Factors Explaining Racial/Ethnic Disparities in Rates of Physician Recommendation for Colorectal Cancer Screening

Abstract

Objectives. Physician recommendation plays a crucial role in receiving endoscopic screening for colorectal cancer (CRC). This study explored factors associated with racial/ethnic differences in rates of screening recommendation.

Methods. Data on 5900 adults eligible for endoscopic screening were obtained from the National Health Interview Survey. Odds ratios of receiving an endoscopy recommendation were calculated for selected variables. Planned, sequenced logistic regressions were conducted to examine the extent to which socioeconomic and health care variables account for racial/ethnic disparities in recommendation rates.

Results. Differential rates were observed for CRC screening and screening recommendations among racial/ethnic groups. Compared with Whites, Hispanics were 34% less likely (P < .01) and Blacks were 26% less likely (P < .05) to receive this recommendation. The main predictors that emerged in sequenced analysis were education for Hispanics and Blacks and income for Blacks. After accounting for the effects of usual source of care, insurance coverage, and education, the disparity reduced and became statistically insignificant.

Conclusions. Socioeconomic status and access to health care may explain major racial/ethnic disparities in CRC screening recommendation rates.

Colorectal cancer (CRC) is the third most common cause of cancer death in the United States and a major cause of cancer morbidity.¹ Annually in the United States, an estimated 150 000 people are diagnosed and 50 000 people die from this disease.² With early detection and removal of precancerous polyps, CRC can be preventable, with a 5-year survival rate as high as 90%.^3,4

Current CRC screening guidelines consist of sigmoidoscopy every 5 years and colonoscopy every 10 years for those aged 50 to 75 years.⁴ Although an annual fecal occult blood test is also recommended, sigmoidoscopy and colonoscopy have higher sensitivity and specificity for the detection of cancerous lesions.⁵ Sensitivity ranges from 25% to 87% for fecal occult blood test, compared with 92% to 95% for endoscopy (sigmoidoscopy and colonoscopy).⁶ Colonoscopy, the most widely used test for CRC detection and prevention, is considered the gold standard because it can detect and remove precancerous polyps, and a positive result from any other screening test must be followed by this advanced modality.^1,7,8 Endoscopy has played a major role in the decreasing trend of CRC mortality in the United States^9,10

Still, minority groups—especially Blacks—carry a disproportionately higher CRC burden.^1,11 From the 1980s to 2007—a period of advance in early detection and treatment—the inequality between Blacks and Whites increased to a 44% difference in CRC mortality rates.¹ This disparity could be attributable in part to the fact that, compared with Whites, members of racial/ethnic minority groups were less likely to be screened and diagnosed at the localized, more treatable stage of CRC.¹ These lower screening rates partially account for the higher rates of late stage detection and subsequent increases in CRC morbidity and mortality rates among racial/ethnic minorities and low-income populations.^1,10,12 Thus, it is necessary to promote utilization of effective screening methods, such as endoscopy, to detect CRC in its early stages, especially among minority groups.

Trends in CRC screening from 1992 to 2005 show widening gaps between Whites and minority groups: the disparity increased to 10.3% for Blacks and to 20.5% for Hispanics.^13,14 Lower rates of cancer screening, including CRC screening, have been strongly associated with lower rates of physician recommendation, particularly among low-income populations, racial/ethnic minority groups, and women.^15,16

The health care provider’s role in CRC screening is essential because a physician recommendation is necessary for endoscopic screening.^17,18 A study found that more than 90% of people who did not undergo endoscopic screening reported not receiving the necessary recommendation.¹⁶ A systematic review discovered that patients identified physician recommendation as the sole CRC-screening facilitator and absence of recommendation as the only barrier.¹⁹ It is thus crucial to identify factors that influence the likelihood of receiving a CRC screening recommendation.

The present study seeks to build on the work of 2 previous analyses of CRC-screening barriers and predictors among respondents to the 2000 National Health Interview Survey (NHIS). Seeff et al. found that “frequent doctor visits in the past year” was the strongest predictor of CRC screening.²⁰ Among the barriers, a “lack of awareness of the need to be screened” was most common, followed by “not receiving a physician recommendation.”²⁰ In bivariate analysis, Coughlin and Thompson found that, among screening-eligible adults who had visited a physician in the past year, members of racial/ethnic minority groups were less likely than Whites to receive an endoscopy recommendation.²¹ The study mainly focused on reasons for having CRC screenings.

We used multivariate analyses and planned sequenced logistic regression to explore patient factors that may explain unequal rates of physician recommendation for endoscopy among Blacks and Hispanics compared with Whites. Factors deemed to have relative importance in terms of statistical significance may help indicate areas of intervention to improve rates of physician recommendation for CRC screening in disadvantaged populations.

METHODS

We analyzed data from the 2000 edition of the Cancer Control Module of the NHIS, which contained detailed information about ethnicity and acculturation not present in subsequent modules. The NHIS is a cross-sectional, nationally representative survey of the noninstitutionalized US population, conducted by in-person interviews. It was developed over decades, progressively adapting its approaches to become more sensitive to differences in education, language, and culture. The sample was obtained through a stratified, multistage design, with oversampling of Black and Hispanic people. Details on NHIS 2000 methods are available to the public.^22,23 The response rate for the cancer module was 72%.

Sample

We used data on 12 477 adults aged 50 years or older. Responses “refused” and “did not know” were considered missing. We excluded 170 who had a history of CRC, 1447 who had not visited a physician in the past year, and 1131 with no information on CRC history or physician visit.

A total of 7709 had visited a physician in the past year and were eligible for endoscopy screening recommendation. Of those, 1149 had already received a recommendation and a sigmoidoscopy or colonoscopy screening in the past 12 months; 6144 had never received endoscopy, and 416 had undergone endoscopy at some point but were not current according to guidelines. From the eligible sample, 1809 with missing information regarding recommendation were excluded. The final sample for analysis consisted of 5900 individuals.

In multiple logistic regression analyses, the sample size varied from model to model depending on missing values in variables, number of variables in the equation, and use of listwise deletion option. These small changes in sample size were unlikely to affect the magnitude of significance of the covariates.

Variables

We developed a dichotomous outcome variable based on a question from the aforementioned national survey: “In the PAST YEAR, has a doctor or other health professional RECOMMENDED that you have a sigmoidoscopy or colonoscopy?” The answer was categorized as yes or no. Those with responses of “refused” or “don’t know” were excluded.

We chose independent variables based on literature review^24–32 and hypothesized the existence of an association with physician recommendation for CRC screening. Sociodemographic characteristics were race/ethnicity, age, gender, marital status, citizenship, education, occupation, income, community size, and geographic region. Health care–related variables were first-degree family history of CRC, insurance status, usual source of care, and number of physician visits in the past 12 months.

The key variables were grouped into 2 simple thematic clusters: health care access (insurance and source of care); and health system interface (education and income) to understand likelihood of individual factors, as well as thematic or combined interplay in receiving a physician recommendation in relation to the extent of racial/ethnic disparity change.

Statistical Analysis

We generated analyses using Stata software (StataCorp LP, College Station, TX) designed to handle weighted data and complex sampling designs. The NCHS provided a national weight variable with public access NHIS data to address design and poststratification effects including oversampling, nonresponse, nonsampling error, missing values, and interviewer effect.²³ In consultation with NHIS experts, we generated all estimates of population proportions and associated standard errors as well as odds ratios and confidence intervals, using Stata 7.0, which uses Taylor series linearization for variance estimation using national weight. Multicollinearity was tested based on correlation coefficients among independent variables. No absolute value of the correlation coefficients were larger than 0.65 to create a serious multicollinearity problem.^33,34

We used logistic regression to calculate crude and adjusted odds ratios of endoscopy recommendation. Planned logistic regressions were performed by entering chosen thematic variables into the model, in sequence of increasing number of variables, in thematic clusters to examine each variable’s relative magnitude of likelihood of receiving a screening recommendation among Blacks and Hispanics compared with Whites. The final model included 14 variables: race/ethnicity, age, gender, marital status, citizenship, occupation, usual source of care, insurance status, income, education, community size, geographic region, family history of CRC, and number of physician visits in the past year.

RESULTS

Bivariate analysis of the sample’s sociodemographic characteristics and crude odds ratios of physician recommendation for screening are shown in Table 1. The majority of participants were insured, had a usual source of care, and had visited a doctor at least twice in the past year. Of the final sample, 26% received a recommendation and 75% of those actually received an endoscopy.

TABLE 1—

Sociodemographic Characteristics of Receiving Physician Recommendation for Endoscopic Screening Among Eligible United States Population: National Health Interview Survey, 2000

Variables	Total	Sample %	% Recommended	Crude OR (95% CI)
Age, y
50–54	1333	22.6	21.6	1.00 (Ref)
55–59	1027	17.4	26.0	1.27* (1.02, 1.59)
60–64	850	14.4	30.8	1.61*** (1.28, 2.03)
65–74	1463	24.8	31.6	1.67*** (1.35, 2.07)
≥ 75	1227	20.8	23.2	1.09 (0.87, 1.38)
Family history of CRC
No	5393	91.4	25.4	1.00 (Ref)
Yes	507	8.6	42.1	2.14*** (1.73, 2.64)
Gender
Male	2578	43.7	29.1	1.00 (Ref)
Female	3322	56.3	24.2	0.78** (0.68, 0.90)
Marital status
Unmarried	2124	36.0	28.5	1.00 (Ref)
Married	3776	64.0	22.4	1.38*** (1.21, 1.58)
Race/ethnicity
White	4785	81.1	27.7	1.00 (Ref)
Hispanic	360	6.1	20.2	0.66** (0.52, 0.85)
Black	531	9.0	22.1	0.74* (0.58, 0.94)
Other	165	2.8	17.3	0.55* (0.32, 0.93)
Citizenship
Noncitizen	148	2.5	12.7	1.00 (Ref)
Citizen	5753	97.5	26.7	2.51*** (1.55, 4.08)
Education
Non–high school graduate	1345	22.8	17.9	1.00 (Ref)
High school graduate	1929	32.7	24.8	1.52*** (1.24, 1.86)
Some college/associate degree	1339	22.7	28.5	1.84*** (1.51, 2.24)
College graduate	1286	21.8	36.5	2.65*** (2.13, 3.29)
Occupation
Managerial	513	8.7	29.3	1.00 (Ref)
Professional	1210	20.5	27.7	0.93 (0.70, 1.22)
Blue collar	1947	33.0	24.0	0.76* (0.59, 0.99)
Retired	1339	22.7	30.0	1.03 (0.81, 1.33)
Unemployed	897	15.2	23.6	0.75 (0.55, 1.01)
Household income, $
< 20 000	1817	30.8	20.9	1.00 (Ref)
20 000—64 999	2555	43.3	27.4	1.43*** (1.20, 1.70)
≥ 65 000	1534	26.0	34.4	1.99*** (1.62, 2.43)
Health insurance
Uninsured	5605	5.0	12.0	1.00 (Ref)
Insured	295	95.0	27.2	2.75*** (1.96, 3.85)
Usual source of care
No	206	3.5	11.5	1.00 (Ref)
Yes	5694	96.5	26.9	2.83*** (1.76, 4.56)
Doctor visits in past 12 mo
1	879	14.9	17.0	1.00 (Ref)
2–3	2862	48.5	25.3	1.65*** (1.31, 2.07)
≥ 4	2159	36.6	32.1	2.30*** (1.85, 2.86)
Community size
< 1 million	3398	57.6	24.0	1.03 (0.81, 1.33)
1–2.49 million	1416	24.0	29.6	1.38* (1.03, 1.84)
2.5–4.9 million	572	9.7	34.8	1.74*** (1.30, 2.33)
≥ 5 million	507	8.6	23.4	1.00 (Ref)
Region
Northeast	1221	20.7	29.3	1.00 (Ref)
Midwest	1457	24.7	25.1	0.81* (0.66, 0.98)
South	2130	36.1	23.6	0.74* (0.62, 0.90)
West	1092	18.5	30.3	1.05 (0.83, 1.33)

Note. CI = confidence interval; CRC = colorectal cancer; OR = odds ratio. The sample size was n = 5900.

*P < .05; **P < .01; ***P < .001.

In multivariate analyses (Table 2), endoscopy recommendation was significantly more likely among those who were aged 55 years and older, had a family history of CRC (odds ratio [OR] = 1.95), were married (OR = 1.30), had more education, had an income of $65 000 or more (OR = 1.36), had insurance (OR = 1.72), had a usual source of care (OR = 3.38), had more doctor visits, and lived in an area with a population of 2.5 to 4.9 million people (OR = 2.07). People living in the South had significantly lower rates of recommendation (OR = 0.67) compared with those in the Northeast.

TABLE 2—

Multivariate Adjusted Odds Ratios of Receiving Physician Recommendation for Colonoscopy and Sigmoidoscopy Screening Among Eligible United States Population Aged 50 Years and Older by Sociodemographic Characteristics: National Health Interview Survey, 2000

Characteristics	All participants (n = 4256), OR (95% CI)	Men (n = 1640), OR (95% CI)	Women (n = 2616), OR (95% CI)	Whites (n = 3069), OR (95% CI)	Blacks (n = 572), OR (95% CI)	Hispanics (n = 510), OR (95% CI)
Age, y
50–54 (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
55–59	1.40** (1.09, 1.79)	1.39 (0.93, 2.08)	1.38* (1.01, 1.92)	1.29 (0.97, 1.72)	1.72 (0.93, 3.16)	1.69 (0.67, 4.26)
60–64	2.18*** (1.59, 2.98)	2.00** (1.23, 3.26)	2.33*** (1.60, 3.41)	2.49*** (1.78, 3.49)	1.24 (0.50, 3.08)	1.70 (0.69, 4.18)
65–74	1.85*** (1.39, 2.47)	1.55* (1.01, 2.39)	2.25 *** (1.53, 3.30)	1.97*** (1.43, 2.70)	1.51 (0.59, 3.89)	1.24 (0.51, 2.98)
≥ 75	1.36* (1.01, 1.87)	1.30 (0.81, 2.29)	1.49* (0.99, 2.24)	1.36 (0.96, 1.92)	1.21 (0.52, 2.85)	0.98 (0.36, 2.65)
Family history of CRC
No (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
Yes	1.95*** (1.50, 2.54)	1.48 (0.96, 2.29)	2.35*** (1.69, 3.26)	2.10*** (1.56, 2.81)	0.82 (0.28, 2.36)	1.63 (0.59, 4.52)
Gender
Male (Ref)	1.00	…	…	1.00	1.00	1.00
Female	0.81* (0.68, 0.95)	…	…	0.85 (0.70, 1.03)	0.49** (0.30, 0.82)	0.60 (0.32, 1.11)
Marital status
Not married (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
Married	1.30** (1.08, 1.57)	1.59** (1.19, 2.12)	1.12 (0.89, 1.41)	1.27* (1.03, 1.56)	1.71* (0.99, 2.96)	1.25 (0.69, 2.28)
Race/ethnicity
White (Ref)	1.00	1.00	1.00	…	…	…
Hispanic	0.98 (0.72, 1.32)	0.97 (0.60, 1.56)	1.04 (0.65, 1.64)	…	…	…
Black	0.99 (0.73, 1.34)	1.32 (0.81, 2.17)	0.76 (0.54, 1.08)	…	…	…
Other	0.68 (0.36, 1.28)	0.60 (0.26, 1.37)	0.96 (0.43, 2.15)	…	…	…
Citizenship status
Noncitizen (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
Citizen	1.60 (0.90, 2.87)	1.21 (0.52, 2.86)	2.39* (1.03, 5.77)	1.35 (0.40, 4.53)	0.84 (0.22, 3.21)	1.27 (0.52, 3.06)
Region
Northeast	1.00	1.00	1.00	1.00	1.00	1.00
Midwest	0.80 (0.63, 1.02)	0.70 (0.48, 1.01)	0.93 (0.67, 1.30)	0.84 (0.64, 1.10)	0.62 (0.27, 1.41)	0.31 (0.05, 1.82)
South	0.67** (0.54, 0.84)	0.72 (0.51, 1.02)	0.65** (0.47, 0.89)	0.67** (0.51, 0.87)	0.47* (0.24, 0.96)	0.83 (0.36, 1.89)
West	1.00 (0.77, 1.30)	0.96 (0.64, 1.43)	1.01 (0.71, 1.44)	1.03 (0.76, 1.38)	0.50 (0.15, 1.62)	0.95 (0.45, 1.97)
Education
Non–high school graduate (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
High school graduate	1.57** (1.22, 2.03)	1.52 (0.98, 2.35)	1.74** (1.25, 2.42)	1.66** (1.23, 2.23)	0.93 (0.49, 1.76)	1.38 (0.60, 3.14)
Some college/associate degree	1.65*** (1.27, 2.14)	1.41 (0.94, 2.12)	1.97*** (1.36, 2.87)	1.70** (1.25, 2.30)	1.63 (0.84, 3.21)	1.49 (0.63, 3.53)
College graduate	2.48*** (1.83, 3.36)	2.48*** (1.58, 3.89)	2.61*** (1.71, 3.97)	2.76*** (1.97, 3.86)	1.05 (0.49, 2.22)	4.43* (1.21,16.24)
Occupation
Managerial (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
Unemployed	0.98 (0.66, 1.45)	0.73 (0.39, 1.35)	1.24 (0.76, 2.02)	1.12 (0.71, 1.73)	0.69 (0.27, 1.77)	0.28 (0.07, 1.14)
Professional	0.97 (0.69, 1.38)	0.88 (0.52, 1.51)	1.08 (0.70, 1.65)	0.93 (0.63, 1.38)	1.13 (0.46, 2.82)	1.89 (0.61, 5.87)
Blue collar	0.82 (0.59, 1.14)	0.85 (0.50, 1.42)	0.79 (0.52, 1.19)	0.85 (0.59, 1.23)	0.51 (0.21, 1.27)	1.35 (0.47, 3.89)
Retired	0.97 (0.70, 1.35)	1.06 (0.62, 1.82)	0.90 (0.59, 1.38)	1.02 (0.71, 1.45)	0.61 (0.27, 1.37)	1.16 (0.38, 3.53)
Annual income, $
< 20 000 (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
$20 000—$64 999	1.15 (0.93, 1.41)	0.94 (0.66, 1.33)	1.39* (1.08, 1.80)	1.12 (0.88, 1.42)	1.48 (0.83, 2.66)	0.97 (0.55, 1.72)
≥ $65 000	1.36* (1.04, 1.78)	1.10 (0.70, 1.72)	1.70** (1.21, 2.38)	1.36* (1.01, 1.85)	0.70 (0.30, 1.63)	1.75 (0.72, 4.21)
Health insurance
Uninsured (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
Insured	1.72** (1.16, 2.57)	2.50* (1.18, 5.32)	1.31 (0.78, 2.19)	1.76* (1.05, 2.95)	1.41 (0.53, 3.73)	3.28* (1.01,10.68)
Usual source of care
No (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
Yes	3.38*** (1.84, 6.21)	2.93* (1.28, 6.70)	3.81** (1.52, 9.58)	3.54*** (1.77, 7.09)	11.9* (1.38, 103.26)	0.70 (0.14, 3.38)
Doctor visits
1 (Ref)	1.00	1.00	1.00	1.00	1.00	1.00
2–3	1.76*** (1.34, 2.30)	1.54* (1.02, 2.34)	2.10*** (1.47, 2.99)	1.79*** (1.35, 2.39)	1.47 (0.54, 3.98)	2.54 (0.79, 8.19)
≥  4	2.38*** (1.85, 3.07)	2.34*** (1.61, 3.40)	2.66*** (1.86, 3.82)	2.29*** (1.72, 3.05)	2.43 (0.95, 6.12)	4.21* (1.22,14.56)
Community size
< 1 million	1.07 (0.77, 1.49)	1.00 (0.58, 1.70)	1.14 (0.75, 1.71)	1.24 (0.85, 1.80)	0.56 (0.23, 1.41)	1.16 (0.48, 2.81)
1–2.49 million	1.40 (0.99, 1.99)	1.33 (0.74, 2.39)	1.45 (0.94, 2.24)	1.64* (1.09, 2.47)	0.97 (0.40, 2.36)	1.53 (0.56, 4.23)
2.5–4.9 million	2.07*** (1.42, 3.01)	1.41 (0.79, 2.53)	2.66*** (1.64, 4.32)	2.63*** (1.72, 4.03)	0.84 (0.29, 2.45)	1.09 (0.29, 4.21
≥ 5 million (Ref)	1.00	1.00	1.00	1.00	1.00	1.00

Note. CI = confidence interval; CRC = colorectal cancer; OR = odds ratio.

*P < .05; **P < .01; ***P < .001.

Gender Disparity

In multivariate analysis, women were (OR = 0.81) less likely than men to receive a recommendation (Table 2). However, within each racial/ethnic group analysis, the disparity was statistically significant only for Black women, who were about half as likely as Black men to receive a recommendation. The likelihood of recommendation significantly increased for women with each increasing level of education; however, among men the increase was significant only for those with a college degree.

Racial/Ethnic Disparity

In bivariate analysis, with only race/ethnicity in the model, Hispanics were 34% (P < .01) and Blacks were 26% (P < .05) less likely than Whites to receive a screening recommendation. Figure 1 depicts planned logistic regression results sequenced in descending magnitude of disparity. The differences between ethnic minority groups compared with Whites remained statistically significant with the model accounting for the effect of usual source of care (Hispanics 33% less likely; Blacks 26% less likely). With the model considering insurance only, the differences decreased by 6 percentage points for Hispanics and by 2 percentage points for Blacks (28% and 24%, respectively). Usual source of care and insurance combined in 1 model gave similar results.

FIGURE 1—

Disparity in likelihood of receiving recommendation for endoscopic screening for colorectal cancer among US adults aged 50 years and older: National Health Interview Survey, 2000.

Note. educ. = education; SOC = source of care. Disparity is defined as the percentage point difference in screening recommendation compared with Whites. Multivariate results adjusted for usual source of care, insurance, income, age, and education. All models included age, gender, marital status, citizenship, occupation, number of doctor visits in past 12 months, community size, and region.

*P <.05; **P = <.01.

The model accounting for income showed a great reduction in disparities, with Hispanics 23% and Blacks 15% less likely than Whites to receive a physician recommendation. Controlling for education by itself reduced the disparity observed between Hispanics and Whites, and these differences ceased to be statistically significant for both Hispanics and Blacks. With the combination of insurance, education and usual source of care the decreasing trend continued. However, a greater effect was observed with the combination of income and education compared with insurance, education and usual source of care taken together.

When controlling for either income or education, the disparity for Blacks compared with Whites decreased to a nonsignificant level. This was also observed for Hispanics compared with Whites when controlling for education alone but not income alone. Multivariate adjustment—controlling for all variables in the model—further reduced screening-recommendation disparities among Hispanics to 2 percentage points and Blacks to1 percentage point compared with Whites, and these differences were statistically insignificant.

DISCUSSION

CRC is a major cause of cancer burden in the United States, and minorities suffer a disproportionate burden of this disease.^1,11 This disparity is due in part to lower screening rates and increased late-stage diagnosis among racial/ethnic minority groups.¹ Lower utilization rates of cancer screening, especially CRC screening, have been strongly associated with lower rates of physician recommendation.^{8,12,15,16,35}

Aligned with the NIH State of the Science Recommendation,⁸ this study’s objective was to identify factors that may influence the likelihood of receiving a physician recommendation for CRC screening. This study sequentially presents the relative importance of single, thematic socioeconomic status and health care variables and their combinations in explaining the extent of the observed disparity. We found that patients’ socioeconomic status and health care access variables are strongly associated with differences in provider’s recommendation by race/ethnicity. Moreover, the single factor that removed statistical differences between Whites and racial/ethnic minority groups was not the same for Blacks (education) and Hispanics (income). The findings of our study could be useful in developing programs to address ethnic-specific barriers to this screening.

Physicians play a central role in CRC screening. Our findings indicate that physicians accurately follow guidelines by considering age and family history of CRC—2 important risk factors.³⁶ However, in the present sample, slightly more than one quarter of eligible adults reported receiving an endoscopic screening recommendation, a noticeably low percentage. Also, socioeconomically disadvantaged members of racial/ethnic minority groups were less likely to receive a recommendation. This finding is consistent with previous research.^8,12,35

Although having insurance and a usual source of care reduced the racial/ethnic disparities in our study, Hispanics and Blacks remained significantly less likely to receive a recommendation than their White counterparts when only these variables were considered. However, the disparities for both Hispanics and Blacks decreased and became statistically insignificant when individual education level was considered. This finding is in line with another study that found patient’s education, but not insurance status, to be associated with CRC-screening recommendation.³⁷

The aforementioned study did not examine predictors for each racial/ethnic group separately. Our findings showed a slight difference in the impact of predictor variables between Blacks and Hispanics, compared with Whites. For Blacks, but not Hispanics, the inclusion of income alone in the model produced effects similar to those observed with the addition of education, reducing the observed disparity to a statistically insignificant level. In the full multivariate model, the combination of education and income achieved the greatest disparity reduction for both minority groups. These results mirror those of a previous study that found that racial/ethnic disparities in likelihood of mammography screening recommendation dissipated when researchers controlled for education and family income.³⁸

Uncovering education’s effect on racial/ethnic disparities in endoscopy recommendation may help explain where the communication gap between physicians and patients occurs. It seems unlikely that physicians would systematically and specifically not recommend CRC screening to members of racial/ethnic minority groups. The unadjusted model indicated a positive linear relationship between likelihood of receiving a recommendation and education, regardless of race/ethnicity. It may be that, in part because of a low educational level, some patients have difficulty understanding what the physician has said and do not realize a screening recommendation was given. A less-educated patient may also hesitate to raise questions or concerns. A previous study found that only 6.2% of physicians reported specifically assessing a patient’s understanding when giving a CRC screening recommendation.³⁹ It is also likely that less-educated patients might be uninformed about the need for CRC screening. Poor knowledge of CRC and screening has been identified by patients and physicians as an important barrier.⁴⁰

Education and a patient’s level of understanding play a role in physician-patient interaction. Those who are more educated are likely to be more aware of CRC’s potential risks and screening’s benefits. In addition, educated patients are more likely to effectively communicate concerns, which seems to increase the probability of patient-centered communication.⁴¹ This, in turn, enhances a patient’s chances of receiving and understanding preventive messages. Previous studies have found that providers reported regularly recommending CRC screening, and 95% of primary care physicians said they recommended colonoscopy for screening purposes.^42,43 The gap between physicians’ reported recommendation rates and patients’ reported lack of recommendation might be attributed to misunderstanding related to patient education levels.

Physicians’ incorrect perception of patients’ barriers may also affect communication of screening messages. In a previous study, a substantial proportion of primary care physicians (PCPs) identified patient issues such as embarrassment or anxiety (56% of PCPs) and cost or lack of insurance (46% of PCPs) as major barriers to CRC screening although less than 1% of an adult population reported these as main barriers.⁴⁰ Other research has shown that physicians perceive Blacks as having a lower level of CRC awareness,⁴⁴ less effective communication skills,²⁸ and lower positive-affect scores in medical conversations.⁴⁵ These factors may hinder the patient–physician relationship, which plays an important role in receiving a screening recommendation. Greater physician trust and a greater likelihood of having the same provider as a usual source of care have been found among White Americans compared with Black Americans, and these factors were positively associated with cancer screening.⁴⁶

An even lower level of quality in patient-physician interactions has been found among Hispanics and was influenced by the patient’s health literacy,⁴⁷ which can be linked to general education level. Educational tools featuring images to support explanations about CRC screening might be useful in increasing physicians’ efficacy when counseling less-educated patients.⁴⁸

Higher education often coincides with greater resources and, as expected, wealthier people were more likely to be referred for endoscopic screening. These findings are supported by a growing body of evidence establishing an association between CRC-screening recommendation and socioeconomic status.^49,50 Accordingly, CRC screening rates among the elderly have shown a significant increase from low-income to middle-income class.⁵¹ This disparity may be explained by the perception that wealthy patients are more able to afford and obtain screening.⁵²

Though income plays a role in the differences in screening recommendations, income alone cannot explain these disparities, especially for Hispanics, among whom education was the most important explaining factor in the present study. Previous research exploring the impact of Medicare coverage on CRC screening has demonstrated that, although insurance coverage reduced the disparity between Blacks and Whites, it increased the disparity between Hispanics and non-Hispanic Whites. This is probably because a higher proportion of Hispanics are in the poor or near poor income group and because a larger proportion of Hispanics are younger and therefore do not qualify for Medicare benefits.⁵¹

Studies on the relationship between CRC screening recommendation and insurance coverage have shown mixed results.^18,37 Patients with no health insurance and no usual source of care may be more likely to receive treatment of urgent health needs rather than preventive care, such as CRC screening. This pattern of health care use provides neither the regularity to build a strong physician–patient relationship nor the time and opportunity for physicians to discuss preventive care. Physicians have identified time as a barrier to screening recommendations.^19,53 Establishing rapport between a patient and a single provider fosters communication and increases the likelihood of screening recommendation.

In the present study, within each racial/ethnic group, a smaller proportion of women received recommendations, which is consistent with previous findings.⁵² This might be explained by the physicians’ knowledge of higher CRC incidence and mortality rates for men.^1,49 It is important to address this disparity because recent findings have shown that women have benefitted less from decreases in CRC incidence rates.¹ Also, after the age of 50 years, women are experiencing lower survival rates than men.^10,54 This may be attributable in part to lower screening rates, which may be associated with lower recommendation rates.

Strengths and Limitations

The use of nationally representative NHIS data allows for this study’s results to be generalized to the US population. Although other studies have shown differences in physician recommendation for CRC screening based on race/ethnicity, few have used a step-by-step, planned analytical model to investigate underlying factors that could account for this disparity.

The survey’s cross-sectional design does not permit temporal or causal inferences. It is possible that data may suffer from self-report bias related to recall problems, but it is less likely that this bias was strong enough to influence the results.

The data used in this study were relatively old. The NHIS cancer module repeats only every 5 years, and we used this data set because of the inclusion of a specific component on acculturation, found only in NHIS 2000, not in subsequent data sets. In addition, we could not combine modules 2000 and 2005 because the coding for race and ethnicity had changed in 2005. Although the accessible clean data were 1 cycle old, the findings are still relevant because CRC screening disparities between racial/ethnic minority groups and Whites persist.¹⁴

It is critical to note that patients’ recall of screening recommendations might not reflect physicians’ actual behavior. Conversely, it has been suggested that patient reports may be more reliable than medical records in measuring screening recommendation, because preventative care data tend to be under-documented in medical files.^52,55 This may be because of the limitations of the traditional medical file format, which does not emphasize recording screening recommendations. It is also possible that some individuals received a recommendation, did not understand it, or did not comply, and did not acknowledge that they actually received the recommendation from the physician.

Conclusions and Recommendations

Population-wide screening is necessary to reduce CRC burden. The findings of this study and others have shown that minority groups have lower rates of endoscopic screening recommendation. The results of our analyses suggest that these disparities are in fact attributable to differences in socioeconomic factors, especially education, income, and access to health care. These factors may influence the likelihood of receiving a screening recommendation through their effect on physician-patient communication.

Increasing physician–patient discourse, identifying and understanding where the communication breakdown is occurring, and employing educational tools at an appropriate literacy level may improve screening rates among underserved populations. Providing physicians with instruments (such as prompters) to deliver preventive messages—even when patients do not ask for them—may help reduce disparities between racial/ethnic groups regarding screening recommendations. It is also important to deliver basic health education to underserved populations to convey the importance of including discussion about cancer screening in their medical visits.