South Asian Ethnicity, Socioeconomic Status, and Psychological Mediators of Faecal Occult Blood Colorectal Screening Participation: A Prospective Test of a Process Model

Objective: Although ethnicity and socioeconomic status (SES) correlate with health inequality, efforts to explain variance in health behavior attributable to these factors are limited by difficulties in population sampling. We used ethnicity identification software to test effects of psychological beliefs about screening as mediators of ethnicity and SES on faecal occult blood colorectal screening behavior in a no-cost health care context. Method: Adults aged 50–67 years (N = 1,678), of whom 28% were from minority South Asian religiolinguistic ethnic groups (Hindu-Gujarati/Hindi, Muslim-Urdu and Sikh-Punjabi), participated in a prospective survey study. Subsequent screening participation was determined from medical records. Results: Screening nonparticipation in the most deprived SES quintile was 1.6 times that of the least deprived quintile. Nonparticipation was 1.6 times higher in South Asians compared with non-Asians. A process model in which psychological variables mediated effects of ethnicity and SES on uptake was tested using structural equation modeling. Self-efficacy and perceived psychological costs of screening were, respectively, positive and negative direct predictors of uptake. Paths from Hindu, Muslim, and Sikh ethnicity, and SES on uptake were fully mediated by lower self-efficacy and higher perceived psychological costs. Paths from South Asian ethnicity to participation via self-efficacy and psychological costs were direct, and indirect via SES. Conclusion: SES is implicated, but does not fully account for low colorectal screening uptake among South Asians. Targeting increased self-efficacy and reduced perceived psychological costs may minimize health inequality effects. Future research should test independent effects of SES and ethnicity on lower self-efficacy and higher psychological costs.


South Asian Ethnicity, Socio-economic Status and Psychological Mediators of FOBt
Colorectal Screening Participation: A Prospective Test of a Process Model.
Despite established inequalities in a range of health outcomes and virtually all health behaviors, there is a paucity of research that has directly evaluated the roles of both ethnicity and socio-economic status together with mediating psychological influences on health related behavior and uptake of health services. This is most likely because the low absolute frequency of minority ethnic people in the population creates significant difficulty in surveying adequate numbers even in large randomised population surveys. For example, the largest minority ethnic group in the UK population is South Asian, representing 5% of the UK population (UK Census 2011). Further, incomplete recording of ethnicity in population databases or in medical records (Iqbal, Johnson, Szczepura, Wilson, Gumber et al., 2012) precludes collecting samples in which South Asian men and women are accurately represented through oversampling. The present study overcame these difficulties by employing name recognition software (Nam Pehchan; Cummins et al.,1999) to pre-screen names in a population database so as to oversample from the South Asian population and achieve an ethnically diverse sample comprising adequate numbers of both South Asian and non-Asian Britons. The goal of this prospective study was to test a process model evaluating the role of socio-economic status and psychological variables in mediating effects of ethnicity on objectively-observed faecal occult blood test (FOBt) colorectal screening participation. It is important to consider the roles of ethnicity and socio-economic status on screening participation to establish the extent to which ethnic disparities in health-seeking behavior can be attributed to deprivation.
We also aimed to establish whether ethnic effects on screening participation could be explained by beliefs about treatment and conditions.
Colorectal cancer is the third most common cancer worldwide (International Agency for Research on Cancer, 2012) and in the UK and US (Office for National Statistics, 2012;American Cancer Society, 2015). Survival rates are favorable when disease is detected at an early stage, but patients presenting with advanced disease have a high mortality rate (Maringe, Walters, Rachet, Butler, Fields et al. (2013). Screening by faecal occult blood testing (FOBt) significantly reduces colorectal cancer mortality and can reduce cancer incidence through detection and removal of colorectal adenomas (Hewitson, Glasziou, Irwig, Towler & Watson, 2007;Hewitson, Glasziou, Watson, Towler & Irwig, 2008). Hewitson et al (2007) reported a 16% reduction in relative risk of colorectal cancer mortality in trial participants allocated to FOBT screening conditions. When their analysis included only those who actually completed screening, the relative risk reduction was 25%, underlining the importance of identifying psychological processes that might explain and promote screening participation.
Screening uptake tends to be low and to vary with socio-economic status (e.g. Decker, Demers, Nugent, Biswanger & Singh, 2015;Joseph, King, Miller & Richardson, 2012). Even in the UK where the National Health Service routinely invites all eligible adults for free screening and any necessary treatment, uptake rates in the most deprived quintile of residential areas are almost half those of the least deprived quintile of areas (35% vs. 61%; von Wagner, et al., 2011). Whereas socio-economic status indicators can be attached to individual patient postal codes in order to examine inequality, estimates of inequalities amongst minority ethnic populations have tended to rely on area-level analyses that cannot be linked to individual screening records. However, the use of name-recognition software to identify South Asian ethnicity showed that South Asians demonstrated significantly lower FOBt screening uptake than non-Asian Britons (32.8% vs. 61.3%) (Szczepura, Price & Gumber, 2008;Price, Szczepura, Gumber & Patnick, 2010).
Screening has been described as a 'risky' health behavior insofar as it involves making a decision to undergo procedures with uncomfortable or upsetting short-term effects to learn of future disease threat and obtain a longer-term health benefit (Rothman & Salovey, 1997;Orbell, Perugini & Rakow, 2004). Motivation for screening participation therefore involves dual psychological influences; motivation to reduce disease threat (vulnerability to and severity of disease) and motivation to engage in a recommended response (by taking up screening) which involves appraisals of likely effectiveness, difficulties and psychological costs associated with unpleasant procedures or outcomes (response efficacy, self-efficacy and response costs). These psychological influences are common to many theoretical accounts of health related behavior (Ripptoe & Rogers, 1987;Janz & Becker, 1984;Weinstein, 1988;Schwarzer, 2008). Application of these theories has been advocated to identify the psychological variables that explain substantive variance in screening behavior. This is considered an important formative step in identifying the target constructs that can be manipulated in behavioral interventions to promote screening.
Application of health behavior thories may also assist in tackling these health inequalities by identifying the psychological variables that account for effects of social structural variables such as ethnicity and deprivation on health behavior. Psychological factors may explain variability in health behavior due to socioeconomic and cultural factors beyond financial constraints that limit access to care. For example, social conditions that cannot cushion short term loss, or which have been characterized by limited efficacy to overcome or prevent negative life experience may enhance the perceived costs of participating in screening or diminish self efficacy to complete the test. There is some empirical evidence that these appraisals may differ by socio-economic status (e.g., Orbell, Johnstone & Crombie, 1996;Whitaker, Good, Miles et al., 2011). However, there is a paucity of studies that have employed population samples, prospectively collected data, objectively observed behavior, or used mediation analyses to examine whether psychological constructs mediate socio-economic status effects on screening participation (von Wagner, Good, Whitaker & Wardle, 2011). Moreover, studies to date have employed largely homogeneous white samples and none have employed a sufficiently diverse sample to enable investigation of ethnicity, socio-economic status and psychological variables in the same anaysis, so that it remains uncertain whether variance attributable to ethnicity and social deprivation might be explained by similar psychological processes. Considerable evidence suggests that ethnicity covaries with deprivation (e.g., Williams, Mohammed, Leavell & Collins, 2010), suggesting the hypothesis that pathways to health behavior may be explained by psychological variables associated with socioeconomic deprivation. The extent to which variability in screening participation attributable to ethnicity cannot be accounted for by socioeconomic deprivation will indicate the need for further investigation of distinct ethnicity influences on health behavior.
If preventive services such as screening are differentially used by different SES and ethnic groups, mortality rates would subsequently show even stronger disparities over time (e.g., Maringe et al., 2013). In the present study we aimed to identify the factors that explain the effects of South Asian ethnicity and socio-economic status on participation in FOBt colorectal screening. We expect to provide valuable insight into the processes by which psychological and social structural variables impact on screening and provide data that may inform intervention development. Specifically, we predict that (a) South Asians will have lower participation in FOBt screening compared to the non-South Asian population, (b) low socio-economic status will be inversely associated with FOBt screening participation, (c) the effect of ethnicity on participation in FOBt screening will be mediated by socio-economic status, and (d) psychological variables will be direct predictors of uptake and mediate the effects of ethnicity and social deprivation on FOBt screening participation.

Setting, Participants and Design
The colorectal cancer screening program in the UK is funded nationally and organized and delivered regionally, without direct involvement of primary care providers. All ageeligible men and women are sent a biennial guaiac-based FOB test to complete at home. Random samples of men and women meeting the eligibility criteria for an invitation to FOBt screening were drawn from screening databases in England and Scotland. Oversampling was utilized to ensure representation of people with lower socio-economic status and of South Asian ethnicity. Over-sampling by deprivation category was derived from Carstairs indexes linked to individual postal codes. In order to ensure that adequate numbers of minority ethnic South Asians were included, name recognition software, Nam Pehchan, for which sensitivity and specificity values of 95% (Gumber, 2006) and 97% (Honer, 2003) have been recorded, was used to assign an ethnicity label to 132,992 men and women in the screening database in England. The program contains a dictionary of South Asian names that are matched against the complete name or the name stem in order to provide a list of South Asians together with a language and religion marker for each person so that individuals can be placed into different religio-linguistic groups: Hindu-Gujarati; Hindu-Other, Muslim-Urdu; Sikh-Punjabi (Szcepura et al., 2003 Appendix 2). The software identifed a total of 6,450 individuals belonging to one of these groups (4.8%) and a stratified sample was drawn from this subsample. For the purposes of the present analyses, the two Hindu subcategories were collapsed into a single category. The response rate was 49%. Response to the questionnaire varied by age and social deprivation but there was no association with gender. Older participants (χ 2 (3) = 48.792, p < .001) and the least deprived (χ 2 (1) = 55. 093, p < .001) were more likely to return a completed questionnaire. Non-Asians were more likely to return a questionnaire than South Asians (χ 2 (1) = 629.878, p < .001).
Linkage to NHS screening records. Data from questionnaires was linked to response to a subsequent FOBT invitation approximately 24 months later using National Health Service identification numbers. NHS matched screening outcome data was available for 1851 questionnaire respondents at follow up. Questionnaire respondents who were not invited to complete an FOB test in the intervening years because they were age ineligible, deceased, undergoing current treatment, had moved away from the screening region or could not be identity matched are summarized in Appendix 1.

Cross validation of ethnicity identifcation. A UK census format ethnicity self report
item was included in the survey. Respondents were asked to assign themselves to one of five categories (Black or Black British, Mixed, Asian or Asian British, White, Chinese/Other) and to further specify their ethnicity within the chosen category. Responses to this item were cross referenced against the ethnicity labels assigned by the Nam Pehcham software (Appendix 2).
Fifty eight people did not provide ethnicity self report data and a further 115 people were misclassified (6%). It was decided that the most approriate strategy in the present context was to exclude these 173 participants whose ethnicity was unverifiable, leaving a final sample of 1678. Characteristics of the final ethnically and socio-economically diverse study sample are summarized in Table 1.

Procedure
All eligible adults were sent a postal questionnaire along with a letter explaining that the purpose of the study was to understand what people think about bowel cancer and what they think about doing the bowel cancer screening test. A freepost return envelope was included. Letters sent to sampled individuals identified a priori by name recognition software as South Asian included a passage translated into five languages inviting people to seek assistance from an English speaker if required. A reminder letter was sent one week later, and a second booklet and reminder letter was sent two weeks later. They were informed that questionnaire completion constituted consent to participate and those returning completed questionnaires were entered into a prize lottery for a £50 high street voucher. Psychological measures. Thirty items were included to assess the five psychological constructs. Focus groups were employed to elicit relevant content and constructs were operationalized according to standard procedures to ensure content validity. All items were scored on six-point Likert scales unless specified otherwise. Severity comprised eight items assessing physical and psychosocial perceived impacts of bowel cancer, for example "If I were to develop bowel cancer; it could almost certainly cause my death (disagree very strongly-agree very strongly". Vulnerability comprised six items (e.g.,"I think that my chances of developing bowel cancer are very low (agree very strongly-disagree very strongly)". Response efficacy comprised eight positive expectancies each scored on a scale from extremely likely to happen-extremely unlikely to happen, for example "Doing a bowel cancer screening test in the future would reduce my chances of dying from bowel cancer".

Measures
Response costs comprised five negative expectancies each scored on a scale from extremely likely to happen-extremely unlikely to happen, for example "Doing a bowel cancer screening test in the future would be embarrassing; would lead to unpleasant treatment if abnormalities were present; would be disgusting; would be unhygienic". Self-efficacy comprised three items "If I am invited to do a bowel cancer screening test in the future; I am certain that I could do it (extremely certain-extremely uncertain)". Full questionnaire items are presented in Appendix 3 as supplemental materials.

Structual equation model testing mediation effects. Structural equation modelling
was employed to test the hypotheses of our process model that included psychological variables and socio-economic status as mediators in a two-stage mediation model. In the first instance, a confirmatory factor analysis (CFA) model was estimated to test whether the covariance matrices among items could be adequately explained by a set of latent and nonlatent variables representing the hypothesized psychological and demographic constructs and a dichotomous measure of participation in the FOBt screen. Specifically, items pertaining to the self-efficacy (n = 3), response efficacy (n = 8), response cost (n = 5), perceived severity (n = 8), and perceived vulnerability (n = 6) were set to indicate latent variables in the model while SES (Carstairs index) was included as a non-latent variable. In addition, we included age and gender as control variables in the model such that each variable was set to predict all other model variables. Consistent with standard practice for CFA models all latent and nonlatent variables were allowed to covary and a single indicator of each latent factor was set to unity to define its scale. Following adequate fit of the CFA model a structural equation model was estimated that included structural parameters representing the hypothesized relations among the model constructs. Specifically, the demographic variables were set as independent predictors of the psychological variables and the psychological variables were proposed as independent predictors of participation. Direct effects of the demographic variables on participation were also freed.
We tested our hypotheses using a structural equation model (SEM). In the model, the three dummy-coded dichotomous variables representing ethnicity group membership (Hindu, Muslim, Sikh) were set as predictors of SES, SES as predictor of each of the latent psychological variables (self-efficacy, response costs, response efficacy, vulnerability, and severity), and the psychological variables as predictors of participation. This model enable us to test a series of three-path sequential indirect effects of each ethnicity variable on participation through SES and each psychological variable (e.g., hindu ethnicity SES selfefficacy participation). We also included direct effects of the ethnicity variables on the psychological variables. This enabled us to test a series of two-path indirect effects of each ethnicity variable on participation through each psychological variable (e.g., muslim ethnicity response costs participation). This test the alternative hypothesis that effects of ethnicity on participation are subsumed by the psychological constructs, but independent of SES. Finally, we also included direct effects of the ethnicity variables and SES on participation to test whether direct effects of these demographic variables in the presence of the indirect effects. This enabled us to test whether the effects of ethnicity on participation are due to variations in deprivation, or beliefs regarding the behavior and condition, both, or neither. The MPlus computer program (Muthén & Muthén, 2015) was used to estimate the specified CFA and SEM models using a robust maximum likelihood method. Multiple criteria were adopted to evaluate model goodness-of-fit including the comparative fit index (CFI), non-normed fit index (NNFI), the standardized root mean square of the model residuals (SRMSR), root mean square error of approximation (RMSEA), and the 95% confidence intervals of the RMSEA (CI95). Values in excess of .90 are indicative of reasonable model fit for the CFI and NNFI indexes (Bentler, 1990), although values approaching or exceeding .95 are preferable (Hu & Bentler, 1999). Cut-off values of .50 and .08 or less for the SRMSR and RMSEA are considered indicative of good fit, with narrow 95% confidence intervals for the RMSEA (Hu & Bentler, 1999). In addition, we also examined the adequacy of the solution estimates of the CFA model, namely, the standardized factor loadings which should exceed .70, the average variance extracted from the items in each factor which should exceed .50, and the composite reliability ( c) estimates which should be greater than .80. 1

FOBt Uptake at Follow-up
Overall 382 respondents (22.8%) did not complete FOBT at follow up. As hypothesized, participation in screening at follow-up varied by ethnicity. Non-participation rates were respectively; 19.6% British white, 30.6% Hindu, 42.6% Muslim and 25.3% Sikh ( 2 (3) = 36.45, p < .001). Non-participation also varied by SES (χ 2 (4) = 14.65, p < .001) and showed a linear association across the distribution of deprivation, rather than a specifically high non-participation amongst the most deprived group. Non-participation rates across five quintiles (most deprived to least deprived) were 29.8%, 24.4%, 21.3%, 23.0% and 18%. No association was observed with age (M = 58.18, SD = 5.14 screened vs. M = 57.96, SD = 5.37 1 It should be acknowledged that alternative models are possible. One such alternative model proposes that SES and ethnicity might moderate the relationship of psychological variables on behavior (e.g., Schüz, 2017). We ran a series (n = 20) of logistic regression models in which participation was regressed in turn on each of the psychological variables along with either SES or one of the ethnicity dummy variables, together with the interation of SES X psychological variable or ethnicity X psychological variable. The interaction term did not obtain a significant relation with participation in any of the regression models. non screened; t (1676) = -.73, p =.462) or gender (χ 2 (1) = .04, p = .846; 23% vs 22.6% nonparticipation for women and men respectively).

Confirmatory Factor Analysis and Structural Equation Model
The CFA supported the construct validity of the latent psychological variables. CFA goodness-of-fit estimates revealed adequate fit of the model according to the multiple criteria  Table 3 and statistically significant paths are illustrated in Figure 1.
Membership of Hindu, Muslim, and Sikh ethnic groups were statistically significant direct predictors of SES, and SES was a statistically significant predictor of self-efficacy and response cost. In addition, there were statistically significant direct effects of Hindu, Muslim, and Sikh ethnic groups on response efficacy, vulnerability, self-efficacy, and response cost.
However, only self-efficacy and response cost were statistically significant direct predictors of participation. Given self-efficacy and response cost were the only predictors of participation, we expected three-path indirect effects of the ethnicity variables on participation with SES and self-efficacy or response cost as multiple sequential mediators. Consistent with our hypotheses, we found statistically significant and negative three-path indirect effects of Sikh, Hindu, and Muslim ethnicity on participation through SES and self-efficacy. However, the effects of ethnicity on participation were not exclusively mediated by SES. There were also statistically significant indirect effects of ethnicity on participation that werethrough the psycholgical variables and not mediated by SES. Specifically, there were statistically significant two-path indirect effects of Sikh, Hindu, and Muslim ethnicity on participation with self-efficacy or response costs as the single mediator. The only exception was the indirect effect of Muslim ethnicity on participation through response cost, which fell short of the conventional level for statistical significance (p = .052). Importantly, there were no direct effects on of any of the ethnicity variables or SES on participation. Effects of ethnicity on participation were therefore mediated by SES and the psychological variables in the threepath indirect effects, or by the psychological variables only in the two-path indirect effects.

Discussion
Uniquely, this study employed indices of SES, ethnicity, psychological variables and behavior assessed at the individual level to evaluate the role of socio-economic status and psychological constructs in mediating effects of ethnicity on colorectal screening uptake. As expected, South Asian ethnic minorities and people with lower SES were under-represented amongst the screened population at follow up. SES also showed a gradient relationship with FOBt uptake, consistent with previous research (e.g. von . The structural equation model showed that the paths from South Asian Hindu, Muslim and Sikh ethnicity, and socio-economic status on uptake were fully mediated by lower self-efficacy and higher perceived response costs. The paths from South Asian ethnicity to participation via selfefficacy and response costs were both direct, and indirect via socio-economic status, indicating a residual influence of ethnicity on uptake that was not attributable to socioeconomic status but which was nonetheless mediated by lower self efficacy and higher response costs. FOBt screening delivered within a cost free health care system involves a selfadministered sampling procedure that does not involve travel to clinics, time off work or contact with health professionals. In this context, perceived psychological costs of completing the test kit and self efficacy to complete the kit fully explained variability in uptake attributable to socioeconomic status. Social and economic conditions that limit opportunities for future planning, or that cannot cushion short-term emotional, social and economic costs, may enhance response costs associated with screening, particularly those occurring in the short term (Orbell, Perugini & Rakow, 2004;Whitaker et al, 2011). These enhanced costs include those that may arise from potential treatment implications of an abnormal result, if the test is taken, such as hospital appointments, medical procedures and time off work, and also from aversive aspects of the self sampling procedure itself, such as disgust and embarrassment. It is not clear why these latter costs might show a gradient relationship with socioeconomic deprivation, although it is explicable for example, that housing conditions might impact upon privacy or embarrassment associated with collecting samples and storing the kit before posting. Screening by FOBt is a complex behavior, requiring confidence to follow instructions to undertake self-sampling (and to do it correctly) and ability to manage negative emotions associated with handling faeces (e.g., embarrassment, disgust). Generally low self agency as a consequence of social experience may explain the SES differentials observed here. Evidence that self-efficacy and response costs are important mediators of both socio-economic status and ethnicity via deprivation suggests that a common strategy might be appropriate to address social sources of self-agency that may impact upon efficacy to plan how to collect samples, or plan to manage negative emotion, for example (Greiner et al., 2014;Schwarzer, 2008). In addition, Orbell et al. showed that emphasising short term benefits of screening participation may be useful in shifting attitudinal focus towards screening participation.
The South Asian samples included in the present study were all less likely to complete a screening kit than non-Asian Britons. Our findings suggest two psychological routes by which ethnicity might exert residual effects on behavior because we obtained direct effects of ethnicity on participation via self efficacy and response costs.The religio-linguistic subpopulations distinguished by these analyses differ on a number of dimensions from the white British sample, including country of origin, religion, language and literacy, and traditional diet (Szczepura, 2010). It is possible that cultural influences impact on self efficacy and enhance the psychological costs of collecting and storing stool samples, and of positive results, if social stigma is attached to a cancer diagnosis, or interactions with medical professionals are perceived to be aversive. South Asian cultures tend to score more highly on collectivism than British culture (Hofstede, Hofstede & Minkov, 2010). Collectivism confers an interdependent self conception in which the self is embedded in social context and defined by social relations. Behavioral motives are guided by avoiding negative outcomes and social group disruption, such as not burdening others, and conformity to community norms and expectations, although much of the previous evidence is based on East Asian samples. It is possible that evidence that collectivist cultures are more responsive to health messages that emphasize avoidance of loss associated with not acting, or that emphasize relational outcomes, or affirm values concerned with avoiding negative things in life (e.g., Sherman, Uskul & Updegraff, 2011) may inform future investigation of non-participation in screening in South Asian communities. Establishing cultural group screening norms and emphasizing community aspects of mass screening programs may also be important.
The threat appraisal variables, severity and vulnerability, were not significantly related to FOBt uptake in our structural model, consistent with evidence that coping appraisal is more reliably associated with a range of health behaviors, perhaps because of its conceptual proximity to behavioral enactment (e.g. Milne, Sheeran & Orbell, 2000). Although not significantly associated with uptake it was interesting to observe significant direct relationships from ethnicity to perceived vulnerability and response efficacy for all three ethnic minorities such that membership of a South Asian group was associated with lower perceived vulnerability to colorectal cancer and lower perceived screening efficacy. These variables were not associated with socioeconomic status in the current structural model. A few studies have suggested that low perceived vulnerability in South Asian populations might be attributable to beliefs that vulnerability is indicated by existing symptoms, (e.g. Lo, Waller, Vrinten, Kobayashi & von Wagner, 2015) consistent with low endorsement of cognitions concerning benefits of early detection and treatment observed in the present study. An alternative, albeit, to date, under-investigated, possibility might be that South Asian populations consider their ethnicity to confer group protection from colorectal cancer. World cancer statistics indicate significantly lower incidence of bowel cancer in South Asia than in Western countries (International Agency for Research on Cancer, 2012) and older British immigrant South Asian populations such as those currently age eligible for screening may therefore perceive low ingroup risk. Historical trends in risk are, however, unlikely to be sustained during acculturalisation and low participation in screening may ultimately lead to a widening gap in cancer survival (Sczepura et al., 2008;Maringe, Mangtani, Coleman & Rachet, 2015). Observed rises in disease incidence and increasingly prevalent behavioral risk in South Asia has led to recent calls for bowel cancer screening (e.g., Bhurgri et al., 2011).
Importantly, current findings indicate that variability in perceived vulnerability was not associated with variability in screening uptake. Increasing perceived vulnerability might therefore have little direct impact on uptake, consistent with meta analytic findings that show small effect sizes for the relation between perceived risk and behavior whether assessed correlationally (Atkinson, Salz, Touza, Yi & Hay, 2015) or experimentally (Sheeran, Harris & Epton, 2014). Efforts to increase perceived vulnerability may have limited impact on behavior change unless also accompanied by interventions that simultaneously address coping appraisal variables by increasing self efficacy and decreasing perceived psychological costs of screening.
The sub-optimal reponse rate might be considered a limitation of the study although the response rate observed in the current study is in line with similar studies. However strengths of the study include the objective assessment of screening participation, stratified random population sampling and the observed prospective relationship of both socioeconomic status and ethnicity to subsequent screening uptake. In this study which included only questionnaire respondents who might be considered to have good literacy, screening non-participation in the most deprived quintile was 1.6 times that of the least deprived quintile. Similarly, non-uptake amongst South Asians was 1.6 times higher than that of non-Asians. Muslims also had the lowest observed uptake amongst South Asian groups, consistent with Szczepura et al (2008). It seems most likely that consideration of questionnaire nonrespondents might only enhance these observed inequalities.
In summary, our process analysis of the effect of ethnicity on screening uptake supports the view that socio-economic status is implicated in, but does not fully explain, variance attributable to South Asian ethnicity. Whilst interventions that target perceived negative psychological costs of screening and enhance self efficacy are indicated to tackle inequality generally, it will also be important to consider how ethnicity might impact directly on these beliefs and develop strategies that address ethnicity specific sources of low selfefficacy and high response costs.   Hindu→SES→Severity→Participation .000 .000 .000 .000 .972 Note. A two-path indirect effect involves the effect of an ethnicity variable on an outcome variable with a single mediator. A three-path indirect effects involves the effect of an ethnicity variable on an outcome variable with two sequential mediators. Effects in boldface are statistically significant (p < .05). a Coefficients are standardized values. CI95 = 95% confidence intervals of the parameter estimate; LB = Lower bound of the 95% confidence interval; UB = Upper bound of the 95% confidence interval. SES = Socio-economic status measured by the Carstairs index(high scores indicate lower SES or more deprivation); Hindu, Muslim, and Sikh variables are dummy-coded dichotomous enthnicity variables with 1 = member of the stipulated ethnic group and 0 = non-member of the stipulated ethnic group.