Design and sample description
These data were collected as part of Open Doors to Health (ODH), an NCI-funded study of colorectal cancer prevention delivered through low income housing. This study used a cluster randomized design, with 12 housing sites as the primary sampling units. Secondary sampling units were individuals within the site. Unequal probability sampling across housing sites was employed due to the varying size of housing sites. In half of the sites (with population size less than 300 residents), the full population was sampled; in the remaining sites, with populations greater than 300, sampling was conducted to obtain an approximate 35% sample with a minimum of 250 subjects per site. The data reported here represent baseline data only, among baseline respondents who were under age 50 at the time of the survey.
Eligibility criteria for the larger study, from which the data reported in this paper are drawn, included: (1) residence in the housing community; (2) age of at least 18 years; (3) English or Spanish fluency; (4) not currently being treated for cancer. Participant recruitment began with housing site representatives sending letters announcing the study to eligible residents, including the opportunity to opt out of contact. Participants provided informed consent, and the interviewer-administered survey was completed in English or Spanish. The study protocol was approved by the Human Subjects Committee at the Harvard School of Public Health.
An initial sample of 3,368 residents was drawn, from which 747 (20%) were deemed ineligible, leaving an eligible sample of 2,941 individuals. Of these, 828 (28%) refused participation, and 559 (19%) were never reached. Enrollment and baseline surveys were obtained on 1,554 participants (53% response rate). Among these participants, there were 718 residents who were under age 50. Of them, 26 (3.6%) reported having already been screened for colon cancer. Since the goal of this paper was to determine associations with screening awareness, we eliminated this group from the analysis. Therefore 692 unscreened residents under the age of 50 constitute the sample for this paper.
The primary outcomes for this paper are awareness of CRC screening and intentions to be screened. We measured key constructs from our Social Contextual Model as factors that may influence screening awareness and participation.
Primary outcomes included awareness of screening guidelines and intentions to be screened among those under age 50. Awareness of CRC screening was assessed by one item, ‘‘Have you ever heard about screening for colon or rectal cancer’’. Participants were also asked their intentions to be screened by one item ‘‘Do you planned to be screened for colon cancer?’’ (response options: yes, no, don't know).
We measured key constructs from our social contextual model as factors that may influence screening awareness and intentions. Three categories of predictor variables were included: (1) sociodemographic characteristics; (2) social contextual factors; and (3) health care.
Sociodemographic variables included gender, race and ethnicity, age, educational attainment, immigration status, and whether English was the first language. We determined whether individuals were at/below or above poverty level by combining yearly household income and the number of people supported by that income based on the 2005 Federal Poverty Guidelines (http://aspe.hhs.gov/poverty/05poverty.shtml
Social contextual variables included social networks, assessed using Berkman's measure of network affiliations. The measure included four items: (1) whether the participant has a spouse or partner; (2) how many relatives they feel close to; (3) how many close friends they have; and (4) whether they are an active member of any groups or clubs [42
]. We assessed social support in three of the four domains of the Inventory of Social Supportive Behaviors, emotional, instrumental, and financial support, using one item from each domain. Participants responded to each item as support ‘‘always available,’’ ‘‘sometimes available,’’ ‘‘available but wouldn't accept help,’’ ‘‘don't need help,’’ and ‘‘not available.’’ Social support was then computed as the number of domains always or sometimes available to the respondent (0 to 3) [43
] (alpha = 0.7689). Role responsibilities was computed as the number of family roles that the respondents has most or all of the responsibility for (0 to 3). The roles asked about include: ‘‘earning money to support the family,’’ ‘‘taking care of your own or other children,’’ and ‘‘taking care of another household.’’ [17
]. Role conflicts was assessed by a single item asking whether one's daily activities make conflicting demands (response options: yes/no) [17
]. To assess social cohesion in the housing community, we asked respondents to report their agreement (on a scale of 1 (strongly agree) to 4 (strongly disagree)) with five statements: (1) people around here are willing to help their neighbors; (2) this is a close-knit neighborhood; (3) people in this neighborhood can be trusted; (4) people in this neighborhood generally do not get along with each other; and (5) people in this neighborhood do not share the same values. Item responses were reversed for the first 3 statements and then responses to the 5 items were averaged. The summary score ranged from 1 to 4, with a higher score indicating social cohesion [44
Health care variables included assessment of whether the respondent had a regular health care provider, how well they felt that their provider knows them across three domains, including: (1) the respondent's responsibilities at work, home, or school; (2) the respondent's worries about their health; and (3) the respondent as a person, and his/her values and beliefs (response options 0–3; summary score reflects number of items for which respondents feel provider knows them somewhat or very well). Decision autonomy was assessed by a single item reflecting agreement with the statement ‘‘I leave all decisions about screening tests to my doctor.’’ We also assessed the number of times in the past year a participant had seen their health care provider, and whether they had health insurance (the state Medicaid plan, private insurance, or no insurance).
For all analyses based on the cluster design, data are weighted up to the population size within each housing site; for the non-screened population under age 50, the weighted size is 1,004. Frequency distributions (for categorical variables) and estimates of means and standard deviations (for continuous variables) were assessed for distributional assumptions and outliers. Social contextual variables were used to create composite variables for social networks, social support, role responsibilities and social cohesion as defined above. Cronbach's alpha values were used to determine the level of agreement between variables where conceptually appropriate. Bivariate associations between the dependent variables, CRC screening awareness and intentions to be screened, and all independent variables that appear in were assessed and variables found to be significant at the p = .15 level in bivariate analyses were retained for use in multivariable modeling. Variables that were used to create composite variables were not tested individually but only in composite form. Based on the bivariate associations and consideration of effect modifiers (i.e., interaction effects) and confounders, multivariable logistic models of the dependent variables were developed. Interaction effects were tested and considered for the models when inclusion of confounders altered coefficients of significant variables by more than 10%. The two interaction effects that fit this criteria were in the model of social awareness. English as a first language and immigrant status was one interaction and the other was gender by insurance status. All data presented reflect weighted analyses and counts unless otherwise specified. SUDAAN Version 9.01 and SAS Version 9.1 statistical software for clustered data were used for all analyses.
The relationship between demographic, social contextual, and health care variables and awareness of CRC screening
Demographic characteristics are presented in . All percentages reported in this table are weighted to reflect population percents. Among the 692 respondents, 78% were women, 54% had completed high school or less education, 23% were immigrants, 44% had English as a second language, and 49% were living at or below the poverty level. Fifty-three percent had Medicaid or free care, 33% had private insurance only, 8% had a combination of public and private, and 6% had no health insurance. Eighty-one percent had a regular health care provider. also presents demographic variables by awareness of CRC and intentions to be screened. Overall, approximately two-thirds of the sample was aware of CRC screening (70%) and intended to be screened in the future (66%).
Demographics by screening awareness and intentions (unweighted n=692)
Factors associated with CRC screening awareness
presents the relationships between the predictor variables and awareness of CRC screening. Bivariate analyses indicated an association between screening awareness and several demographic variables, including age, education, gender (p = .07), immigrant status, and having English as a 2nd language. Screening awareness was higher among women, those who were older, those with more education, immigrants, and native English speakers. Social contextual factors that were associated with screening awareness included role responsibilities and role conflicts (p = .07), with higher awareness among those with more role responsibilities and role conflicts. Health care variables that were associated with awareness included having a provider that knows you well (p = .07), and having insurance other than Medicaid or Medicare. Those with private insurance only had significantly higher awareness levels compared with those with no insurance, Medicaid or Medicare.
In multivariate analyses, the association remained between awareness and both age and education. Interaction effects were observed between immigrant and English as a second language; immigrants who had English as a second language had lower awareness. An interaction was also observed between insurance status and gender; females tended to have higher awareness if they had private insurance than if they had public ± private insurance, whereas there were no significant differences seen for males.
Factors associated with intentions to be screened (see )
Bivariate analyses indicated associations between intentions and gender, frequency of seeing one's physician in the past year, role responsibilities, role conflict, and social cohesion. Intentions were higher among men, among those who had seen their provider between 3 and 11 times in the past year, had more role responsibilities, had more role conflicts, and those who reported higher levels of social cohesion in their housing site. In multivariable analyses, the association persisted between intention and gender, role responsibilities, role conflicts, and social cohesion.
The relationship between demographic and social contextual variables and screening intentions (n = 692 unweighted)
Colorectal cancer screening rates in the US are suboptimal, particularly among lower income and racial/ethnically diverse groups. Targeting education efforts at those age 50 and older fits with screening guidelines recommending CRC screening for average risk adults starting at age 50. However, if specific populations have limited awareness or understanding of the importance of screening when they reach age 50, there will likely be delay in screening adoption, which may ultimately contribute to disparities in RC morbidity and mortality.
This study investigated awareness of CRC and intentions to be screened at age 50 among a large population of low income, racial, and ethnic minority adults. The majority of respondents (61%) were between ages 30 and 49, and interestingly largely insured through the state Medicaid program, presumably a function of the population's income qualification for housing subsidies. Overall, there were relatively high levels of awareness of CRC screening, with about 70% of the sample reporting that they had heard about screening for CRC. Intentions were also reasonably high, with 66% of the sample reporting intentions to be screened at age 50. Both awareness and intentions were higher than we had expected and than has been found in other studies [45
], although more recent studies suggest higher levels of awareness and intentions to be screened, [47
] and nationally screening rates are starting to increase [48
]. It is possible that our findings reflect sampling bias, and that we reached those who were most likely to be aware. However, there was a wide range of variability on factors that are associated in national databases with screening behavior and large numbers of people with characteristics associated with lower screening rates (e.g., education, income). It is more likely that the higher awareness and screening intention levels in this sample reflect the extensive public health campaigns about CRC screening that have targeted lower income populations. That there were no differences in race and ethnicity related to screening awareness or intentions is also notable, and may reflect the impact of these campaigns. It will be important to follow younger, diverse populations such as this that may have been exposed to public health messages about CRC to determine if these early indicators of readiness translate into actual screening behavior. These findings suggest that some optimism related to reduction of screening disparities by race/ethnicity and SES may be warranted, but longitudinal analyses are needed to answer these questions.
Bivariate analyses illustrated a dose-response relationship between age and screening awareness, with increasing awareness at each age level; this association remained in multivariable analyses. This finding is encouraging because it demonstrates that as minorities age, they become more aware of the need for CRC screening. However, this greater awareness did not translate into greater intention to be screened. In addition, although women were more likely than men to be aware of CRC screening, they were significantly less likely to report intention to be screened at age 50 years. This follows the national trend of differential CRC screening by gender, with more men than women being screened [8
]. One possible explanation is that women perceive their risk for CRC to be lower compared to men and therefore do not need to be screened. Clearly, additional studies need to flush out these varying relationships, with the goal of identifying culturally-appropriate interventions to increase screening intentions among women and among all age groups, but particularly those approaching age-eligible screening.
Interestingly, few social contextual variables were associated with screening awareness, or were any of the health care variables other than insurance status. In contrast, screening intentions were associated with social contextual variables. Intentions to be screened were almost three times higher among those with more role responsibilities compared to those with fewer, and almost half as likely among those who had fewer role conflicts. Social cohesion was also associated with screening intentions. None of the health care variables were associated in multivariable analyses with screening intentions, including health insurance status. It is interesting that social context is related with screening intentions, but not awareness. It would make sense that contextual factors are bigger drivers of, or impediments to, taking a preventive health action, vs. knowledge. Individuals who have higher role responsibilities and role conflicts may in part be motivated to maintain their health by their need to care for others. That social cohesion is also associated with screening intentions, albeit somewhat weakly, is interesting—it would be worth exploring in longitudinal analyses whether this actually facilitates higher participation in screening. These may be important targets for intervention efforts, particularly related to making screening more convenient and conducive, given life's many demands on this population. It is encouraging that there was little variation in screening intentions by sociodemographic variables, which often have persistent and intractable associations with health behaviors.
Several limitations of this study should be noted. First, the response rate ideally would have been higher, although low income housing is a challenging environment in which to conduct research. In addition, specific challenges experienced in the initial two housing sites (during the study's startup phase) most negatively influenced the response estimate. We do not have data on non-responders, and thus cannot address potential response bias with certainty. These data are cross-sectional and thus the causal direction of the findings cannot be established. Longitudinal analyses are needed to determine causality. Several of the mediating variables were assessed using abbreviated measures. This was necessary to limit the response burden on this less educated sample. One factor that we anticipated being an important mediator was family history of CRC. However, only three participants reporting having first degree relatives diagnosed with CRC before age 50 and 18 had relatives diagnosed at a later age; thus it was not possible to examine the role of family history in screening awareness and intentions to be screened.
This study also had several strengths. This is a large sample of low income, diverse adults who are largely insured. This study is an opportunity to understand factors that may influence screening uptake in this somewhat ideal situation where health insurance rates are high, by virtue of the fact that the population's income eligibility for low income housing also makes them eligible for Medicaid or other insurance programs. Use of a social contextual framework makes it possible to evaluate a wide range of factors that can influence health behaviors.
If we are to make a meaningful impact on disparities in CRC morbidity and mortality by race/ethnicity and socioeconomic position, it is critical that we identify opportunities for maximizing screening uptake among those not yet age-eligible for screening. Different factors mediate knowledge about CRC screening and intention to get screened. Education efforts need to address factors associated with effectively communicating with populations with lower education, whereas efforts aimed at increasing screening compliance need to also address social contextual and structural factors to be effective. These findings provide some possible mechanisms that may need to be addressed in order to maximize screening rates to help people form intentions to be screened at age 50. Future research should evaluate whether this approach will increase screening, and ultimately contribute to reduction of CRC disparities.