months previously, and more pronounced and longer lasting in younger women (20–29 years) who had not screened previously (results not shown).
There was a statistically significant association between the campaign and the increases in Pap screens in women due for a screen and in underscreened and unscreened women, but not in women whose screening interval was 0–21 months, although after adjustment it was significant (table 1).
Time series regression confirmed the association. No significant secular trends in screening were found, so the differencing was confined to the corresponding weeks of each year. After adjusting for autoregressive and/or moving average processes in the time series of the (differenced) weekly screens, the campaign was found to be associated with a mean weekly increase in total screens of 1852 or 15% over those expected to occur for 2007 in the same period if no campaign had occurred (table 2). With time series adjustment, the increase in screening in the 0–21 months screening group was found to be also smaller (6% vs 8%) but non-significant.
In the high priority NSW local government areas, the campaign was associated with a statistically significant excess in mean weekly Pap screens of 388 (21%) over what would be expected for the same period in 2007 if no campaign had been conducted. Significant increases for women due and overdue for their screen (18–29%) as well as in women not previously recorded with a Pap test on the PTR (18%) were also found. However, there was a significant increase in screens in women whose last screen occurred within the previous 21 months (around 8%).
Discussion
This study has shown that a mass media campaign to encourage women to have a Pap test can be effective in achieving increased cervical screening, both in women due for a screen and in underscreened and unscreened women. The Don't Just Sit There television advertising campaign was associated with a 15% increase in cervical screening in women in NSW during the 9 weeks of its airing compared with what would have been expected for the same period without the advertising campaign. Furthermore, the potential for significant overscreening in women not due for a screen (the 0–21 months screening interval group in this analysis) did not eventuate for NSW overall, although in the high priority areas there was a significant but not substantially higher screening increase in these women (8%) over that expected. Otherwise, the campaign was associated with significantly and substantially higher Pap testing in women from these high priority areas with historically low screening rates.
A notable feature of the time plots of cervical screens is the week-by-week regularity of screening patterns which are quite sensitive, particularly to holiday periods. For instance, the Easter and ANZAC Day (an Australian and New Zealand day of war remembrance) holiday breaks occurred just before the campaign in 2007 and in the corresponding period in 2006, during which screening declined markedly and then subsequently picked up. The campaign started just at the time of the pick up in screening following this holiday period and aimed to maximise this expected seasonal increase. Also occurring within the advertising period was the Queen's Birthday Holiday weekend, as evidenced by the dip in week 24. Clearly, without taking the week-by-week seasonality in screening into account, by comparing 2006 and 2007 screening on a weekly basis, it is possible that incorrect inferences regarding the effectiveness of the campaign would result. In particular, analysis of weekly screens allows more precise location of the beginning and end points of the campaign. The ARIMA time series analytical approach is appropriate here also because the outcome measure is of screens (not screened women) and, as the number of screens comprises a subset of women who screened more than once during the study period, there would be autocorrelation inherent in the time series of screening events. The other issue addressed by an ARIMA modelling approach in a natural way is underlying population trends or other influences fuelling secular trends in screening events, which are eliminated by analysing time series de-trended by appropriate differencing.
While the effects of the campaign on screening are clearly tangible, it is notable that there was an apparent sustained effect in the 48+ months screening group, and screening activity in women previously not recorded on the PTR was also substantially higher over most of 2007 compared with 2006. The trend in the former group can be explained readily by a 48-month reminder letter initiative conducted by the NSW PTR which was occurring in the background throughout 2007. This initiative was implemented to encourage women who had not screened in ≥4 years to have a Pap test and was staged so that the first reminder letters were sent to women with the longest lapsed period since their last screen. Accordingly, longer screening intervals are expected to occur more commonly in older women, and this is consistent with the positive age gradient found in sustained screening following the campaign.
The difference in the unscreened group between 2007 and 2006 is less easy to explain and it might be tempting to relate this to increased screening in younger unscreened women wishing to avail themselves of the HPV vaccine which was introduced in April 2007. The more pronounced effect in younger women certainly suggests this, reinforced by the fact that higher screening in younger than in older women was more sustained following the campaign across the screening groups (figure 2).
Some explanation for the immediate screening response to the advertising is indicated by a follow-up survey (n=1000) conducted to evaluate the NSW Cervical Screening campaign.21 After the campaign, 55% of respondents who recalled viewing the advertisement stated that they would not do anything in response, but a strong and highly significant predictor of this particular response was recently having a Pap test. This was consistent with screening rates in the 0–21 months screening group observed during both the NSW and Victorian Don't Just Sit There campaigns, where the number of screens overall was not significantly different in women whose last screen was ≤21 months previously. Moreover, 16% of survey respondents also indicated that, as a result of seeing the advertisement, they planned to have a Pap test soon, very similar to the follow-up evaluation survey of the Victorian campaign.20 If we assume this figure to be an estimate of the increase in Pap testing attributable to the advertisement, then the figure of 15% for NSW overall in the present study is remarkably close to this and close to the screening increment found in Victoria of 18%.
Themes emerging from focus groups of unscreened and underscreened women in particular indicated that the intended message of the campaign got through21: that the temporary discomfort of a Pap test is a small sacrifice compared with having cervical cancer and it's only two-yearly: ‘… it's saying that it might be uncomfortable to do this, but the comfort is in knowing that you're okay’; and ‘… it doesn't take a long time to do it. You don't do it every month …’. Moreover, the acknowledgement of the discomfort aspect was perceived as lessening fear and anxiety: ‘I liked the fact that they showed discomfort because that's exactly how I feel before I go and I sit there and do that …’; and ‘Just go and have a Pap test … don't be scared, it's over in a few minutes’. There was also some evidence that women reported being more likely to act if they had learnt something new from the messages being conveyed. In short, it is evident that frank honesty and empathy, combined with new information and a stark choice, was personally relevant and therefore an effective prompt for many women to act.
While the present study was ecological in design, it is not only the supporting survey evidence that suggests the advertising campaign was effective. When we examine the study results against the criteria for causation, as outlined originally by Austin Bradford Hill,23 we find that Hill's criteria for causation are satisfied:
Temporality: our findings clearly show the purported outcome did not precede the purported cause, as shown in Figures 1–3.
Strength: the statistical tests showed a strong statistically significant temporal correlation between the advertising campaign and cervical screening, after accounting for seasonal effects and autoregressive and moving average processes inherent to time series data.
Dose response: this is not a necessary condition for causality since threshold effects can also occur, but the time series regression modelling showed a statistically significant and positive relationship between advertising exposure and screening indicating a dose-response relationship within the range of advertising intensities modelled.
Consistency: our results are consistent with a similar intervention conducted previously in the State of Victoria.
Plausibility: the relationship is plausible in that many women were exposed to advertising that urged them to have a Pap test as the best way of avoiding cervical cancer, and approximately 2200 more women per week than normally expected had a Pap test during the advertising period.
Consideration of alternative explanations: the only plausible alternative explanation for the increases in screening coinciding with the advertising campaign is seasonal effects. However, the time series modelling approach accounts for seasonality as a factor and still the association between screening and the advertising campaign remained.
An experiment would alter the findings: we would expect that, if individuals or areas were randomised to the advertising campaign and there was no contamination between them, the exposed individuals would be more likely to screen than unexposed individuals and exposed areas would experience a rise in screening while unexposed areas would experience no significant comparable rise (seasonality-adjusted). This is similar to the consistency criterion above and, if one considers the Victoria and NSW advertising campaigns as a natural experiment, there was no increase in cervical screening in NSW when the Victorian advertising campaign was conducted in 2006, and similarly in Victoria for the NSW campaign of 2007.
Specificity: while not a necessary condition for causation since a cause of something can co-exist with other causes, we would contend that the advertising campaign was specific because no plausible alternative explanation for the observed trends has been raised.
Coherence: no logical inconsistencies have arisen as a result of the study.
Despite the present study being an ecological time series analysis, from survey results and from the Bradford Hill criteria above it does provide strong evidence for the effectiveness of the advertising campaign in increasing cervical screening rates.
Future mass media campaigns to further increase cervical screening are planned for NSW. It would be of interest to examine in 2009—and preferably in the absence of a campaign—the extent to which the increased screening of the 2007 campaign has been sustained as the additional screeners return for their biennial Pap screen, and whether younger women in particular have remained on the screening pathway 2 years hence.
Conclusions
Mass media campaigns can change health behaviours and the NSW Cervical Screening Campaign has shown strong evidence for this. It is also apparent that a motivational advertisement that is personally relevant to women is as important as its quantity (audience reach, coverage) if health behaviours are to be influenced for the good. The addition of strategies targeting both doctors and women to coincide with mass media campaigns would be expected to increase the effectiveness of such strategies.18 19 Multiple communication channels to remind women to screen on a 2-yearly basis are required to maintain adequate screening coverage, and a mass media campaign is a key element of this strategy.
What is already known on this subject
Approximately two-thirds of cases of cervical cancer occur in women who are unscreened or underscreened. Higher cervical cancer incidence and mortality rates occur among women of lower socioeconomic status, in migrant women from some country-of-birth groups and indigenous women. Women in these groups are also more likely to be unscreened or underscreened. Mass media campaigns have been shown to impact on screening behaviour and to be an effective strategy to increase participation in cervical screening.
What this study adds
The findings of this study demonstrate the effectiveness of a well-targeted television advertising campaign to increase cervical screening participation. Moreover, the campaign was associated with higher screening participation in unscreened and underscreened women overall, as well as in historically low-screening communities with high proportions of women from culturally and linguistically diverse backgrounds and/or low socioeconomic status. There was no evidence of overscreening associated with the campaign.
Acknowledgments
Cervical screening data were supplied by the NSW Pap Test Register (PTR). The authors acknowledge the contributions of staff members of the PTR and Elizabeth Tracey (Senior Epidemiologist, Cancer Institute NSW) for determining the high priority local government areas.
References
Footnotes
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.