Equity of colorectal cancer screening: cross-sectional analysis of National Bowel Cancer Screening Program data for South Australia
Paul R. Ward, Professor of Public Health, Discipline of Public Health, Flinders University, South Australia; e-mail: email@example.com
Objective: The National Bowel Cancer Screening Program (NBCSP) is a population-based screening program based on a mailed screening invitation and immunochemical faecal occult blood test. Initial published evidence from the NBCSP concurs with international evidence on similar colorectal cancer screening programs about the unequal participation by different population sub-groups. The aim of the paper is to present an analysis of the equity of the NBCSP for South Australia, using the concept of horizontal equity, in order to identify geographical areas and population groups which may benefit from targeted approaches to increase participation rates in colorectal cancer screening.
Method: De-identified data from the NBCSP (February 2007 to July 2008) were provided by Medicare Australia. Univariate and multivariate statistical analyses were undertaken in order to identify the predictors of participation rates in the NBCSP.
Results: The overall participation rate was 46.1%, although this was statistically significantly different (p<0.001) by gender (42.6% for males and 49.5% for females), socioeconomic status (40% in most deprived quintile through to 48.1% in most affluent quintile) and remoteness (45.6% for metropolitan, 46% for remote and 48.6% for rural areas). These findings were confirmed in multivariate analyses. Of the NBCSP participants, 0.24% (CI 95% 0.20–0.30) identified themselves as Indigenous and 8% (CI 95% 7.7–8.3) reported speaking a language other than English at home.
Conclusion: Findings from this study suggest inequities in participation in the NBCSP on the basis of gender, geographical location, Indigenous status and language spoken at home.
In Australia, colorectal cancer (CRC) is the second most commonly diagnosed cancer and cause of death from malignant disease,1 which is similar to recent global data from 20 world areas, as defined by the United Nations Population Division, on both incidence and mortality rates.2 Coupled with the prediction that CRC incidence will rise by 30% between 2002 and 2011,3 the early detection of CRC is a major clinical and public health concern.
In response to the increased burden of disease posed by CRC, a number of randomised controlled trials have been undertaken that demonstrate the effectiveness of CRC screening for reducing CRC population mortality4,5 and reducing CRC incidence when a test sufficiently sensitive to aid detection of colorectal adenomas is employed.6 However, the impact on mortality and incidence in these trials is limited not just by the technology and the fact that simple screening tests such as guaiac-based faecal occult blood tests (FOBT) have a limited sensitivity for the target lesion, but also by the willingness of the eligible population to participate in screening in the first place, or the ability to access high quality diagnostic assessment and treatment if test results indicate neoplasia. Equity of access to CRC screening, education and a range of social and economic factors might well influence these non-technological limitations, including factors such as the importance of primary health care practitioners in detecting symptomatic patients.7,8
The aim of this paper is to analyse the equity of the Australian CRC population based screening program, the National Bowel Cancer Screening Program (NBCSP), which was implemented (Phase 1) in South Australia in February 2007, and to identify particular population groups and geographical areas that may benefit from targeted programs to increase participation rates in the NBCSP. This paper focuses on the first phase of the screening pathway (i.e. FOBT), although it is acknowledged that access to all stages of the screening process including diagnostic and treatment services is required for benefit. Our analysis is based on a de-identified extract of the NBCSP Register for the South Australian population invited to participate in the screening program. Early findings from the NBCSP9–11 and international CRC screening programs12 found that even in situations where universal access to the screening program was in place, there were differences in participation rates between population sub-groups.7 We argue that the difference in participation rates between groups should more specifically be seen as inequity in access, because they are based on factors other than clinical need. This is not just a semantic distinction, but represents a fundamental shift in thinking, since inequity is synonymous with social injustice and unfairness.
In terms of defining the ‘need’ for CRC screening, research suggests that CRC incidence increases with age13 and is higher for males than females.13,14 In terms of socioeconomic status, findings are unclear, with some research suggesting an increased CRC incidence in lower socioeconomic (SES) groups15 while other research found increased CRC incidence in males in higher SES groups, but no difference between SES groups for women.16 In terms of Indigenous status, it has been suggested that CRC incidence is lower in Indigenous groups than non-Indigenous groups, although there are potential limitations due to poor data quality.17,18 Therefore, it may be argued that the need for CRC screening is greater for men and in terms of increasing age, although we remain uncertain with respect to SES and Indigenous status. Based on these data, an equitable screening program should have higher participation rates for groups with higher needs (i.e. men and those aged 65 years) and comparable participation rates for groups with similar needs (i.e. SES groups and Indigenous/non-Indigenous groups). This is the principle of horizontal equity,19 which we use to interpret the findings from our analyses. Two main forms of health care equity are identified: vertical equity (preferential treatment for individuals or groups with greater health needs), and horizontal equity (equal treatment for those with similar levels of health care need).20,21 Therefore, within this paper, any systematic difference in participation rates within the NBCSP, which are not justified on the basis of need, may be horizontal inequities in healthcare. Obviously, participation in the NBCSP is not the only important factor for analysis of equity – it would also be extremely important to analyse equity in terms of the stage of disease at diagnosis, but we did not have access to these data within this study.
Context for the study
The NBCSP currently offers free CRC screening in the form of an immunochemical FOBT test to people turning 50, 55 and 65 years of age in any given year. However, at the time of our study, invitations were restricted to people of 55 and 65 years only. The NBCSP Register mails a pre-invitation letter alerting eligible participants to the arrival of an immunochemical FOBT test kit, before sending the invitation package which includes the test kit. Invitees are requested to mail their FOBT samples to a central pathology service for analysis. The Australian program is supported by a tracking system that sends out reminder letters. Participants who return a positive result are advised by mail to visit their nominated General Practitioner (GP), who is also advised of their result, for the purpose of arranging appropriate clinical follow-up.
Methods and data sources
De-identified data for the South Australian population invited to participate in Phase 1 of the NBCSP (between February 2007 and July 2008) were provided by Medicare Australia. The dataset included the age, sex and postcode for those people sent the FOBT (i.e. invitees) and age, sex, postcode, Indigenous status and language spoken at home for those who completed the FOBT (i.e. participants). Ethics committee approval was granted by the Departmental Ethics Committee of the Commonwealth Department for Health and Ageing and by the Social and Behavioural Research Ethics Committee of Flinders University.
The postcode variable was converted into two new separate variables for use in the analysis. Firstly, each postcode was coded according to the Index of Relative Social Disadvantage (IRSD),22 a composite measure based on selected census variables such as income, educational attainment and employment status. The IRSD scores for each postcode were then grouped into quintiles for analysis, where the highest quintile comprises the 20% of postcodes with the highest IRSD scores (most advantaged areas). Secondly, postcodes were converted into a measure of ‘remoteness’, using the Accessibility/Remoteness Index of Australia (ARIA).23 This is an index of the proximity of postcodes to service centres, or conversely of remoteness of postcodes. The ARIA has both a 5-point and a 3-point scale. We chose to use the 3-point scale which includes the categories of metropolitan, rural, and remote postcode areas.
Data were analysed using the Statistical Package for the Social Sciences (SPSS) version 15.0. In total there were 92,279 invitees during Phase 1 of NBCSP (January 2007 to July 2008 in South Australia), including the 17,497 who had been involved in the pilot phases of the NBCSP. The pilot invitees were removed from the data analysis because their prior exposure to CRC screening may have had a confounding effect on NBCSP participation. Therefore, our final dataset for analysis included 74,782 South Australians who had been invited to undertake CRC screening for the first time by the NBCSP. It was not possible to ascertain if these people had previously been offered, or participated in, CRC screening.
Initial analyses were undertaken using χ2 testing and univariate odds ratios in order to analyse the associations between participation in the NBCSP and socio-demographic variables (sex, Indigenous status, language spoken at home, IRSD and ARIA). All variables that were associated with NBCSP participation rate at the p<0.25 level24 at a univariate level were then entered as independent variables into a logistic regression analysis with participation rate in the NBCSP as the dependent variable (Indigenous status and language spoken at home were not included due to data only being available for participants). The final multiple regression model was checked for collinearity and all variables included at the initial step remained statistically significant.
Of the 74,782 invitees, 34,480 participated in the NBCSP, which represents an overall participation rate of 46.1%. Participation rates varied by place of residence, age, sex, IRSD, ARIA score, Indigenous status and language spoken at home.
Analysis of the South Australia data shows that the metropolitan region is socioeconomically differentiated in terms of the southern and eastern suburbs being traditionally more affluent, with the western and northern suburbs being more materially deprived. There were generally higher participation rates in the south and east, and lower participation rates in the centre, west and outer-north of the Adelaide metropolitan area.
Table 1 shows the participation rates for the IRSD quintiles and geographical accessibility (as measured by ARIA). Participation rates were statistically significantly different (p<0.001) by IRSD quintiles and IRIA classifications, with results revealing a gradient in participation linking increasing affluence with increasing participation, although after the middle quintile, participation rates began to plateau. Participation rates are similar in metropolitan and remote areas (45.6% and 46.0% respectively), although rates are slightly higher in rural areas (48.6%).
Table 1. NBCSP participation rates by IRSD quintiles and ARIA classification.
While the overall participation rate was 46.1%, men had a significantly lower participation rate (p<0.0001) than women (42.6% and 49.5% respectively). In addition, people aged 65 years had a statistically significantly higher participation rate (p<0.0001) than those aged 55 years (51.2% and 42.8% respectively).
Self-reported Aboriginal and Torres Strait Islander status (in this paper we call this group Indigenous, although we do recognise the potential problems in doing this) was available only for participants who completed the FOBT, with the Indigenous status of invitees who failed to return their FOBT unknown. The total proportion of the Indigenous participants was 0.24% (CI 95% 0.20–0.30%). Given that the reported proportion of the Indigenous people of the same age group within South Australia in the 2006 Census was 0.54%,25 and as this proportion is not within the confidence interval of the sample, this suggests that the proportion of Indigenous people who participated in the NBCSP was statistically significantly lower than we would expect. However, given that we do not have sufficient details about the Indigenous status of invitees, we cannot compute an overall participation rate.
The dataset included details of the language spoken at home for NBCSP participants (but not for invitees), which is particularly important for a postal screening program. Given the large number of languages spoken, we created a dichotomous variable for comparison purposes: spoke English only at home versus spoke a language other than English at home. Of the people who participated in the NBCSP, 8.0% (CI 95% 7.7–8.3%) reported speaking a language other than English at home, compared to 17% for the same age group in South Australia in the 2006 Census,25 and as this proportion is not within the confidence interval of the sample, this indicates that the proportion of NBCSP participants who spoke a language other than English at home was statistically significantly lower than we would expect.
The result from the South Australia data is consistent with the national results on participation, which shows people who preferred to correspond in English were 1.6 times more likely to accept the invitation to screen than those who indicated that their preferred language is other than English.26 This suggests an inequity on the basis of language spoken at home.
We computed a multivariate logistic regression model using the variables that were available for both NBCSP invitees and participants. The resulting model is shown in Table 2. Consistent with the univariate results, participation in the NBCSP was higher for females and in higher socioeconomic status areas. Participation rates were also higher in rural than metropolitan areas, but there was no difference between remote and metropolitan areas.
Table 2. Multivariate logistic regression of participation rates in NBCSP.
| ||Female||1.33 (1.29–1.37)||<0.001|
| ||Low||1.25 (1.19–1.31)||<0.001|
| ||Middle||1.38 (1.27–1.39)||<0.001|
| ||High||1.42 (1.34–1.48)||<0.001|
| ||Highest||1.44 (1.39–1.53)||<0.001|
| ||Rural||1.19 (1.15–1.24)||<0.001|
| ||Remote||1.03 (0.94–1.12)||0.559|
Overall, our analyses reveal lower NBCSP participation rates for men compared to women, socioeconomically disadvantaged groups compared to more affluent groups, and people from metropolitan and remote areas compared to those from rural areas in South Australia between 2007 and 2008. Among invitees who participated in the NBCSP, comparison with the 2006 Census data indicated that South Australians who reported speaking a language other than English at home and those who reported an Indigenous background were under-represented. These differences in participation rates, while potentially explainable in part by other factors, highlight the high likelihood of horizontal inequity for CRC screening in South Australia. Analysis of the NBCSP at national level also revealed that higher proportion of females and those aged 65 years have participation than males and those aged 55 years in all states.26 This report also shows that participation was significantly lower in Aboriginal and Torres Strait Islanders and those who preferred to correspond in a language other than English. These findings are also consistent with results from other cancer screening programs which suggest that inequitable patterns of participation may arise from a variety of factors including those associated with gender,27 ethnicity,28 socioeconomic status29 and Indigenous status.19,30 The existing disparities suggest that even in countries such as Australia where the population screening programs are highly structured and universal access to the program is in place, individuals and some population sub-groups may face a range of barriers hindering the actual utilisation of those services. Nevertheless, we cannot discount the possibility that some people may have made conscious and rational decisions not to participate in the NBCSP. For example, a qualitative evaluation of the NBCSP pilots31 found that some non-participants in the pilot made decisions not to participate for a variety of reasons such as fear of cancer, lack of knowledge of bowel cancer and having what they perceived to be more pressing concerns to deal with at that moment in time. Therefore, any policy response needs to take into account such reasons, in addition to implementing targeted interventions to increase the equity of participation.
While the national monitoring of the NBCSP9,10 is useful, it does not provide the necessary details at a state-based level (and lower levels of aggregation such as postcodes) required to inform service planning. Moreover, it does not highlight the specific geographical areas that might benefit from renewed attempts at targeted interventions designed to improve the equity of participation. For these reasons, we have developed maps of NBCSP participation rates across South Australia (for a copy of these maps, contact the corresponding author), which reveal geographical differences in participation rates between postcodes. Such maps may be useful to healthcare policy makers, planners and practitioners who can now identify distinct geographical areas for further intervention in order to increase participation rates. For example, postcodes with high levels of participation would be useful in order to understand the particular reasons and facilitators for participation, which if applicable to policy transfer, may be transferred to other areas. For example, they might reflect a greater awareness of screening, stronger positive belief in illness prevention or prior exposure to screening programs. Conversely, postcodes with low levels of participation may be in need of targeted approaches to screening, including public education campaigns and interventions by health workers. In either case, more in-depth research is needed in order to further understand the reasons for the inequities in CRC screening participation identified in this paper so that targeted interventions can be developed and implemented with a view to increasing equity.
The authors wish to acknowledge SA Health for funding the project, under the Strategic Health Research Program and Medicare Australia for providing access to the data.