Effect of Language on Colorectal Cancer Screening Among Latinos and Non-Latinos

Colorectal cancer (CRC) remains the third most common cancer in both men and women and the second leading cause of cancer-related deaths in the United States. (1). National guidelines, including those from the U.S. Preventive Services Task Force and the American Cancer Society, recommend CRC screening for all men and women age 50 years and older. Although there is no ideal screening method, recommended screening modalities include fecal occult blood testing (FOBT), sigmoidoscopy, colonoscopy, double-contrast barium enema (2), as well as computed tomography colonography and fecal DNA analysis (3). Although studies suggest that up to 60% of CRC deaths could be prevented with routine CRC screening (4), a major barrier to decreasing CRC morbidity and mortality is poor utilization of CRC screening procedures (5).

CRC is the second most commonly diagnosed cancer among Latinos, yet previous studies show that even after accounting for socioeconomic factors, CRC screening rates are lower in Latinos compared with non-Latino Whites (6, 7). Although previous studies suggest that some Latinos have misperceptions and erroneous understandings of cancer that may effect their preventive behavior (8, 9), language may also be an important potential barrier to screening as limited English proficiency has been implicated as a barrier to preventive care in general (10) and in breast and cervical cancer screening services (11, 12).

The purpose of this study was to examine the relationship between preferred language use (English versus Spanish) and self-reported receipt of CRC screening tests among Latinos and non-Latinos.

We conducted a cross-sectional study using data from the Centers for Disease Control's 2006 Behavioral Risk Factor Surveillance System (BRFSS) (13). Following guidelines for CRC screening, adults, the study population for this analysis, included adults at least age 50 years (2, 3). Although the Centers for Disease Control provided states with English and Spanish versions of the survey in 2006, not all states administer the survey in Spanish. States that had data on fewer than 50 surveys completed in Spanish were excluded and given differences in health care delivery between the U.S. territories and the states, the territories were excluded.

Our dependent variable was the reported receipt of CRC screening tests. We operationalized this variable by using BRFSS questions regarding FOBT and lower endoscopic procedures and testing. Respondents were considered to have been tested for CRC if they reported completing FOBT testing within the past 1 year or lower endoscopy within the past 10 years.

Our main independent variable was preferred language of survey respondent, English versus Spanish. We classified participants as “responding-in-English” if the corresponding survey was coded in the BRFSS data set as being conducted in English or “responding-in-Spanish” if the survey was coded as being conducted in Spanish. The BRFSS provides no information about the level of English proficiency, so this variable remained dichotomous. Respondent language was further stratified based on Hispanic/Latino race/ethnicity. The main independent variable was thus divided into three categories of exposure: non-Latinos responding-in-English (non-Latinos), Latinos responding-in-English, and Latinos responding-in-Spanish. There were insufficient data from non-Latinos responding-in-Spanish to create a fourth exposure category. Entries that were missing a language identifier or Hispanic/Latino status were excluded from the analysis.

Demographic characteristics for the study population included age, gender, marital status, health insurance status, and geographic region. Race and ethnicity were incorporated into the exposure variable. Education and income were combined to define an indicator variable for low socioeconomic status (SES). Low SES was defined as less than a 12^th grade education and/or annual income of less than $15,000 (14). In addition to these factors, we included as potential confounders the presence of an identified health care provider, smoking status, and respondent's perceived general health.

Respondent characteristics were calculated using standard means for continuous variables and proportions/frequencies for categorical variables. χ² tests were used to examine the relationships between the outcome of interest, receipt of CRC screening tests, and ethnicity/language category as well as each potential confounder. Logistic regression was used to estimate crude odds ratios (OR) between the three ethnicity/language categories and the receipt of CRC screening tests and to calculate crude ORs between potential confounder variables and receipt of CRC screening tests. Variables were considered to be confounders and, hence, included in a final multivariable logistic model, if the OR for the ethnicity/language variable adjusted for each potential confounder resulted in at least a 10% difference from the crude unadjusted OR. Additional variables were retained in the multivariable model if their inclusion was supported by prior research (6, 15, 16). The data were analyzed using SUDAAN version 9 (Research Triangle Institute). Sampling weights were included in all analysis. These sampling weights take into account the disproportionate stratified sampling design of the BRFSS. Additionally, the weights are adjusted poststratification to accommodate nonresponse and noncoverage within the sample (13).

The final study population of 99,895 represented individuals age ≥50 years old in 23 states. The majority (n = 94,346; 94.4%) of the study population was composed of non-Latinos who completed the survey in English. The remainder of the population consisted of Latinos who responded in English (n = 3,660; 3.7%) and Latinos who responded in Spanish (n = 1,889; 1.9%). A greater percentage of non-Latinos had health insurance, a health care provider, and higher education and income levels compared with Latinos responding-in-English, who in turn had greater percentages of these characteristics compared with Latinos responding-in-Spanish (Table 1.).

Of the total study population, 59.5% reported having received FOBT within the last year and/or lower endoscopy within 10 years. There were no significant differences in reported test receipt among groups of non-Hispanic/Latino race/ethnicity subgroups, so these groups, White non-Hispanic, Black non-Hispanic, other non-Hispanic, and mixed non-Hispanic, were collapsed and categorized as non-Latino. Overall, 61.6% of non-Latinos versus 43.6% of Latinos reported having received at least one screening test. In the adjusted model, compared with non-Latinos, Latinos as a group were less likely to report having received CRC screening tests, either FOBT and/or lower endoscopy [adjusted OR, 0.74; 95% confidence interval (CI), 0.65-0.85].

When divided into the 3 exposure categories to account for language and Hispanic/Latino ethnicity, 33.4% of Latinos responding-in-Spanish reported having CRC screening tests, compared with 51.0% of Latinos responding-in-English, and 61.6% of non-Latinos. In the adjusted model, both Latinos responding-in-English (OR, 0.84; 95% CI, 0.73-0.98) and Latinos responding-in-Spanish (OR, 0.57; 95% CI, 0.44-0.74) were less likely to report receiving CRC screening tests compared with non-Latinos. Latinos responding-in-Spanish were 36% less likely than Latinos responding-in-English to report having been screened (OR, 0.64; 95% CI, 0.48-0.84).

We also explored the effect of language and Latino/Hispanic ethnicity on CRC tests use by examining the relationship among respondents who reported having medical insurance and an identified health care provider. Among those with a health care provider and medical insurance, Latinos responding-in-English (OR, 0.83; 95% CI, 0.71-0.98) and Latinos responding-in-Spanish (OR, 0.56; 95% CI, 0.41-0.75) were less likely to report test use compared with non-Latinos. These OR results are very similar to the previously stated full model analysis. We further examined the effect of education and income on screening by stratifying on SES. In the low SES strata, compared with non-Latinos, Latinos responding-in-English (OR, 0.70; 95% CI, 0.54-0.91) and Latinos responding-in-Spanish (OR, 0.50; 95% CI, 0.39-0.65) were again less likely to report CRC test use. In the higher SES strata, although no longer statistically significant among Latinos responding-in-English (OR, 0.84; 95% CI, 0.70-1.0), Latinos responding-in-Spanish remained less likely to report screening compared with non-Latinos (OR,0.39; 95% CI, 0.24-0.64) (Table 2).

Despite recent increases in CRC screening nationally (15, 17), screening rates remain low, particularly among Latinos. Understanding disparities in CRC screening are especially important, given that Latinos and other ethnic minority groups are more likely than Whites to be diagnosed with CRC in more advanced stages and have higher mortality rates (18). In this population-based study of American households, not only were Latinos significantly less likely to report having CRC screening tests than non-Latinos, but when language was considered, Latinos who responded to the BRFSS survey in Spanish were significantly less likely to report CRC screening test use than Latinos who responded in English.

In an analysis of the 2001 California Health Interview Survey, Etzioni (16) found that among adults age <65 years, Hispanics were the only ethnic group less likely than Whites to have received recent CRC testing. However, in a multivariate model, those with limited English proficiency did not have lower rates of self-reported testing. A potential for reason for the discordant results between this study and the present study is that Etzioni et al. (16), included an assessment of acculturation based on English proficiency and the percent of years lived in the United States, which may partially explain the affect of language on CRC screening (19).

Several limitations of this study are noteworthy. First, we categorized Latinos as responding-in-English versus responding-in-Spanish based on respondents' language in the BRFSS. Although this may represent respondents' language preference, respondents who answered in Spanish may also be proficient in English and vice versa. More formal measures of English proficiency were not used in the 2006 BRFSS. The definition of the screening outcome variable is also a potential limitation. We relied upon self-report and considered those who had reported FOBT within the past year or lower endoscopy within the past 10 years as having been screened for CRC. Although current screening guidelines suggest intervals of every 5 years for sigmoidoscopy and every 10 years for colonoscopy, the BRFSS assesses the use of sigmoidoscopy and colonoscopy together in the same questions. We therefore included both sigmoidoscopy and colonoscopy testing done within the previous 10 years as part of the screening outcome variable. Although this measurement error may overestimate the true screening rates, they are likely to affect all three exposure groups equally and therefore would bias our results toward the null. Finally, as a telephone survey of noninstitutionalized adults, the survey may not be representative of those of lower SES who may not have telephones and of those who primarily use cellular telephones. Of particular concern with the BRFSS is the finding described by Link et al. (20) who used race, ethnicity, and language variables from the 2000 U.S. Census data with 2003 BRFSS data and estimated that counties with larger percentages of Spanish-only speakers had lower BRFSS participation rates compared with counties in which all respondents spoke English. The BRFSS data set does not provide response rate by race/ethnicity or language so we could not assess the effect of potential differential response rates based on these variables.

The results of the present study suggest a negative relationship with screening between Spanish language respondents and English language respondents implying that non- or limited-English speaking Latinos are at greater risk of not being screened for CRC. Limited English proficiency has been implicated as a barrier to medical comprehension (19), which may be particularly important given the multiple CRC screening options, the relatively complicated required preparation, and the invasiveness of lower endoscopy compared with other commonly performed cancer screening tests. The mechanism by which Spanish language use for survey responses is associated with less screening indicates that it may be a proxy measure for acculturation and/or may simply represent a communication barrier. Although the BRFSS does not include further analysis of respondents' level of acculturation, language itself is a major component of most acculturation scales (19). On the other hand, using data from the Hispanic Health and Nutrition Examination Survey, Solis et al. (21) suggest that the effect of language on screening practices should be considered as a factor related to access to care as opposed to a cultural factor. Regardless of the mechanism, cultural and language differences between patients and providers contribute to poorer health communications (22). When counseling patients about CRC screening and other complex medical decisions, providers should recognize that limited English proficiency may be a marker for lower acculturation, lower health literacy, and lower access to health care. Barriers to effective health communication may be overcome with relatively simple interventions such as culturally and linguistically appropriate telephone counseling targeting non-English speaking populations, which have been shown to improve CRC screening rates (23). In addition, adoption of the Department of Health and Human Services' National Standards for Culturally and Linguistically Appropriate Services in Health Care would help minimize cultural and linguistic barriers to health care and CRC screening (19, 24).

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.