Adjusted proportions of participants reporting up-to-date colorectal cancer screening by race/ethnicity, adjusted for age, sex, metropolitan statistical area residence, and region (A) plus educational level, annual income, insurance status, availability of usual source of care, self-rated health, language spoken at home, and nativity (B). FOBT indicates fecal occult blood testing. Error bars represent SD.
Jerant AF, Fenton JJ, Franks P. Determinants of Racial/Ethnic Colorectal Cancer Screening Disparities. Arch Intern Med. 2008;168(12):1317-1324. doi:10.1001/archinte.168.12.1317
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
The contributions of demographic, socioeconomic, access, language, and nativity factors to racial/ethnic colorectal cancer (CRC) screening disparities are uncertain.
Using linked data from 22 973 respondents to the 2001-2005 Medical Expenditure Panel Survey and the 2000-2004 National Health Interview Survey, we modeled disparities in CRC screening (fecal occult blood testing [FOBT], endoscopy, and combined FOBT and endoscopy) between non-Hispanic whites and Asians, blacks, and Hispanics, sequentially adjusting for demographics, socioeconomic status, clinical and access variables, and race/ethnicity–related variables (language spoken at home and nativity).
With demographic adjustment, minorities reported less CRC screening (all measures) than non-Hispanic whites. Disparities were largest for combined screening in Asians (adjusted odds ratio [AOR], 0.40; 95% confidence interval [CI], 0.32-0.49) and Hispanics (AOR, 0.43; 95% CI, 0.39-0.48) and for endoscopic screening in Asians (AOR, 0.41; 95% CI, 0.33-0.50) and Hispanics (AOR, 0.43; 95% CI, 0.38-0.48). With full adjustment, all Hispanic/non-Hispanic white disparities and black/non-Hispanic white FOBT disparities were eliminated, whereas Asian/non-Hispanic white disparities remained significant (FOBT: AOR, 0.72 [95% CI, 0.52-1.00]; endoscopic screening: AOR, 0.63 [95% CI, 0.49-0.81]; and combined screening: AOR, 0.66 [95% CI, 0.52-0.84]).
Determinants of racial/ethnic CRC screening disparities vary among minority groups, suggesting the need for different interventions to mitigate those disparities. Whereas socioeconomic, access, and language barriers seem to drive the CRC screening disparities experienced by blacks and Hispanics, additional factors may exacerbate the disparities experienced by Asians.
Colorectal cancer (CRC) screening uptake is suboptimal, lagging behind other evidence-based cancer screening tests.1,2 In the 2004 Behavioral Risk Factor Surveillance System, only 57% of adults 50 years or older overall reported up-to-date CRC screening status.1 Marked disparities in CRC screening also seem to exist nationally between non-Hispanic whites and other racial/ethnic groups in the United States.3- 12 For example, an analysis of pooled 1987-2003 National Health Interview Survey (NHIS) data found that up-to-date CRC screening was reported by 47% of white men and 44% of white women but in only 30.4% of Hispanic men and 31% of Hispanic women, in 43% of black men and 38% of black women, and in 29% of men and 32% of women of other ethnicity (including Asian).6 Such disparities in screening may contribute to the higher CRC incidence13,14 and mortality14- 16 rates observed in racial/ethnic minorities relative to non-Hispanic whites.
Previous studies have suggested several explanations for minority/non-Hispanic white screening disparities, including relatively lower socioeconomic status,5,8- 12,17- 23 reduced access to care,5,8,10,11,17,18,20- 23 and language or acculturation barriers.4,20,22,23 However, previous studies8- 12,17- 25 often excluded 1 or more of the 3 largest US racial/ethnic minority groups (Asians, blacks, and Hispanics), making comparisons of the importance of the various factors that affect CRC screening behavior across racial/ethnic groups difficult. Many studies were also confined to individuals in a single geographic region21- 25 or did not adjust for 1 or more known correlates of cancer screening behavior.4,5,8- 12,18,19,22- 24
To address these limitations in the literature, we examined the correlates of CRC screening among all 4 major US racial/ethnic categories (non-Hispanic white, Asian, black, and Hispanic individuals) using linked data from the 2001-2005 Medical Expenditure Panel Survey (MEPS) and the 2000-2004 NHIS. By combining 5 panels of MEPS data, we derived a large enough nationally representative sample (>22 000 individuals) to compare CRC screening up-to-date rates among major US racial/ethnic subgroups. We constructed 4 models, adjusting progressively for age, sex, panel, and region of the United States (basic demographics [model 1]), additionally for socioeconomic status (model 2), additionally for access to care and self-rated health (model 3), and, finally, additionally for race/ethnicity–related factors (language spoken at home and nativity [model 4]).
The primary source of data was the 2001-2005 MEPS,26 a nationally representative survey of health care use and costs in the US civilian, noninstitutionalized population conducted by the Agency for Healthcare Research and Quality. It uses an overlapping panel design and oversamples Hispanics and blacks. Data are collected for individuals during a 2-year period via a baseline interview and 5 follow-up interviews. The MEPS Household Component collects information on language spoken at home, country of origin, sociodemographic information, usual source of care, and health insurance coverage. The MEPS Household Component sample is drawn from a subsample of households included in the previous year's NHIS, an annual in-person household survey conducted by the National Center for Health Statistics.27 The NHIS collects information on several variables not included in MEPS that we expected might affect racial/ethnic CRC screening disparities: years of residence in the United States, citizenship status, and birth status (continental United States or elsewhere). Thus, we linked data for these variables in the 2000-2004 NHIS to the MEPS data. The MEPS point-in-time response rates for the 5 panels of public use data that we used were as follows: 2000, 70.5%; 2001, 71.4%; 2002, 69.2%; 2003, 68.9%; 2004, 68.2%; and 2005, 66.5%.
The MEPS respondents self-identify their racial category (white, black, Asian, Hawaiian native or other Pacific Islander, American Indian or Alaskan native, or multiple races) and ethnicity (Hispanic or non-Hispanic). Responses to race and ethnicity questions are crossed to derive combined race/ethnicity categories (eg, non-Hispanic white). The analyses in this article focus on adults 50 years or older classified in the MEPS as non-Hispanic white, Asian, black, or Hispanic.
The MEPS respondents were asked whether they had ever undergone fecal occult blood testing (FOBT) or “flexible sigmoidoscopy or colonoscopy” (endoscopic screening, a single item) and, if so, the interval (in the past year; 2, 3, or 5 years ago; or >5 years ago). In these analyses, respondents were considered to be up-to-date for screening if they reported FOBT in the previous 2 years (based on randomized controlled trial data28) or endoscopic testing at any time (because there are no evidenced-based intervals for these tests).2
The sociodemographic variables examined in the analyses were age (categorized as 50-54, 55-59, 60-64, 65-74, and ≥75 years), sex, rurality (living in a metropolitan statistical area [MSA] or not), household annual income level (<100%, 100%-124%, 125%-199%, 200%-399%, or ≥400% of the federal poverty level), educational attainment (less than high school, some high school, high school graduate, some college, or college graduate), geographic region (Northeast, Midwest, South, or West), and panel year.
The access-to-care variables examined were insurance status (private, public, or uninsured) and having a usual source of health care (yes or no). Self-rated health was assessed using a 5-point Likert scale response to the question, “In general, would you say your health was excellent, very good, good, fair, or poor?”
Respondents noted whether the language spoken at home was English or another language.
The NHIS respondents indicated whether they were born in the continental United States. Additional analyses explored length of time in the United States and citizenship, but the results are not reported herein because of significant proportions of missing data. Furthermore, analyses of nonmissing cases revealed that citizenship and length of time in the United States made no significant contribution beyond language and nativity.
Data were analyzed using Stata version 10.0 (Stata Corp, College Station, Texas), adjusting for the complex survey design of the MEPS. Analyses incorporated the longitudinal strata and primary sampling units and were weighted to yield appropriate standard errors and estimates representative of the US civilian, noninstitutionalized adult population.
We constructed 3 sets of analyses with 4 sequential logistic regression models to determine the relationship between CRC screening and race/ethnicity (non-Hispanic white, Asian, black, and Hispanic) using CRC screening as the dependent variable in all the models. The 3 sets of analyses in each sequential logistic regression model examined up-to-date rates for the 2 studied CRC screening modalities combined (endoscopy and FOBT), for endoscopy alone, and for FOBT alone. The first model in each set (model 1) examined the relationship between CRC screening and race/ethnicity, adjusting only for basic demographics (age, sex, MSA residence, and region of the United States) and survey year. Model 2 adjusted additionally for socioeconomic status (annual income and educational level). Model 3 then included adjustment for access to care (insurance status and availability of a usual source of care) and self-rated health, and model 4 included race/ethnicity–related factors (language spoken at home and nativity). We also examined interactions between racial/ethnic group and the other covariates, but none were statistically significant, and they are not reported herein. Because adjusted odds ratios can be misleading when the prevalence of the outcome is greater than 10%, we also report adjusted percentages of persons screened by ethnic group.29
Complete data were available for 95.3% of 22 973 eligible participants. Incomplete data were more common in minorities, for whom complete response rates varied from 90.3% for blacks to 93.0% for Hispanics (Table 1). The unadjusted combined (FOBT and endoscopy) CRC screening rate was 54.1% and was significantly lower in minorities, especially Asians and Hispanics. The same general pattern was observed for FOBT and endoscopic screening considered separately. Colorectal cancer screening for both studied modalities combined and for endoscopic screening alone increased throughout the study, whereas for FOBT it declined (data not shown). Women were less likely than men to report being screened for CRC by means of FOBT or endoscopy.
Table 1 provides the distribution of the variables examined by racial/ethnic group. Compared with non-Hispanic whites, Hispanics were less likely to be 75 years or older, had less education, had a lower annual income, had less private insurance, were less likely to have a usual source of care, had poorer self-rated health, were less likely to speak English at home, were less likely to be born in the United States, and were more likely to reside in MSAs. Patterns were similar for other minority groups, with some notable differences: compared with non-Hispanic whites, blacks were as likely to speak English at home and to be born in the continental United States, whereas Asians were less likely to be women and had more persons with less than 9 years and greater than 16 years of education.
Adjusted odds ratios for the logistic regression models are given in Table 2 (both CRC screening modalities combined), Table 3 (endoscopic screening), and Table 4 (FOBT). After adjustment for age, sex, MSA residence, region, and year, compared with non-Hispanic whites, all minorities, and especially Asians and Hispanics, were significantly less likely to report up-to-date CRC screening (model 1 in Tables 2-4). This disparity was more marked for endoscopic screening and was absent for blacks and FOBT. Although the disparity was most marked for Asians in model 1, the difference between Hispanics and Asians was not statistically significant in this model (results not shown).
With additional socioeconomic adjustment, the disparities in CRC screening were attenuated for Hispanics and blacks (and, for blacks, eliminated for FOBT) relative to non-Hispanic whites, but there was little change in Asian/non-Hispanic white disparities (model 2 in Tables 2-4). The gradient in screening was steepest for educational level, with the most educated group (≥16 years of schooling) having adjusted odds ratios of greater than 2.00 relative to the least educated group (<9 years of schooling).
Adjustment for access and self-rated health further attenuated Hispanic/non-Hispanic white screening disparities but had little effect on Asian/non-Hispanic white disparities (model 3 in Tables 2-4). For FOBT, blacks were more likely to report being up-to-date than were non-Hispanic whites. Apart from their effects on Hispanic/non-Hispanic white disparities, addition of the access and self-rated health variables had little effect on the CRC screening odds ratios for the other minority groups. Those with worse self-rated health, availability of some insurance, and a usual source of care were more likely to report screening.
Finally, with the inclusion of language and nativity, Hispanic/non-Hispanic white disparities were attenuated such that they were no longer statistically significant, whereas Asian/non-Hispanic white disparities were attenuated but remained significant (model 4 in Tables 2-4). Speaking English at home and being born in the continental United States were associated with greater CRC screening. Analyses that included language and nativity separately suggested that language, rather than nativity, was the main driver of the attenuation in disparities in model 4 (data not shown but available on request from the authors).
The Figure illustrates how the odds ratios for race/ethnicity, adjusted for demographics only (model 1) and for demographics, socioeconomic status, access to care, self-rated health, language spoken at home, and nativity (model 4), translate into adjusted proportions of racial/ethnic groups screened. It underscores that Asian/non-Hispanic white CRC screening disparities remained clinically and statistically significant for all 3 study screening outcomes after full adjustment (model 4) (Figure). In contrast, after full adjustment, black/non-Hispanic disparities in CRC screening persisted only for FOBT (favoring blacks) and endoscopy (favoring non-Hispanic whites), whereas Hispanic/non-Hispanic white disparities were eliminated for all 3 screening outcomes.
The present analyses overcome some of the limitations of previous studies, including the exclusion of 1 or more large minority groups,8- 12,17- 25 the use of nonnational samples,21- 25 and the lack of adjustment for key correlates of cancer screening behavior,4,5,8- 12,18,19,22- 24 and demonstrate the differing contributions of demographics, socioeconomic factors, access and self-rated health, language, and nativity to CRC screening disparities that affect Asian, black, and Hispanic individuals in the United States.
We verified that statistically significant disparities in CRC screening exist for each of these racial/ethnic minority groups relative to non-Hispanic whites. The initial analyses, adjusted only for basic demographics, revealed that disparities in combined CRC screening (FOBT and endoscopy) are more marked for Hispanic and, especially, Asian individuals than for blacks, relative to non-Hispanic whites. A variety of previous national studies4- 8,10,11,20 have suggested a substantial Hispanic/non-Hispanic white CRC screening disparity, but few studies4,5,22 have explored an Asian/non-Hispanic white disparity. The present findings call attention to a marked Asian/non-Hispanic white CRC screening disparity and suggest that federal public health and research initiatives aimed at increasing screening uptake in Asians should be implemented to complement those targeted to blacks and Hispanics.30
We further found that after sequential adjustment for an array of key correlates of CRC screening behavior—basic demographics, socioeconomic variables, access and self-rated health, and language spoken at home and nativity—black/non-Hispanic white and Hispanic/non-Hispanic white disparities in combined CRC screening were eliminated. Beyond socioeconomic factors, which disproportionately affect minorities, these findings suggest the effect of access and, for Hispanics, language-appropriate care on CRC screening uptake. In contrast, after full adjustment in these models, Asian/non-Hispanic white disparities in combined CRC screening remained statistically significant. Although this study design does not permit firm conclusions regarding the reason for this finding, the implication is that unmeasured cultural factors may contribute to the Asian/non-Hispanic white disparity in CRC screening. Less acculturated Asian individuals in the United States may have core health beliefs and values that differ from those in the “Western” health model, leading them to decline FOBT or endoscopy offered in the absence of worrisome symptoms.31 They may also be less likely to be offered CRC screening. Thus, culturally targeted interventions (focused on physicians and patients) might help address the Asian/non-Hispanic white CRC screening disparity. In contrast, enhancing access to health care might help mitigate black/non-Hispanic white disparities, whereas maximizing access to and linguistically appropriate provision of health care and information might help mitigate Hispanic/non-Hispanic white disparities.
The minority/non-Hispanic white disparities we observed were more pronounced for endoscopy than for FOBT. This finding suggests a current racial/ethnic technology diffusion gap32,33 and the potential for widening minority/non-Hispanic white CRC screening disparities across time because the primary driver of the recent secular increase in CRC screening in the United States has been the greater use of colonoscopy (with a concomitant decline in FOBT screening).1 Increasing access to and insurance coverage of colonoscopy in minority groups could narrow this gap but may be difficult to achieve given that access to colonoscopy is suboptimal even in non-Hispanic whites.34 It may be more feasible to increase uptake of FOBT in minorities, a much less expensive and more accessible test than colonoscopy34 and, to our knowledge, the only screening modality shown in randomized controlled trials to reduce CRC mortality rates.28 Quantitative FOBT holds particular promise for reducing ethnic/minority disparities because the positive cutoff point can be adjusted to match the risk of CRC in the targeted population.35,36
Finally, we found lower CRC screening rates for women than for men across all ethnic/minority groups, screening modalities, and analytic models. Although the reasons for this finding are unclear, specific efforts to engage women of all ethnicities and races in CRC screening seem to be justified.
This study has some limitations. First, FOBT and endoscopy are self-reported by MEPS respondents. Previous research37,38 suggests the potential for overreporting of screening by minorities, probably because of a social desirability effect. This may have contributed to the apparent black/non-Hispanic white FOBT disparity favoring blacks. To the degree that such overreporting of CRC screening may have occurred in minorities in this sample, these findings may underestimate minority/non-Hispanic white disparities in screening. It is also not possible to distinguish in the MEPS whether self-reported colonoscopy was for screening or diagnosis. Thus, the apparent minority/non-Hispanic white technology diffusion gap we observed could be because of less access to screening endoscopy for minorities, less access to follow-up endoscopy (eg, after abnormal FOBT results), or a combination. Regardless, such a technology diffusion gap may contribute, along with the overall CRC screening disparities presented herein, to the increased CRC incidence13,14 and mortality13,14 rates for minorities relative to non-Hispanic whites that are observed in some studies.
An additional limitation is that the MEPS data set we used in these analyses includes only a simple dichotomized language variable (English or another language spoken at home). Yet, many US Hispanic and Asian adults are fluently bilingual and might speak English and another language, depending on the context. For example, some Hispanic individuals may regularly speak Spanish when conversing with parents or spouses but may speak English with their children. Because the MEPS and the NHIS do not include questions to ascertain multilingualism, whether and how it might affect these findings are unknown. However, we did find that the dichotomous language variable is significantly associated with CRC screening. Finally, we used 3 large, composite, racial/ethnic groupings in the analyses to allow adequate power to explore minority/non-Hispanic white disparities in CRC screening. However, each of the categories encompasses various national origin and immigration cohorts, among whom significant differences in CRC screening may exist.23,39
In conclusion, disparities in CRC screening exist for each of the 3 largest racial/ethnic minority groups (Asians, blacks, and Hispanics) relative to non-Hispanic whites in the United States. The disparities seem to be most marked for Asians, a group that has previously attracted little investigation. Furthermore, the underlying determinants of racial/ethnic disparities in CRC screening seem to differ in each minority group, in turn implying the need for different strategies to mitigate the CRC screening disparities. Finally, there is evidence that current racial/ethnic CRC screening disparities are larger for endoscopic screening than for FOBT, suggesting a technology diffusion gap that could contribute to widening disparities across time,32,33 and for women relative to men. Each of these findings begins to provide needed direction to those seeking to develop and implement interventions to eliminate racial/ethnic CRC screening disparities in the United States.
Correspondence: Anthony F. Jerant, MD, Department of Family and Community Medicine, University of California Davis School of Medicine, 4860 Y St, Ste 2300, Sacramento, CA 95817 (firstname.lastname@example.org).
Accepted for Publication: January 13, 2008.
Author Contributions: Dr Jerant had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. Study concept and design: Jerant, Fenton, and Franks. Acquisition of data: Franks. Analysis and interpretation of data: Jerant, Fenton, and Franks. Drafting of the manuscript: Jerant and Franks. Critical revision of the manuscript for important intellectual content: Jerant, Fenton, and Franks. Statistical analysis: Franks. Administrative, technical, and material support: Franks. Study supervision: Jerant and Franks.
Financial Disclosure: None reported.