Completion of Colonoscopy After Positive Fecal Occult Blood Test Screening at a Community Health Center

ABSTRACT

Introduction

Screening for colorectal cancer decreases mortality. Fecal occult blood testing (FOBT) decreases some barriers to screening, however positive results require colonoscopy. We evaluated factors associated with colonoscopy completion for community health center (CHC) patients after positive FOBT.

Methods

We identified patients of one CHC with positive FOBT from 1/1/2018‐12/31/2021. We performed chart reviews for demographics, insurance status, FOBT date, colonoscopy referral date and site, and colonoscopy results.

We performed descriptive analysis. We fitted a logistic model and employed stepwise selection to evaluate effect of variables influencing likelihood of colonoscopy. We employed forward and backward model selection to identify the reduced model with the lowest Akaike Information Criterion score. Using this model, we calculated hypothesis tests for each coefficient using Wald tests with an alpha level of 0.05.

Results

Overall, 50% of participants completed colonoscopy. Mean time between positive FOBT and colonoscopy completion was 237.4 days (SD 187.9). The reduced logistic model included age and health insurance covariates. Insured patients were 221.7% more likely to complete colonoscopy than uninsured patients. A seven‐year increase in age was associated with 41.2% increase in likelihood of colonoscopy. An increase of 38 miles to the colonoscopy site was associated with 17.7% decrease in likelihood of completion.

Conclusions

Only half of participants with positive FOBT completed colonoscopy. On average, nearly eight months elapsed between FOBT and colonoscopy. Having insurance was the strongest predictor of colonoscopy, despite available financial aid programs. Longer distances to colonoscopy sites decreased likelihood of completion.

1. Introduction

Colorectal cancer (CRC) is the second‐leading cause of cancer‐related deaths in the U.S. [1]. Although screening improves mortality, only two‐thirds of U.S. adults are appropriately screened [2, 3]. Individuals from disadvantaged populations are less likely to complete screening [4]. Racial disparities in CRC screening may account for half of the survival disparities [5].

CRC screening modalities include high‐sensitivity fecal occult blood testing (FOBT), fecal immunochemical testing, colonoscopy, computerized tomography colonography, and flexible sigmoidoscopy [6]; colonoscopy is the most sensitive, specific, and expensive. In one study of Medicaid patients, those with the most disadvantaged social determinants were more likely to receive FOBT than colonoscopy for screening [7]. Although FOBT is a lower‐barrier method, positive FOBT requires subsequent colonoscopy.

Community health centers (CHCs), which serve under‐resourced patients, may utilize FOBT screening, referring for colonoscopy after positive results. Access to gastroenterologists is limited by geography and insurance. According to the Health Resources & Services Administration, the 15 222 gastroenterologists practicing in the U.S. in 2020–2021 (4.6 gastroenterologists per 100 000 population) were not evenly distributed [8]. Colonoscopy sites for uninsured patients may be further limited to practices offering financial support. In addition to cost, transportation is a common barrier among underserved and minority patients [9, 10, 11]. Colonoscopies also require time for bowel preparation and absence from work.

We studied CRC screening at one CHC within the Washington, DC Metropolitan Area (DC Metro Area), which has 5.6 gastroenterologists per 100 000 population [8, 12]. This CHC provides CRC screening using FOBT; patients with positive FOBT are referred for colonoscopy at a private medical center 15 miles away within the DC Metro Area (Site A) or a public university medical center 115 miles away outside of the DC Metro Area (Site B) [13]. Both sites offer financial assistance for qualified uninsured patients; it is easier to qualify for Site B's assistance. We evaluated factors associated with colonoscopy completion after positive FOBT. We hypothesized there would be a low rate of colonoscopy completion and that rates would be higher among insured patients and those traveling shorter distances to colonoscopy sites.

2. Methods

Participants included all patients at a CHC in northern Virginia with a positive FOBT screening result from January 1, 2018 to December 31, 2021. We reviewed CHC electronic medical records (EMRs) of participants, collecting demographics, insurance status/type, FOBT result date, colonoscopy referral date and site, and colonoscopy reports. We reviewed EMRs at referral sites to confirm colonoscopy completion and results. We recorded whether a gastroenterology visit occurred before colonoscopy and whether this visit addressed issues other than a positive FOBT. Data were collected and managed using Research Electronic Data Capture (REDCap) at George Washington University [14, 15].

We performed descriptive analysis using mean and standard deviation. We excluded date entries from analysis if dates were not appropriately chronological (e.g., FOBT date was later than colonoscopy referral date). We calculated the distance between participants' homes and colonoscopy sites as the distance between zip codes [16], excluding participants without documented home zip codes. To compare demographic characteristics between participants who did and did not complete colonoscopy, we employed Chi‐squared tests of independence for categorical variables and two‐sided t‐tests assuming unequal variances with Welsh approximation for continuous variables.

We fitted a logistic model and employed stepwise model selection to evaluate the effect of variables that influence the likelihood of completing a colonoscopy. The response variable was the log odds of completing a colonoscopy, which we modeled as a linear function of the following covariates: age; sex; race; ethnicity; primary language; employment; income; insurance; colonoscopy referral site; distance to the colonoscopy site; and days elapsed between FOBT result and colonoscopy referral. We scaled numeric variables to facilitate model convergence.

We employed forward and backward model selection to identify the reduced model with the lowest Akaike Information Criterion (AIC) score. Using this model, we calculated the significance of hypothesis tests for each coefficient using Wald tests with an alpha level of 0.05. If the distance between home and colonoscopy site were not included in the reduced model, we planned to fit a second logistic model including distance as a covariate in addition to the covariates in the reduced model. We performed descriptive analyses, fitted and selected models, and generated tables and figures using R statistical computing software via the RStudio integrated development environment.

George Washington University's Institutional Review Board declared this research exempt (NCR224019). The University of Virginia (UVA) and Inova Health System (Inova) IRBs granted exempt concurrences.

3. Results

See Table 1 for participant demographics. The majority were white, non‐Hispanic women who were uninsured and earned less than $1000 per month. Overall, 50% of participants (65/130) completed colonoscopy, including 53% of those referred to Site A and 40% of those referred to Site B.

TABLE 1.

Demographic characteristics of study participants, sites to which they were referred for colonoscopy, and mean distances between participants' homes and the sites to which they were referred.

	Overall (n = 130), number (%)	Did not complete colonoscopy (n = 65), number (%)	Completed colonoscopy (n = 65), number (%)	p
Sex
Female	72 (55.4%)	37 (56.9%)	35 (53.8%)	0.860
Male	58 (44.6%)	28 (43.1%)	30 (46.2%)
Race
African American/Black	35 (26.9%)	17 (26.2%)	18 (27.7%)	0.837
Caucasian/White	66 (50.8%)	32 (49.2%)	34 (52.3%)
Asian	13 (10.0%)	6 (9.2%)	7 (10.8%)
Nat. Amer./Pac. Island	0 (0.0%)	0 (0.0%)	0 (0.0%)
Other	13 (10.0%)	8 (12.3%)	5 (7.7%)
Not documented	3 (2.3%)	2 (3.1%)	1 (1.5%)
Ethnicity
Hispanic	56 (43.1%)	29 (44.6%)	27 (41.5%)	0.799
Not Hispanic	73 (56.2%)	35 (53.8%)	38 (58.5%)
Not documented	1 (0.8%)	1 (1.5%)	0 (0%)
Primary language
English	46 (35.4%)	20 (30.8%)	26 (40.0%)	0.470
Spanish	55 (42.3%)	29 (44.6%)	26 (40.0%)
Other	28 (21.5%)	16 (24.6%)	12 (18.5%)
Not documented	1 (0.8%)	0 (0%)	1 (1.5%)
Employment status
Employed	55 (42.3%)	27 (41.5%)	28 (43.1%)	0.456
Unemployed	61 (46.9%)	33 (50.8%)	28 (43.1%)
Not documented	14 (10.8%)	5 (7.7%)	9 (13.8%)
Income
Under $1000	69 (53.1%)	32 (49.2%)	37 (56.9%)	0.313
$1001–2000	28 (21.5%)	18 (27.7%)	10 (15.4%)
$2001–3000	7 (5.4%)	2 (3.1%)	5 (7.7%)
$3001–4000	1 (0.8%)	1 (1.5%)	0 (0%)
Over $4000	1 (0.8%)	0 (0%)	1 (1.5%)
Not documented	24 (18.5%)	12 (18.5%)	12 (18.5%)
Insurance
Medicaid	27 (20.8%)	9 (13.8%)	18 (27.7%)	0.148
Medicare	4 (3.1%)	2 (3.1%)	2 (3.1%)
Commercial	0 (0.0%)	0 (0.0%)	0 (0.0%)
Uninsured	99 (76.2%)	54 (83.1%)	45 (69.2%)
Mean age	60.1 years (SD 6.7)	59.2 years (SD 6.5)	61.0 years (SD 6.9)	0.122
Mean distance from home to colonoscopy site	29.7 miles (SD 36.8)	34.4 miles (SD 37.9)	25.2 miles (SD 35.5)	0.170
Referred to
Site A: Private Med Ctr	100 (76.9%)	47 (72.3%)	53 (81.5%)	0.298
Site B: Pub Univ Med Ctr	30 (23.1%)	18 (27.7%)	12 (18.5%)

The mean time elapsed between positive FOBT and colonoscopy referral was 47.0 days (SD 119.3). (Data not shown.) Among those who were approved for financial assistance, the mean time between initiating and being approved for assistance was 161.9 days (SD 216.9, range 6–776). The mean time elapsed between positive FOBT and colonoscopy completion was 235.5 days (SD 187.0); this was 221.9 days (SD 228.9) for insured participants and 241.7 days (SD 167.7) for uninsured participants.

Among the 65 participants who completed colonoscopy, 48 (73.8%) had preceding gastroenterology visits. (Data not shown.) At those visits, 60.4% (29/48) addressed only positive FOBT; 39.6% (19/48) addressed something else. Among the 65 participants who did not complete colonoscopy, 9.2% (6/65) attended gastroenterology visits. One of these visits (16.7%) addressed only positive FOBT; five (83.3%) addressed another problem.

Sixty‐one of 65 colonoscopy reports (93.8%) were in the CHC EMR. Pathology results were present for 97.6% (41/42) of colonoscopies that included polyp removal. 75.6% (31/42) of reports that included polyp removal included recommendations for next surveillance.

Following model selection, the reduced model (AIC = 147.76) included age and insurance status. Other characteristics were not significantly associated with colonoscopy completion. Patients with insurance were 221.7% more likely to complete colonoscopy than patients without, adjusting for age (Table 2; Figure 1). One standard deviation increase in age (approximately 7 years) was associated with a 41.2% increase in the likelihood of completing colonoscopy, adjusting for insurance (Table 2; Figure 1).

TABLE 2.

Logistic model summaries on linear predictor scale with converted odds ratio (OR) and 95% confidence intervals (CI).

Model	Coefficient	Estimate	Std. error	z value	OR	CI	p
Reduced model	Intercept	−0.2077	0.2225	−0.934	0.81	0.52–1.25	0.351
	Age	0.3451	0.2060	1.675	1.41	0.95–2.14	0.094
	Insurance	1.1684	0.5115	2.284	3.22	1.22–9.31	0.022
Reduced model + distance	Distance	−0.19589	0.21366	−0.917	0.82	0.54–1.25	0.359

Note: Baseline level for health insurance status is uninsured. Age and distance are scaled.

FIGURE 1.

Predicted probabilities (solid line) and 95% confidence intervals (dotted lines) for completing colonoscopy from the logistic model on the response scale. Input values for the age covariate spanned the minimum and maximum of our dataset. Circles represent actual data; circles at 0% represent patients who did not receive a colonoscopy, and circles at 100% represent patients who did receive a colonoscopy.

Because the distance to the colonoscopy site was not included in the reduced model, we fitted another logistic model including age, insurance status, and distance to the colonoscopy site as covariates. One standard deviation increase in distance (approximately 38 miles), was associated with a 17.7% decrease in the likelihood of completing colonoscopy, adjusting for age and insurance (Table 2; Figure 2).

FIGURE 2.

Predicted probabilities (solid line) and 95% confidence intervals (dotted line) for completing colonoscopy from the logistic model on the response scale. Input values for the distance from the participant's home to the colonoscopy site covariate spanned the minimum and maximum of our dataset. Input values for the age covariate were held constant at 59.5 years, which is the median of our dataset. Circles represent actual data; circles at 0% represent patients who did not receive a colonoscopy, and circles at 100% represent patients who did receive a colonoscopy.

4. Conclusion

In this sample, half of the participants with a positive screening FOBT subsequently completed colonoscopy. Institutions utilizing FOBT for CRC screening should vigilantly support those with positive results to ensure colonoscopy completion. It is likely that some of the same barriers leading institutions to choose FOBT over colonoscopy for screening will present challenges for patients who need colonoscopies after positive FOBT. These may be compounded by other challenges specific to individuals.

Having insurance was the strongest predictor of completing colonoscopy after FOBT, despite the existence of financial assistance programs. Those with insurance were more than twice as likely to complete colonoscopy as those without it. Nearly all insured participants in this sample had Medicaid, underscoring its importance for under‐resourced patients to access guideline‐recommended preventive care. Medical center financial assistance programs are a necessary but insufficient support for uninsured individuals to access specialty care.

Increasing age was also associated with an increased likelihood of completing colonoscopy. This cannot be explained by Medicare acquisition, given few patients with Medicare (3.1%) and the equal likelihood of colonoscopy completion or not among those with it. Further study could elucidate this finding, including how the prioritization of colonoscopy relative to competing interests may change with age.

Although not statistically significant, further distance traveled to obtain colonoscopy was a barrier to its completion. Nearly one‐quarter of participants were referred to a colonoscopy site 115 miles from their primary care location, despite living in an area with a population density of gastroenterologists higher than the national average. This highlights the need to incentivize specialists to care for un‐ and underinsured patients.

On average, almost 8 months elapsed between positive FOBT results and colonoscopy. Some of this delay may be explained by financial assistance applications, however the difference between mean time elapsed between those with insurance versus those without was only 20 days. One way to decrease delays may be “fast‐track” colonoscopy referrals, in which patients are referred directly to colonoscopy procedures without preceding gastroenterology office visits. For the majority of participants with visits before colonoscopy, no additional needs were addressed. One participant, who traveled 101.5 miles for a visit at which no additional needs were addressed, did not return for colonoscopy. These findings support further study of “fast‐track” colonoscopy as a potential way to decrease barriers to timely completion.

In this study, closed‐loop communication between colonoscopists and referring providers was excellent. Nearly all colonoscopy and pathology reports were filed in the CHC's EMR. We did not ascertain whether this was the result of high‐quality workflows by staff at the CHC, at colonoscopy sites, or both.

This study has several strengths. EMR data were available from three institutions. Chart review resulted in largely complete data acquisition. Because this study includes participants from one CHC in a metropolitan area, it may have limited generalizability to other populations or geographical regions.

This preliminary work exposes important gaps in CRC screening for CHC patients. It highlights that CRC screening programs utilizing FOBT should support patients with positive results to ensure subsequent colonoscopy completion. It demonstrates the importance of Medicaid in helping under‐resourced patients access preventive care, and the insufficiency of financial assistance programs for uninsured individuals to do so. Finally, it highlights the barrier of travel distance to specialty care for CHC patients, even among those in a large metropolitan area. Future research should gather data from more CHCs and include qualitative interviews to assess ways to help patients overcome individual and system‐level barriers to colonoscopy completion. Research is also needed to elucidate the existence of financial assistance programs among gastroenterology practices and possible incentives to promote them.

Conflicts of Interest

The authors declare no conflicts of interest.