Colorectal cancer screening has increased substantially in New York City in recent years. However, screening uptake measured by telephone surveys may not fully capture rates among underserved populations. We measured screening completion within 1 year of a primary care visit among previously unscreened patients in a large urban safety-net hospital and identified sociodemographic and health-related predictors of screening.

We identified 21,256 patients ages 50 to 75 who were seen by primary care providers (PCP) in 2014, of whom 14,425 (67.9%) were not up-to-date with screening. Because PCPs facilitate the majority of screening, we compared patients who received screening within 1 year of an initial PCP visit to those who remained unscreened using multivariable logistic regression.

Among patients not up-to-date with screening at study outset, 11.5% (1,658 patients) completed screening within 1 year of a PCP visit. Asian race, more PCP visits, and higher area-level income were associated with higher screening completion. Factors associated with remaining unscreened included morbid obesity, ever smoking, Elixhauser comorbidity index of 0, and having Medicaid/Medicare insurance. Age, sex, language, and travel time to the hospital were not associated with screening status. Overall, 39.9% of patients were up-to-date with screening by 2015.

In an underserved urban population, colorectal cancer screening disparities remain, and overall screening uptake was low. Because more PCP visits were associated with modestly higher screening completion at 1 year, additional community-level education and outreach may be crucial to increase colorectal cancer screening in underserved populations.

Colorectal cancer screening, most commonly by colonoscopy or fecal occult blood testing (FOBT), has been shown to reduce colorectal cancer incidence and mortality (1–4). Current guidelines recommend screening for colorectal cancer starting at age 50 for average-risk individuals (5). An estimated 62% of the U.S. population is up-to-date with colorectal cancer screening (6). In New York City (NYC), screening has increased substantially over the past 15 years, corresponding to dedicated efforts by the Citywide Colorectal Cancer Control Coalition (C5) and the NYC Department of Health and Mental Hygiene (DOHMH; ref. 7). Screening data in NYC are obtained using the NYC Community Health Survey, an annual telephone survey conducted by the DOHMH. These survey data showed that NYC colorectal cancer screening rates increased from 42% in 2003 to 70% in 2010 and has since remained stable (8).

Although these statistics are encouraging, self-reported telephone surveys like the Community Health Survey may not represent all demographics or fully capture disparities (9). Nationally, there are well-documented screening disparities by age, education level, income, insurance status, and healthcare access (10). In the 2015 National Health Interview Survey (NHIS), non-Hispanic whites had the highest screening uptake at 65%, followed by African Americans (62%), Hispanics (50%), and Asians (49%; ref. 11). Other surveys have shown similar racial/ethnic differences, and lower screening rates in immigrants and non-English speakers (12–14).

The Community Health Survey reported that racial/ethnic disparities in screening had been eliminated in 2013 (7). Importantly, although the survey was conducted in a variety of languages, it could not reach individuals without a telephone, interviewed only one adult per household, and excluded adults living in group quarters (e.g., college dormitories, nursing facilities). In addition, only 40.5% of all eligible participants responded to the survey (8). Therefore, the NYC Community Health Survey may not capture the true screening rates within NYC's large medically underserved populations.

Therefore, we evaluated the rates of screening completion in primary care patients at Bellevue Hospital Center, the oldest and one of the largest public safety-net hospitals in the United States. Bellevue Hospital provides care to a diverse population with disproportionately low income and high rates of uninsured (15). Because most patients are informed about cancer screening during primary care visits, our main aim was to measure 1-year colorectal cancer screening completion rates—the rate at which previously unscreened patients underwent colorectal cancer screening after seeing a primary care provider (PCP). Second, we aimed to identify sociodemographic and medical predictors of screening completion within this uniquely diverse patient population.

Study design

We conducted a retrospective chart review of primary care patients who were not up-to-date with colorectal cancer screening at Bellevue Hospital Center. Data from PCP visits from 2004 to 2014, guaiac-based FOBT (gFOBT) results from 2013 to 2015, and colonoscopies and sigmoidoscopies performed from 2004 to 2016 were extracted from the electronic medical record. Patients who were unscreened at the initial PCP visit in 2014 were considered screened at 1 year if they completed gFOBT, colonoscopy, or sigmoidoscopy within the year (4, 16). We compared participants who were screened at 1 year to those who remained unscreened at 1 year. The study was conducted in accordance with the ethical guidelines of the Belmont Report, and was approved by the NYU School of Medicine Institutional Review Board (Study 16-01503).

At Bellevue, existing strategies to increase colorectal cancer screening included outreach mailings for patients, annual report cards to providers about screening rates, a part-time patient navigator who assisted with colonoscopy scheduling and bowel preparation education, and expedited pre-procedure appointments with a gastroenterology nurse practitioner. One notable barrier to screening was the lack of an electronic medical record reminder for providers to order colorectal cancer screening, which was implemented after the study period.

Participants

We identified all patients age 50 to 75 years who had at least one PCP office visit in 2014. Individuals with missing ZIP code data were excluded. Next, we excluded patients who were up-to-date with colorectal cancer screening at the time of their initial visit, defined as having completed a colonoscopy in the past 10 years, a sigmoidoscopy in the past 5 years, or gFOBT in the past 1 year, based on procedure codes in the medical record. The remaining patients—those not up-to-date with colorectal cancer screening—were included in the primary analysis.

Variables

Individual-level variables of interest included age, sex, race, ethnicity, country of origin, preferred language, ZIP code, BMI, smoking status, and insurance type. The modified Elixhauser comorbidity index—a weighted summary score of 31 medical conditions—was calculated using ICD 9/10 codes (17). The Elixhauser comorbidity index has been shown to outperform the Charlson comorbidity index in predicting mortality (18–20). We used the number of years that patients visited their PCPs between 2004 and 2014 as a measure of healthcare utilization, although the number of visits per year was unavailable.

Area-level data were obtained at the level of Primary Care Service Areas (PCSA) using the Dartmouth Atlas of Health Care (21). Area-level variables included population density, PCPs per capita, specialists per capita, median household income, education level, and percentage of white/black/Hispanic/Asian residents within the PCSA. In addition, we calculated average travel time on public transit from each residential ZIP code to the hospital using Google Maps based on a noon arrival time.

Outcomes

The primary outcomes were screening completion rates at 1 year after the initial PCP visit and sociodemographic and medical factors that predict screening completion. We measured screening completion rate over the study period rather than overall cross-sectional proportion screened, to identify specific modifiable factors or groups that can be potential targets for interventions.

Analysis

We compared individuals who were screened versus unscreened at 1 year on bivariate analysis using chi-squared and t tests. Variables with P < 0.10 on bivariate analysis were then included in a multivariable logistic regression model. A two-tailed P < 0.05 was considered statistically significant in this model. To assess whether screening completion increased with longer follow up, we conducted a sensitivity analysis examining gFOBT completion through 2015 (one additional year) and colonoscopy/sigmoidoscopy completion through 2016 (two additional years). All statistical analyses were performed using R version 3.5.0 (Foundation for Statistical Computing).

We identified 21,256 patients from 190 countries with at least one PCP visit in 2014 and available ZIP code data. Of these, 6,831 (32.1%) were up-to-date with colorectal cancer screening at the time of their initial visit and excluded from the analysis. Supplementary Table S1 compares characteristics of these patients who were up-to-date with screening with those not up-to-date. Age was inversely correlated with being up-to-date with screening, and there was a 15% difference in screening uptake between those in the 50 and 54 (24.7%) and 70 and 75 (40.2%) age groups. Hispanics (34.1%) and Blacks (33.4%) had a higher baseline screening rate than Asians (26.4%). Groups with low screening uptake at baseline included individuals who spoke Chinese, had low to normal BMI, had only one PCP visit in 2014, and had an Elixhauser score of 0.

After excluding participants who were up-to-date with screening, a total of 14,425 patients (67.9%) were included for the primary analysis. In this group, 1,658 (11.5%) patients completed screening within 1 year of their initial PCP visit. Table 1 compares characteristics of patients who received screening with those who remained unscreened at 1 year The largest absolute difference observed was between categories of healthcare utilization—9.1% screened among patients with PCP visits in two separate years versus 15.9% screened among patients with visits in 3 or more years with at least two visits before 2012 (6.8% absolute difference, P < 0.01). A statistically significant difference was also found in screening uptake between whites and Asians (8.5% vs. 13.9%, P < 0.01). Smaller but statistically significant differences were found between subgroups of ethnicity, language, country of origin, BMI, insurance type, smoking status, comorbidity score, and area-level median household income.

Table 1.

Baseline characteristics of patients not up-to-date with colorectal cancer screening, by screening status at 1 year.

VariableUnscreened at 1 year, N (%)Screened at 1 year, N (%)P
 Total 12,767 (88.5) 1,658 (11.5)  
Age Age mean (SD) 59.6 (6.7) 59.2 (6.7) 0.06 
Sex Female 6977 (88.4) 920 (11.6) 0.52 
 Male 5,790 (88.7) 738 (11.3)  
Race White 1,299 (91.5) 121 (8.5) <0.01 
 Black 1,939 (89.7) 223 (10.3)  
 Asian 2,730 (86.1) 441 (13.9)  
 Unknown 6,799 (88.6) 873 (11.4)  
Ethnicity Non-Hispanic or Latino 5,663 (89.1) 692 (10.9) <0.01 
 Hispanic or Latino 3,551 (88.9) 443 (11.1)  
 Unknown 3,553 (87.2) 523 (12.8)  
Language English 5,754 (89.4) 684 (10.6) <0.01 
 Spanish 3,720 (88.6) 479 (11.4)  
 Chinese 2,018 (85.9) 331 (14.1)  
 Other 1,275 (88.6) 164 (11.4)  
Country of origin United States 3,790 (90.3) 407 (9.7) <0.01 
 Mexico 551 (85.3) 95 (14.7)  
 China 1,674 (86.4) 264 (13.6)  
 Dominican Republic 1,192 (87.9) 164 (12.1)  
 Ecuador 509 (87.7) 71 (12.2)  
 Puerto Rico 658 (89.0) 81 (11.0)  
 Bangladesh 227 (89.0) 28 (11.0)  
 Other 4,166 (88.4) 548 (11.6)  
BMI 10–24.9 2,815 (87.1) 418 (12.9) <0.01 
 25–29.9 3,257 (87.4) 468 (12.6)  
 30–34.9 1,872 (88.5) 243 (11.5)  
 35+ 1,061 (90.3) 114 (9.7)  
 Unknown 3,762 (90.1) 415 (9.9)  
Smoking status Never used 3,778 (85.6) 635 (14.4) <0.01 
 Ever used 1,748 (89.5) 206 (10.5)  
 Unknown 7,241 (89.9) 817 (10.1)  
Healthcare Utilization Visit in 2014 only 3,267 (88.9) 407 (11.1) <0.01 
 Visits in 2 years 1,751 (90.9) 175 (9.1)  
 Visits in 3+ years, ≥1 visit in 2012–2013 7,378 (88.0) 1006 (12.0)  
 Visits in 3+ years, none in 2012–2013 371 (84.1) 70 (15.9)  
Insurance Private 1,061 (87.8) 147 (12.2) <0.01 
 Medicaid 1,552 (91.2) 150 (8.8)  
 Medicare 2,386 (89.7) 274 (10.3)  
 No insurance 5,445 (87.0) 811 (13.0)  
 Other/Unknown 2,323 (89.4) 276 (10.6)  
Travel time 1–29 minutes 1,797 (89.9) 203 (10.2) 0.11 
 30–59 minutes 8,111 (88.4) 1065 (11.6)  
 60+ minutes 2,859 (88.0) 390 (12.0)  
Median household income in PCSA <45,000 2,806 (89.1) 343 (10.9) 0.04 
 45,000–60,000 6,682 (87.8) 928 (12.2)  
 60,000–80,000 1,377 (88.7) 176 (11.3)  
 80,000–100,000 1,251 (90.5) 132 (9.5)  
 >100,000 651 (89.2) 79 (10.8)  
Elixhauser comorbidity score <0 3,604 (87.3) 524 (12.7) 0.02 
 4,323 (89.4) 514 (10.6)  
 1–6 2,338 (88.6) 301 (11.4)  
 7+ 2,502 (88.7) 319 (11.3)  
% of Hispanics in PCSA Highest quartile (33.1–80.5) 2,911 (89.0) 361 (11.0) 0.06 
 Q3 (24.0–33.1) 4,149 (87.5) 594 (12.5)  
 Q2 (13.2–24.0) 2,535 (88.9) 316 (11.1)  
 Lowest quartile (2.2–13.2) 3,172 (89.1) 387 (10.9)  
% of Blacks in PCSA Highest quartile (24.7–93.6) 3,334 (88.8) 419 (11.2) 0.13 
 Q3 (9.0–24.7) 1,496 (90.0) 166 (10.0)  
 Q2 (3.5–9.0) 4,639 (88.1) 628 (11.9)  
 Lowest quartile (0.12–3.5) 3,298 (88.1) 445 (11.9)  
% of Whites in PCSA Highest quartile (62.9–98.1) 3,049 (90.1) 336 (9.9) <0.01 
 Q3 (41.7–62.9) 2,521 (88.8) 318 (11.2)  
 Q2 (32.2–41.7) 4,188 (87.3) 609 (12.7)  
 Lowest quartile (1.7–41.7) 3,009 (88.4) 395 (11.6)  
% of Asians in PCSA Highest quartile (36.1–61.9) 3,315 (87.2) 486 (12.8) 0.02 
 Q3 (15.2–36.2) 3,297 (88.5) 430 (11.5)  
 Q2 (5.3–15.2) 2,798 (89.1) 341 (10.9)  
 Lowest quartile (0.34–5.3) 3,357 (89.3) 401 (10.7)  
Population density in PCSA Highest quartile (75,910–143,300) 3,784 (88.3) 503 (11.7) 0.33 
 Q3 (50,120–75,910) 2,385 (87.7) 333 (12.3)  
 Q2 (29,680- 50,120) 3,443 (88.7) 439 (11.3)  
 Lowest quartile (92.5–29,680) 3,155 (89.2) 383 (10.8)  
PCPs per capita in PCSA Highest quartile (0.0013–0.004) 2,981 (89.3) 358 (10.7) 0.36 
 Q3(0.0012–0.0013) 1,012 (88.1) 137 (11.9)  
 Q2 (0.0006–0.0012) 5,616 (88.1) 758 (11.9)  
 Lowest quartile (0.00004–0.0006) 3,158 (88.6) 405 (11.4)  
Specialists per capita in PCSA Highest quartile (0.003–0.012) 3,106 (89.0) 384 (11.0) 0.38 
 Q3 (0.001- 0.003) 2,675 (88.4) 351 (11.6)  
 Q2 (0.0007–0.001) 3,942 (87.9) 543 (12.1)  
 Lowest quartile (0.00009–0.0007) 3,044 (88.9) 380 (11.1)  
% High school graduates in PCSA Highest quartile (85.6–98.6) 3,508 (89.5) 413 (10.5) 0.07 
 Q3 (77.3–85.6) 2,794 (88.8) 353 (11.2)  
 Q2 (69.0–77.3) 2,627 (88.2) 352 (11.8)  
 Lowest quartile (52.3–69.0) 3,838 (87.7) 540 (12.3)  
VariableUnscreened at 1 year, N (%)Screened at 1 year, N (%)P
 Total 12,767 (88.5) 1,658 (11.5)  
Age Age mean (SD) 59.6 (6.7) 59.2 (6.7) 0.06 
Sex Female 6977 (88.4) 920 (11.6) 0.52 
 Male 5,790 (88.7) 738 (11.3)  
Race White 1,299 (91.5) 121 (8.5) <0.01 
 Black 1,939 (89.7) 223 (10.3)  
 Asian 2,730 (86.1) 441 (13.9)  
 Unknown 6,799 (88.6) 873 (11.4)  
Ethnicity Non-Hispanic or Latino 5,663 (89.1) 692 (10.9) <0.01 
 Hispanic or Latino 3,551 (88.9) 443 (11.1)  
 Unknown 3,553 (87.2) 523 (12.8)  
Language English 5,754 (89.4) 684 (10.6) <0.01 
 Spanish 3,720 (88.6) 479 (11.4)  
 Chinese 2,018 (85.9) 331 (14.1)  
 Other 1,275 (88.6) 164 (11.4)  
Country of origin United States 3,790 (90.3) 407 (9.7) <0.01 
 Mexico 551 (85.3) 95 (14.7)  
 China 1,674 (86.4) 264 (13.6)  
 Dominican Republic 1,192 (87.9) 164 (12.1)  
 Ecuador 509 (87.7) 71 (12.2)  
 Puerto Rico 658 (89.0) 81 (11.0)  
 Bangladesh 227 (89.0) 28 (11.0)  
 Other 4,166 (88.4) 548 (11.6)  
BMI 10–24.9 2,815 (87.1) 418 (12.9) <0.01 
 25–29.9 3,257 (87.4) 468 (12.6)  
 30–34.9 1,872 (88.5) 243 (11.5)  
 35+ 1,061 (90.3) 114 (9.7)  
 Unknown 3,762 (90.1) 415 (9.9)  
Smoking status Never used 3,778 (85.6) 635 (14.4) <0.01 
 Ever used 1,748 (89.5) 206 (10.5)  
 Unknown 7,241 (89.9) 817 (10.1)  
Healthcare Utilization Visit in 2014 only 3,267 (88.9) 407 (11.1) <0.01 
 Visits in 2 years 1,751 (90.9) 175 (9.1)  
 Visits in 3+ years, ≥1 visit in 2012–2013 7,378 (88.0) 1006 (12.0)  
 Visits in 3+ years, none in 2012–2013 371 (84.1) 70 (15.9)  
Insurance Private 1,061 (87.8) 147 (12.2) <0.01 
 Medicaid 1,552 (91.2) 150 (8.8)  
 Medicare 2,386 (89.7) 274 (10.3)  
 No insurance 5,445 (87.0) 811 (13.0)  
 Other/Unknown 2,323 (89.4) 276 (10.6)  
Travel time 1–29 minutes 1,797 (89.9) 203 (10.2) 0.11 
 30–59 minutes 8,111 (88.4) 1065 (11.6)  
 60+ minutes 2,859 (88.0) 390 (12.0)  
Median household income in PCSA <45,000 2,806 (89.1) 343 (10.9) 0.04 
 45,000–60,000 6,682 (87.8) 928 (12.2)  
 60,000–80,000 1,377 (88.7) 176 (11.3)  
 80,000–100,000 1,251 (90.5) 132 (9.5)  
 >100,000 651 (89.2) 79 (10.8)  
Elixhauser comorbidity score <0 3,604 (87.3) 524 (12.7) 0.02 
 4,323 (89.4) 514 (10.6)  
 1–6 2,338 (88.6) 301 (11.4)  
 7+ 2,502 (88.7) 319 (11.3)  
% of Hispanics in PCSA Highest quartile (33.1–80.5) 2,911 (89.0) 361 (11.0) 0.06 
 Q3 (24.0–33.1) 4,149 (87.5) 594 (12.5)  
 Q2 (13.2–24.0) 2,535 (88.9) 316 (11.1)  
 Lowest quartile (2.2–13.2) 3,172 (89.1) 387 (10.9)  
% of Blacks in PCSA Highest quartile (24.7–93.6) 3,334 (88.8) 419 (11.2) 0.13 
 Q3 (9.0–24.7) 1,496 (90.0) 166 (10.0)  
 Q2 (3.5–9.0) 4,639 (88.1) 628 (11.9)  
 Lowest quartile (0.12–3.5) 3,298 (88.1) 445 (11.9)  
% of Whites in PCSA Highest quartile (62.9–98.1) 3,049 (90.1) 336 (9.9) <0.01 
 Q3 (41.7–62.9) 2,521 (88.8) 318 (11.2)  
 Q2 (32.2–41.7) 4,188 (87.3) 609 (12.7)  
 Lowest quartile (1.7–41.7) 3,009 (88.4) 395 (11.6)  
% of Asians in PCSA Highest quartile (36.1–61.9) 3,315 (87.2) 486 (12.8) 0.02 
 Q3 (15.2–36.2) 3,297 (88.5) 430 (11.5)  
 Q2 (5.3–15.2) 2,798 (89.1) 341 (10.9)  
 Lowest quartile (0.34–5.3) 3,357 (89.3) 401 (10.7)  
Population density in PCSA Highest quartile (75,910–143,300) 3,784 (88.3) 503 (11.7) 0.33 
 Q3 (50,120–75,910) 2,385 (87.7) 333 (12.3)  
 Q2 (29,680- 50,120) 3,443 (88.7) 439 (11.3)  
 Lowest quartile (92.5–29,680) 3,155 (89.2) 383 (10.8)  
PCPs per capita in PCSA Highest quartile (0.0013–0.004) 2,981 (89.3) 358 (10.7) 0.36 
 Q3(0.0012–0.0013) 1,012 (88.1) 137 (11.9)  
 Q2 (0.0006–0.0012) 5,616 (88.1) 758 (11.9)  
 Lowest quartile (0.00004–0.0006) 3,158 (88.6) 405 (11.4)  
Specialists per capita in PCSA Highest quartile (0.003–0.012) 3,106 (89.0) 384 (11.0) 0.38 
 Q3 (0.001- 0.003) 2,675 (88.4) 351 (11.6)  
 Q2 (0.0007–0.001) 3,942 (87.9) 543 (12.1)  
 Lowest quartile (0.00009–0.0007) 3,044 (88.9) 380 (11.1)  
% High school graduates in PCSA Highest quartile (85.6–98.6) 3,508 (89.5) 413 (10.5) 0.07 
 Q3 (77.3–85.6) 2,794 (88.8) 353 (11.2)  
 Q2 (69.0–77.3) 2,627 (88.2) 352 (11.8)  
 Lowest quartile (52.3–69.0) 3,838 (87.7) 540 (12.3)  

Table 2 shows results of the multivariable logistic regression model. Independent predictors of screening completion included Asian race (OR = 1.58; 95% CI, 1.23–2.05), Mexican country of origin (OR = 1.43; 95% CI, 1.09–1.86), three or more PCP visits since 2004 (OR = 1.61; 95% CI, 1.21–2.13), and higher area-level income (>100,000 USD: OR = 1.44; 95% CI, 1.03–2.00). Predictors of remaining unscreened included morbid obesity (OR = 0.74; 95% CI, 0.59–0.93), ever smoking (OR = 0.75; 95% CI, 0.63–0.90), having Medicaid (OR = 0.69; 95% CI, 0.54–0.88) or Medicare (OR = 0.79; 95% CI, 0.62–0.99) compared with private insurance, and having an Elixhauser comorbidity score of 0 (OR = 0.75; 95% CI, 0.65–0.86). Age, sex, language, travel time to the hospital, and area-level education were not significantly associated with screening status.

Table 2.

Predictors of PCSA colorectal cancer screening completion at 1 year in previously unscreened patients (N = 14,425).

VariableOR95% CI
Age Mean (SD) 1.00 0.99–1.01 
Sex Female REF  
 Male 1.03 0.92–1.14 
Race White REF  
 Asian 1.58 1.23–2.05 
 Black 1.24 0.97–1.58 
 Unknown 1.19 0.96–1.50 
Ethnicity Non-Hispanic or Latino REF  
 Hispanic or Latino 1.02 0.86–1.20 
 Unknown 1.25 1.09–1.43 
Languagea English REF  
 Spanish 0.97 0.83–1.13 
 Chinese 1.10 0.87–1.38 
 Other 0.99 0.81–1.19 
Country of originb United States REF  
 Mexico 1.43 1.09–1.86 
 China 1.05 0.83–1.33 
 Dominican Republic 1.24 1.00–1.53 
 Ecuador 1.20 0.89–1.59 
 Puerto Rico 1.14 0.87–1.48 
 Bangladesh 0.97 0.62–1.45 
 Other 1.08 0.94–1.26 
BMI 10–24.9 REF  
 25–29.9 0.98 0.84–1.13 
 30–34.9 0.89 0.75–1.06 
 35+ 0.74 0.59–0.93 
 Unknown 0.75 0.65–0.88 
Smoking status Never Used REF  
 Ever Used 0.75 0.63–0.90 
 Unknown 0.66 0.59–0.74 
Healthcare utilization Visit in 2014 only REF  
 Visits in 2 years 0.86 0.71–1.04 
 Visits in 3+ years, ≥1 visit in 2012–2013 1.10 0.96–1.26 
 Visits in 3+ years, none in 2012–2013 1.61 1.21–2.13 
Insurance Private REF  
 Medicaid 0.69 0.54–0.88 
 Medicare 0.79 0.62–0.99 
 No insurance 1.03 0.85–1.25 
 Other/unknown 0.79 0.63–0.98 
Travel time 1–29 minutes REF  
 30–59 minutes 1.04 0.85–1.26 
 60+ minutes 1.08 0.86–1.36 
Median household income in PCSA <45,000 REF  
 45,000–60,000 1.19 1.03–1.37 
 60,000–80,000 1.20 0.95–1.52 
 80,000–100,000 1.33 0.96–1.84 
 100,000+ 1.44 1.03–2.00 
% High school graduates in PCSAc  0.94 0.87–1.01 
% of Whites in PCSAc  0.95 0.88–1.01 
Elixhauser comorbidity score <0 REF  
 0.75 0.65–0.86 
 1–6 0.89 0.76–1.04 
 7+ 0.89 0.76–1.04 
VariableOR95% CI
Age Mean (SD) 1.00 0.99–1.01 
Sex Female REF  
 Male 1.03 0.92–1.14 
Race White REF  
 Asian 1.58 1.23–2.05 
 Black 1.24 0.97–1.58 
 Unknown 1.19 0.96–1.50 
Ethnicity Non-Hispanic or Latino REF  
 Hispanic or Latino 1.02 0.86–1.20 
 Unknown 1.25 1.09–1.43 
Languagea English REF  
 Spanish 0.97 0.83–1.13 
 Chinese 1.10 0.87–1.38 
 Other 0.99 0.81–1.19 
Country of originb United States REF  
 Mexico 1.43 1.09–1.86 
 China 1.05 0.83–1.33 
 Dominican Republic 1.24 1.00–1.53 
 Ecuador 1.20 0.89–1.59 
 Puerto Rico 1.14 0.87–1.48 
 Bangladesh 0.97 0.62–1.45 
 Other 1.08 0.94–1.26 
BMI 10–24.9 REF  
 25–29.9 0.98 0.84–1.13 
 30–34.9 0.89 0.75–1.06 
 35+ 0.74 0.59–0.93 
 Unknown 0.75 0.65–0.88 
Smoking status Never Used REF  
 Ever Used 0.75 0.63–0.90 
 Unknown 0.66 0.59–0.74 
Healthcare utilization Visit in 2014 only REF  
 Visits in 2 years 0.86 0.71–1.04 
 Visits in 3+ years, ≥1 visit in 2012–2013 1.10 0.96–1.26 
 Visits in 3+ years, none in 2012–2013 1.61 1.21–2.13 
Insurance Private REF  
 Medicaid 0.69 0.54–0.88 
 Medicare 0.79 0.62–0.99 
 No insurance 1.03 0.85–1.25 
 Other/unknown 0.79 0.63–0.98 
Travel time 1–29 minutes REF  
 30–59 minutes 1.04 0.85–1.26 
 60+ minutes 1.08 0.86–1.36 
Median household income in PCSA <45,000 REF  
 45,000–60,000 1.19 1.03–1.37 
 60,000–80,000 1.20 0.95–1.52 
 80,000–100,000 1.33 0.96–1.84 
 100,000+ 1.44 1.03–2.00 
% High school graduates in PCSAc  0.94 0.87–1.01 
% of Whites in PCSAc  0.95 0.88–1.01 
Elixhauser comorbidity score <0 REF  
 0.75 0.65–0.86 
 1–6 0.89 0.76–1.04 
 7+ 0.89 0.76–1.04 

aCountry was removed from model to obtain estimates for language.

bLanguage was removed from model to obtain estimates for country

cThose with missing data for % high school graduates in PCSA and % of whites in PCSA were excluded from the model.

A sensitivity analysis extending the time frame for screening completion for gFOBT through 2015 and for colonoscopy and sigmoidoscopy through 2016 resulted in a slightly higher screening uptake of 14.7%.

Among individuals overdue for colorectal cancer screening, screening completion within 1 year of an initial PCP visit was low overall at 11.5%. Extending the screening time frame by an additional 1 to 2 years only increased the absolute screening uptake by 3.2%. On multivariable logistic regression, Asian race, Mexican country of origin, more PCP visits, and higher median household income were all statistically significant predictors of screening completion. Conversely, morbid obesity, positive smoking history, insurance with Medicare or Medicaid, and an Elixhauser comorbidity index of 0 were associated with incomplete screening. However, the absolute differences in screening completion between all categories were modest.

Adding together the individuals who were up-to-date with screening at the outset of the study with those who were screened within 1 year, up to 39.9% of primary care patients would have been up-to-date with screening at the end of 2015. The true figure would be lower, because a proportion of previously up-to-date individuals would have become overdue for screening during this year. Because even a screening rate of 39.9% is substantially lower than the 70% uptake reported in the NYC Community Health Survey, these results support our hypothesis that telephone surveys may overestimate screening in an underserved population. Compared with the population surveyed by the Community Health Survey, our study population included a higher proportion of Hispanics (27.6% vs. 23.4%) and Asians (15.0% vs. 10.0%) and fewer individuals who preferred English primarily (44.6% vs. 65.9%; ref. 22). At Bellevue Hospital, uninsured visits made up 31% of all clinic visits, compared with 11% on average at other voluntary nonprofit NYC hospitals (15). These figures highlight the substantial differences between the population sampled by the telephone survey and NYC's medically underserved population. Therefore, despite great improvements in colorectal cancer screening in NYC overall, there remains a clear screening gap in our safety-net hospital population.

Reported colorectal cancer screening rates in other underserved populations have varied widely. Among unscreened patients at a safety-net hospital in Fort Worth/Tarrant County, Texas, colorectal cancer screening completion at 1 year was 12.1%, similar to our figure (23). At Parkland Hospital System in Dallas, Texas, screening completion rate among unscreened patients was 29.6% at 1 year (24). A later study at the same center observed a rate of 45.1% after 18 months (25). With respect to cross-sectional screening rates, 46.8% of patients were up-to-date with screening in 2015 in a study of federally qualified health centers (FQHC) in Oregon and California (26), comparable to our overall up-to-date rate of 39.9%. Data from US Health Resources & Services Administration (HRSA) Health Centers, which are designated primary care centers that provide services regardless of patients' ability to pay, showed CRC screening rates of 40% to 44% nationally in 2016 to 2018, with rates of 44% to 50% in New York state (27). These findings taken together consistently show that screening rates in underserved patient populations are lower than corresponding national- and state-level screening rates reported in the general population.

Asians had the lowest screening uptake at baseline (Supplementary Table S1) but had 5.4% higher absolute screening completion (58% higher in relative terms) than whites at 1 year. This finding is consistent with prior studies that have shown Asian Americans have the lowest cross-sectional screening rate in the United States, but have higher screening completion than other racial/ethnic groups when actively engaged by the healthcare system (11, 28–30).

Prior studies have found Hispanics to have higher screening uptake rates compared with Caucasians (24, 28, 31). This relationship was not seen in our study, although patients born in Mexico were more likely to complete screening. In addition, language was not associated with screening completion in our population, though 55% of individuals preferred a language other than English. This may reflect the widespread use of interpreter services within our institution that minimized the impact of patient-physician language discordance.

We found that a greater number of PCP visits distributed over a longer period of time was associated with higher screening completion. A higher frequency of office visits has similarly been shown to predict colorectal cancer screening completion in other studies (24–26). These suggest that the length of the patient–provider relationship is important for obtaining appropriate preventive care. Perhaps providers under time constraints are unable to address non-acute issues such as screening until a second or third appointment. Higher PCP visit frequency may also be a proxy for increased health awareness, access, and willingness to undergo medical interventions—all factors that contribute to higher screening uptake.

Strengths and limitations

The strengths of our study include its large sample size and diverse, underserved population. A few limitations should also be noted. First, because our institution is the flagship medical center for a network of public hospitals and clinics in NYC, it is possible that patients received primary care at our institution but underwent colorectal cancer screening at another facility. Our data does not capture screening at other facilities and therefore may underestimate the true screening uptake. We calculated travel time between residential ZIP codes and our hospital on the assumption that individuals who lived further away may be more likely to undergo screening at another facility, and we found no difference in screening rates by travel time. However, we acknowledge that individuals may have obtained screening at other facilities for reasons other than distance. Second, some patients are seen in Bellevue's primary care clinic as follow up from an emergency visit or hospitalization, which may lead to lower screening rates than would be expected in a more stable outpatient panel. Nevertheless, 75% of our study population had multiple primary care visits between 2004 and 2014, which suggests some measure of long-term care. Finally, there was substantial missing data for race, ethnicity, smoking status, and BMI in our electronic health record. However, there is no reason to believe this led to differential exposure misclassification.

In a diverse, medically underserved population, screening uptake within 1 year of an initial PCP visit was low overall at 11.5%. This was substantially lower than estimates from a citywide telephone survey of the general population. Asian race and more frequent PCP visits predicted screening completion, but absolute differences between subgroups were small. These findings suggest that even with counseling at PCP visits, it is difficult to successfully screen patients in an underserved population. Therefore, combining PCP visits with additional community-based targeted interventions may be needed to improve overall screening rates.

No potential conflicts of interest were disclosed.

Conception and design: K. Ni, P.S. Liang

Development of methodology: K. Ni, A.B. Wallach, M. Du, P.S. Liang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Anand, S.C. Yakoubovitch, P.S. Liang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Ni, K. O'Connell, M. Du, P.S. Liang

Writing, review, and/or revision of the manuscript: K. Ni, K. O'Connell, S. Anand, S.C. Yakoubovitch, S.C. Kwon, R.A. de Latour, A.B. Wallach, S.E. Sherman, M. Du, P.S. Liang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.C. Yakoubovitch, P.S. Liang

Study supervision: P.S. Liang

The authors thank the Data Sciences Department at New York City Health + Hospitals for their help with data extraction. M. Du and K. O'Connell were supported by NCI Grant P30CA008748 P.S. Liang was awarded ReMission Foundation grant and NCI grant K08CA230162. S.C. Kwon was supported by NIH NCATS grant UL1TR001445 and NIH NCI grant P30CA016087.

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.

1.
Mandel
JS
,
Church
TR
,
Bond
JH
,
Ederer
F
,
Geisser
MS
,
Mongin
SJ
, et al
The effect of fecal occult-blood screening on the incidence of colorectal cancer
.
N Engl J Med
2000
;
343
:
1603
7
.
2.
Muller
AD
,
Sonnenberg
A
. 
Protection by endoscopy against death from colorectal cancer. A case-control study among veterans
.
Arch Intern Med
1995
;
155
:
1741
8
.
3.
Winawer
SJ
,
Zauber
AG
,
Ho
MN
,
O'Brien
MJ
,
Gottlieb
LS
,
Sternberg
SS
, et al
Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup
.
N Engl J Med
1993
;
329
:
1977
81
.
4.
Lin
JS
,
Piper
MA
,
Perdue
LA
,
Rutter
CM
,
Webber
EM
,
O'Connor
E
, et al
Screening for colorectal cancer: updated evidence report and systematic review for the US Preventive Services Task Force
.
JAMA
2016
;
315
:
2576
94
.
5.
Rex
DK
,
Boland
CR
,
Dominitz
JA
,
Giardiello
FM
,
Johnson
DA
,
Kaltenbach
T
, et al
Colorectal cancer screening: recommendations for physicians and patients from the U.S. Multi-Society Task Force on Colorectal Cancer
.
Am J Gastroenterol
2017
;
112
:
1016
30
.
6.
White
A
,
Thompson
TD
,
White
MC
,
Sabatino
SA
,
de Moor
J
,
Doria-Rose
PV
, et al
Cancer screening test use: United States, 2015
.
MMWR Morb Mortal Wkly Rep
2017
;
66
:
201
6
.
7.
Itzkowitz
SH
,
Winawer
SJ
,
Krauskopf
M
,
Carlesimo
M
,
Schnoll-Sussman
FH
,
Huang
K
, et al
New York Citywide colon cancer control coalition: a public health effort to increase colon cancer screening and address health disparities
.
Cancer
2016
;
122
:
269
77
.
8.
New York City Department of Health and Mental Hygiene Community Health Survey
.
2003–2016. NYC Health Data EpiQuery Tool. Accessed August 7
, 
2018
.
9.
Call
KT
,
Davern
M
,
Boudreaux
M
,
Johnson
PJ
,
Nelson
J
. 
Bias in telephone surveys that do not sample cell phones: uses and limits of poststratification adjustments
.
Med Care
2011
;
49
:
355
64
.
10.
Hall
IJ
,
Tangka
FKL
,
Sabatino
SA
,
Thompson
TD
,
Graubard
BI
,
Breen
N
. 
Patterns and trends in cancer screening in the United States
.
Prev Chronic Dis
2018
;
15
:
E97
.
11.
Siegel
RL
,
Miller
KD
,
Fedewa
SA
,
Ahnen
DJ
,
Meester
RGS
,
Barzi
A
, et al
Colorectal cancer statistics, 2017
.
CA Cancer J Clin
2017
;
67
:
177
93
.
12.
Wee
CC
,
McCarthy
EP
,
Phillips
RS
. 
Factors associated with colon cancer screening: the role of patient factors and physician counseling
.
Prev Med
2005
;
41
:
23
9
.
13.
Liss
DT
,
Baker
DW
. 
Understanding current racial/ethnic disparities in colorectal cancer screening in the United States: the contribution of socioeconomic status and access to care
.
Am J Prev Med
2014
;
46
:
228
36
.
14.
Lin
SC
,
McKinley
D
,
Sripipatana
A
,
Makaroff
L
. 
Colorectal cancer screening at US community health centers: Examination of sociodemographic disparities and association with patient-provider communication
.
Cancer
2017
;
123
:
4185
92
.
15.
Pressman
M
,
Bohlen
S
.
2013 Bellevue Hospital Center Community Health Needs Assessment and Implementation Strategy
.
New York
:
Bellevue Hospital Center
; 
2013
.
16.
Levin
B
,
Lieberman
DA
,
McFarland
B
,
Smith
RA
,
Brooks
D
,
Andrews
KS
, et al
Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology
.
CA Cancer J Clin
2008
;
58
:
130
60
.
17.
van Walraven
C
,
Austin
PC
,
Jennings
A
,
Quan
H
,
Forster
AJ
. 
A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data
.
Med Care
2009
;
47
:
626
33
.
18.
Southern
DA
,
Quan
H
,
Ghali
WA
. 
Comparison of the Elixhauser and Charlson/Deyo Methods of comorbidity measurement in administrative data
.
Med Care
2004
;
42
:
355
60
.
19.
Stukenborg
GJ
,
Wagner
DP
,
Connors
AF
 Jr.
Comparison of the performance of two comorbidity measures, with and without information from prior hospitalizations
.
Med Care
2001
;
39
:
727
39
.
20.
Chu
YT
,
Ng
YY
,
Wu
SC
. 
Comparison of different comorbidity measures for use with administrative data in predicting short- and long-term mortality
.
BMC Health Serv Res
2010
;
10
:
140
.
21.
Primary Care Service Area (PCSA) Project - 2010 Census Tract
.
The Dartmouth Atlas of Health Care
2010
;
3
.
1
.
22.
Rastogi
N
,
Xia
Y
,
Inadomi
JM
,
Kwon
SC
,
Trinh-Shevrin
C
,
Liang
PS
. 
Disparities in colorectal cancer screening in New York City: an analysis of the 2014 NYC Community Health Survey
.
Cancer Med
2019
;
8
:
2572
9
.
23.
Gupta
S
,
Halm
EA
,
Rockey
DC
,
Hammons
M
,
Koch
M
,
Carter
E
, et al
Comparative effectiveness of fecal immunochemical test outreach, colonoscopy outreach, and usual care for boosting colorectal cancer screening among the underserved: a randomized clinical trial
.
JAMA Intern Med
2013
;
173
:
1725
32
.
24.
Singal
AG
,
Gupta
S
,
Tiro
JA
,
Skinner
CS
,
McCallister
K
,
Sanders
JM
, et al
Outreach invitations for FIT and colonoscopy improve colorectal cancer screening rates: a randomized controlled trial in a safety-net health system
.
Cancer
2016
;
122
:
456
63
.
25.
Hughes
AE
,
Tiro
JA
,
Balasubramanian
BA
,
Skinner
CS
,
Pruitt
SL
. 
Social disadvantage, healthcare utilization, and colorectal cancer screening: leveraging longitudinal patient address and health records data
.
Cancer Epidemiol Biomarkers Prev
2018
;
27
:
1424
.
26.
Petrik
AF
,
Le
T
,
Keast
E
,
Rivelli
J
,
Bigler
K
,
Green
B
, et al
Predictors of colorectal cancer screening prior to implementation of a large pragmatic trial in federally qualified health centers
.
J Community Health
2018
;
43
:
128
36
.
27.
2018 National Health Center Data
. 
Health Resources and Service Administration
; 
2018
.
28.
Inadomi
JM
,
Vijan
S
,
Janz
NK
,
Fagerlin
A
,
Thomas
JP
,
Lin
YV
, et al
Adherence to colorectal cancer screening: a randomized clinical trial of competing strategies
.
Arch Intern Med
2012
;
172
:
575
82
.
29.
Liang
PS
,
Wheat
CL
,
Abhat
A
,
Brenner
AT
,
Fagerlin
A
,
Hayward
RA
, et al
Adherence to competing strategies for colorectal cancer screening over 3 years
.
Am J Gastroenterol
2016
;
111
:
105
14
.
30.
Burnett-Hartman
AN
,
Mehta
SJ
,
Zheng
Y
,
Ghai
NR
,
McLerran
DF
,
Chubak
J
, et al
Racial/ethnic disparities in colorectal cancer screening across healthcare systems
.
Am J Prev Med
2016
;
51
:
e107
15
.
31.
Hudson
SV
,
Ferrante
JM
,
Ohman-Strickland
P
,
Hahn
KA
,
Shaw
EK
,
Hemler
J
, et al
Physician recommendation and patient adherence for colorectal cancer screening
.
J Am Board Fam Med
2012
;
25
:
782
91
.

Supplementary data