Disparities in the incidence of and mortality from chronic diseases such as colorectal cancer (CRC) contribute to a lower life expectancy for black men (mean age, 71.8 y) compared with white men (mean age, 76.5 y) in the United States (1–3). In New York City (NYC) in 2003, vital statistics indicated significant racial disparities in mortality from CRC. Meanwhile, disparate rates of timely screening colonoscopy by race were noted among participants in the Community Health Survey (CHS), an annual random-digit–dial (RDD) survey of approximately 10,000 NYC residents; the lowest screening rates were found among racial groups with the highest CRC-related mortality (4). As a result, the NYC Department of Health and Mental Hygiene (DOHMH) instituted a multifaceted program to increase CRC screening rates and reduce disparities (4). By 2007, the CHS indicated reduced racial disparities and substantial improvements in CRC screening rates among all older New Yorkers, prompting the NYC Citywide Colon Cancer Control Coalition to declare that “powerfully, racial and ethnic colon cancer screening disparities among blacks, whites, Hispanics and Asians have been eliminated” (4,5). Although this achievement was remarkable as a population-wide health strategy, racial disparities in CRC mortality persisted over the same 5-year period. Therefore, it is unclear whether CRC screening rates improved across all subgroups, particularly in communities that may be excluded by RDD sampling such as people from vulnerable subsets of the population (ie, those without consistent telephone access), potentially obscuring disparities within subgroups.

We tested this hypothesis by examining 2 sampling methods used to capture population-level data for older black men: a weighted RDD sample and a location-based community sample. Among the location-based sample, we compared differences in self-reported history of CRC screening by 1) people with access to a working phone and 2) the type of community setting.

All participants were black men aged 50 or older. Compared with the CHS sample, the MHI sample had lower educational attainment, worse self-reported health, and a lower rate of CRC screening (49% vs 63%, P < .001) (Table 1). Ten percent of the MHI sample had no working telephone. Participants without a working telephone had lower educational attainment, worse self-reported health, and lower CRC screening rates (Table 2). After excluding MHI participants with no working telephone, the CHS sample still had higher rates of screening colonoscopies (62.8% vs 49.3%, P < .001).

MHI participants from churches had the highest educational attainment compared with participants from other settings; 55.4% of church participants had at least some college education (Table 3). Only 3.5% of participants at churches lacked access to a working telephone compared with 12.5% of participants recruited from social service agencies (P < .001). Results also varied by setting type with regard to self-reported general health; only 18.2% of participants from churches reported fair or poor health compared with 35.1% of participants from senior centers (P < .001). Finally, 72.7% of senior center participants and 71.9% of church participants reported having ever had a screening colonoscopy compared with 47.1% of participants from social service locations, 55.0% of health fair participants, 46.1% of barbershop participants, and 30.6% of mosque participants (P < .001).

To decrease racial disparities in health, population-level interventions must reach those who are most in need. Likewise, accurate documentation of progress in reducing health disparities relies on the inclusion of diverse populations, including vulnerable subgroups, in surveillance efforts. We found considerable differences between location-based and RDD samples of older black men in NYC in terms of education and self-reported health, with the most striking difference being for CRC screening. Our data indicate that surveillance data must include methods for reaching people who may be more vulnerable than those reached in RDD samples to sufficiently capture disparities. Moreover, community-based interventions should include varied settings rather than concentrating efforts in 1 location to ensure reaching those who are most in need.

Despite underrepresentation in research, studies indicate that black men in the United States experience worse health outcomes than any other racial/ethnic or gender group (8–11). The institutionalization of racism and structural inequalities have created lasting health and socioeconomic inequalities affecting blacks in the United States, and black men specifically face structural disadvantages that undermine their likelihood of being included in surveillance efforts and health promotion programs (12–14). For example, black men are more likely than others to be imprisoned, and prisoners are excluded from most surveillance sampling frames (14–16). As many as 1 in 3 black men will be imprisoned at some point in their lives (16). Black men are also more likely to be unemployed and to have unstable housing or experience homelessness (15). Therefore, black men are less likely than others to be represented in RDD samples, other household-based samples, or community-based recruitment. In medical settings, black men may not be reached for intervention or surveillance purposes, because they are less likely than others to receive regular health care or to have a primary care provider (15). In addition, mistrust of research or medicine by blacks may result in the active avoidance of health-related programs and research in any setting (17). Furthermore, those reached using various techniques may be segmented, with no one setting or sampling strategy being truly representative of the population.

Our findings indicated important differences between people with and without working telephones, suggesting a potential for noncoverage bias in RDD population estimates of CRC screening. Also of note, 9.5% of our location-based sample had no working telephone, which is almost twice the national estimate of households with no telephone (18). Among the church-based sample, only 3.5% lacked a working telephone, indicating that these people would be more likely to be sampled using an RDD approach.

Although random sampling strategies are generally considered more representative of the general population, few studies have empirically examined the representativeness of these samples, perhaps due to a lack of appropriate comparison groups. One study found that location intercept-sampling, similar to our approach, resulted in a sample that had greater connection with their community, resulting in potential selection bias when compared with household-based sampling (19). However, this finding may have been due to the types of locations included. Our sample included a diverse selection of settings, which we believe attracted men from diverse socioeconomic backgrounds, regardless of their engagement with other community activities or the health care system. This inclusion also enabled us to examine the potential differences between samples intercepted in different settings. Understanding the discrepancy in CRC screening by setting and sampling technique has public health implications, as black men have higher CRC incidence and mortality than do other populations in the United States (3). In NYC, citywide RDD sampling indicates that disparities in screening by race have been dramatically reduced, which should lead to reductions in disparities in incidence and mortality (4). However, this apparent progress may reflect noncoverage bias and the omission of more disadvantaged segments of the population.

In recent years, churches have become a popular venue for implementing health programs, including interventions to promote cancer screening (20–22). Past research on interventions conducted in religious organizations point to the social structure of such organizations, such as the innate social support system that may benefit congregants’ health, and available resources for health programming (21,23). Thus, it is not surprising that our data indicate that black men intercepted at churches were significantly more likely than others to report prior CRC screening. Among our participants, 72% of men from churches had a prior screening colonoscopy, which was greater than the 63% RDD estimated screening rate for black men in NYC. Faith-based recruitment and interventions remain important vehicles for health promotion in black communities, but consideration should extend to community sites that serve marginalized segments of this population to ultimately eliminate racial disparities in CRC screening, morbidity, and mortality. Our data indicate there may be a need to broaden the reach of interventions beyond church congregations to include the surrounding community. As more affluent congregants have moved from the inner city to the suburbs, many black churches have remained in their original locations, in part due to a commitment to social justice and serving those left behind through social services and outreach (24). Partnerships with church-based social services such as food pantries and soup kitchens offer a viable way to reach those most in need, and such interventions would also reduce disparities. Participants at senior centers were also more likely than other participants to have had a colonoscopy. However, senior center attendees are older on average (mean age, 76 y) than the older adult population of NYC (25). Thus, senior center attendees may be more likely than others to qualify for Medicare and to have had more time to meet CRC screening recommendations. Similar to those of churches, the core functions of senior centers include providing social engagement, links to resources and services, and promoting health (25).

Participants from mosques, social services, and barbershops exhibited CRC screening rates that were far below the citywide RDD estimates for black men. Participating mosques were largely those serving the African immigrant community, congregants of which may lack access to services due to socioeconomic and immigration status, and availability of culturally appropriate care may contribute to the low screening rates among these participants (26). Despite their potential to reach many immigrant communities, mosques are included in few studies as settings for health promotion programs (26).

Conversely, many barbershop-based interventions target health issues similar to those targeted by church-based interventions. Because barbershops are important community centers for the black community, many men spend time at and around barbershops in their neighborhoods even when they are not getting their hair cut (27,28). Moreover, although we did not collect information on place of residence, we observed that many of the men recruited from barbershops were attracted from the surrounding neighborhood. This finding is congruent with Wright and Calhoun’s findings that barbershops are locations where men from the surrounding community are able to spend time together and escape the solitude of their homes (29). Thus, barbershops have salience for interventions targeting hard-to-reach populations such as black men.