Promoting Colonoscopy Screening among Low-Income Latinos at Average Risk for Colorectal Cancer: A Randomized Clinical Trial

Abstract

Background:

Screening colonoscopy (SC) for colorectal cancer (CRC) is underused by Latinos. This randomized-clinical trial examined the impact of three interventions: 1) patient navigation, 2) patient navigation plus standard CDC print materials, and 3) patient navigation plus culturally-targeted print materials for Latinos referred for SC. Demographic, personal/health history, and psychometric factors associated with SC were also explored.

Methods:

Urban Latinos (N=344) ages 50 to 85 with no personal/immediate family history of CRC before age 60, no personal history of gastrointestinal disorder, no colonoscopy in the past five years, with insurance coverage, and a referral for SC were consented. Participants were randomized to patient navigation (20%), patient navigation plus standard CDC print materials (40%), and patient navigation plus culturally-targeted print materials (40%). Completion of SC was assessed at 12 months.

Results:

The interventions had an overall SC rate of 82%. Counterintuitively, patients with an average income less than $10,000 had higher SC rates (87%) than those with greater income (75%).

Conclusion:

The addition of standard or culturally-targeted print materials did not increase SC rates above that of patient navigation. Indeed, after controlling for other variables, culturally-targeted print materials were associated with lower SC rates for Puerto Ricans.

Condensed Abstract:

This randomized-clinical trial found that patient navigation led to high SC rates among Latinos. Adding print materials did not increase SC rates above that of patient navigation alone.

Introduction

Approximately 140,250 new colorectal cancer (CRC) cases will be diagnosed in 2018, making it the third most common cancer among men and women in the United States (US)^[1]. When detected early, CRC can be treated effectively with a five-year 90% survival rate^{[2, 3]}. Its high diagnostic effectiveness and ability to conduct a simultaneous preventative polypectomy make screening colonoscopy (SC) the gold standard CRC screening test^[4]. However, only 39% of CRC cases in the US are detected early with even lower rates among Latinos (38%) who account for over half of the US population growth in the last decade^{[3, 5]}. For example, 47.6% of Latinos over age 50 underwent endoscopic CRC (e.g. SC) screening in 2015 vs. 63.3% of non-Latino whites and 59.3% of African-Americans^{[3, 6]}. The current study was designed to address this health disparity and promote SC among Latinos at average risk for CRC.

Originally suggested by Freeman and colleagues as a method for patients to gain better access to medical services following an abnormal screening test, patient navigation provides a designated person within the healthcare setting to help patients navigate and obtain medical care, including cancer screening^{[5, 7]}. In a recent (2018) review of interventions to increase colorectal cancer screening, navigation, particularly in the context of multicomponent interventions, was associated with increased screening for CRC^[8]. An earlier (2011) review of navigation specifically for SC found that while it increased SC rates overall, with patient navigation, only 36.4% of Latino patients received colonoscopy compared to 44.4% of white patients^[9]. Furthermore, even with patient navigation, national screening rates remain below the National Colorectal Cancer Roundtable (NCCR) goal of 80% by 2018^[10].

Culturally-targeted (CT) print materials have also been explored as a method to boost CRC screening. At the time of study design, compared to a standard intervention, CT interventions were rated more favorably, and were more likely to increase targeted health-promoting behaviors^{[11, 12]}. More recently, a study was conducted of SC for Latino patients receiving culturally-enhanced and standard patient navigation^[13]. Culturally-enhanced navigation did not have a greater impact on SC. However, this study did not utilize culturally-targeted print materials and the extent of cultural-targeting was limited to emphasizing screening importance and barriers for Latinos. The current study examined the impact of three interventions: 1) patient navigation, 2) patient navigation plus standard CDC print materials, and 3) patient navigation plus culturally-targeted print materials in the context of Latinos referred for SC. The study hypothesis was that the patient navigation plus culturally-targeted print materials group would lead to greater SC rates as compared to the other two intervention groups.

The study also explored demographic, personal health history, and psychometric factors associated with motivation for SC. Theory-based studies have the advantage of using systematic approaches to examine cancer screening barriers, including the Health Belief Model (HBM) and the Transtheoretical Model (TTM)^[14–16]. These models propose that people’s beliefs and attitudes are the primary determinants of their decisions and behavior. The present study includes concepts from the TTM (pros/cons), HBM (pros/cons, perceived severity), and the Dual-Process Model (worry and fear) because of their wide application to cancer screening and associations identified in previous research^[14–18]. The HBM proposed that behavior for disease prevention or screening test for early detection depends on the: “1) desire to avoid illness …and 2) the belief that a specific health action will prevent (or ameliorate) illness….”^[14–18] The HBM constructs include: perceived susceptibility, severity, benefits and barriers. Although perceived risk is usually calculated as perceived susceptibility plus perceived seriousness in the HBM, in this study, items measuring perceived risk were from the Health Information National Trends Survey^[19]. In addition, cancer fatalism (i.e., belief that detection and diagnosis will result in death) has been associated with reduced routine cancer screening and is examined in the present study^[20]. Data were limited as to whether optimism relates to SC, so it was included in this study as an exploratory variable and potential confounder; however, our prior analyses for this study found optimism to be associated with lower SC^[21]. Acculturation (i.e., process by which individuals adopt another culture’s attitudes/values/customs/beliefs/behaviors) is a socio-cultural variable that was expected to moderate the intervention impact^[22]. For Latino women, greater acculturation was associated with undergoing a clinical breast exam after program completion^[23].

Methods

Study Sample and Recruitment

Three-hundred and eighty-six individuals were approached between May 2012 and December 2013 in an IRB-approved randomized clinical trial (RCT) R01 CA140737–01A2 “Improving CRC Screening for Diverse Hispanics in an Urban Primary Care Setting”; ClinicalTrials.gov identifier: NCT01569620) at the Icahn School of Medicine at Mount Sinai’s primary care facility, a health center within an urban general hospital providing services to low-income individuals. Eligibility criteria included: 1) self-identified Latino; 2) ages 50–85; 3) no personal history of CRC and no immediate family history of CRC before age 60; 4) no history of gastrointestinal disorder; 5) no colonoscopy procedure in last five years; 6) colonoscopy referral by PCP; 7) English or Spanish-speaking; 8) having a telephone/cell phone service; 9) having insurance coverage; and 10) being navigated. The eligibility criteria were designed to recruit participants of average risk for CRC and eliminate insurance and PCP referral as barriers. It is important to note that as per the participating institution’s clinical practice, participants were referred for screening colonoscopy if they had not had a colonoscopy within the past five years.

The protocol can be obtained from the study authors. Individuals were assigned to a patient navigator after receiving a SC referral from their primary care physician (PCP), then approached for study enrollment while still at clinic. Of those referred, 344 (89.1%) eligible individuals agreed to participate and consented. Three interviews were conducted by trained bilingual research assistants: 1^st assessment (baseline, T1), 2^nd assessment (3 months post-baseline/ approximately 2 weeks before colonoscopy, T2), and 3rd assessment (6 months post-baseline or 1 month after colonoscopy, T3). In the present study, data from the 2^nd and 3^rd assessment were not included. Respondents received $20 for each interview and SC was provided at no cost to the participants. Medical chart reviews were conducted 12 months after the colonoscopy appointment, or 12 months following baseline for those who did not schedule a colonoscopy. Consented participants were randomized via permuted blocks conducted by Winkel to patient navigation (20%), patient navigation plus standard CDC print materials (40%), and patient navigation plus culturally-targeted print materials (40%). Research assistants were blind to randomization assignment until after completion of baseline interview. After randomization to either of the two print conditions, participants were giving the appropriate print materials. Patient Navigation followed procedures used in prior work and included three telephone calls: 1) initial scheduling call, which occurred following the consent process, 2) follow-up call two weeks before scheduled procedure, and 3) final appointment reminder 3 days prior to procedure^[13]. Participants who never received patient navigation services (N=40) were assessed for comparability but excluded from analysis (see Figure 1).

Figure 1:

Consort Figure

Development of Print Materials

We conducted six focus groups with a total of 34 participants (equal number of men and women). We used our prior research to develop and refine this study’s culturally-targeted print materials. As previously reported, focus group analyses confirmed that brochures were culturally relevant for diverse, low-income Latinos and were motivating for SC^[24]. Culturally-targeted print materials also addressed four culture-based factors associated with cancer screening: acculturation, medical mistrust, fatalism, and ethnic identity. Experiences of peers who were ethnically, culturally, and socioeconomically similar to the study population were included in the materials to bolster messaging^[25][26]. Elements of family importance and Latino culture were also included.

Measures

Socio-demographic data, personal and family cancer history, and psychometric measures were collected at baseline. Imputation was used for 14 cases with missing income data. Spanish-language versions of the measures were used from our previous study of 400 Latinos with the exception of the risk and optimism measures. The questions of risk (e.g., “Compared to the average (man/woman) your age, would you say you are…?” Rated from 1 “More likely to get colorectal cancer” to 3 “About as Likely”) were those used in 2005 HINTS and the Spanish-language version of the items that measure optimism (e.g., “I am always optimistic about my future” rated from 1 “I Disagree a lot” to 5 “ I agree a lot”) was already published^[27].

Acculturation was studied with the 12-item Marin Acculturation scale, modified to include a question regarding language of PCP^[28].

Psychosocial measures included: fear of colonoscopy (e.g. “How fearful are you of... The procedure being painful” rated from 1 “Not at all fearful” to 5 “Extremely fearful’); fatalism (e.g. “ I believe that is someone gets colorectal cancer, his/her time to die is near”, rated “Yes” or “No”.; worry (e.g., “I am afraid of having an abnormal colonoscopy result” rated from 1 “Strongly disagree” to 4 “Strongly agree”; perceived preventability and curability (e.g., “Colorectal cancer can be prevented” rated from 1 “Strongly disagree” to 5 “Strongly agree”; perceived risk; pros/cons (e.g., “ A colonoscopy can find growths that are not yet cancer but could become cancer” rated from 1 “Strongly disagree” to 5 “Strongly agree” ); optimism (Life Orientation Test – Revised (LOT-R or the Spanish LOT-R)^{[14, 19, 21, 27, 29–32]}. SC (the primary outcome) was obtained from medical chart review.

Data and Power Analysis

Statistical analyses were executed using SAS 9.4 software package. Analyses are based on a type I error rate of 0.05 and two-sided. Individual baseline characteristics, patient navigation status, and print groups (e.g. CDC standard or CT) were assessed for unadjusted association with SC rates through a series of logistic regression models and Chi-square analyses. Differences in distributions of baseline characteristics with randomization groups, among those not navigated, were similarly assessed.

SC rates (primary outcome) were modeled on print group with adjustment for all covariates found significantly associated with SC in unadjusted analyses of baseline socio-demographic, cancer history, and psychosocial variable described above. Optimism (LOT-R), which has been reported elsewhere as being associated with outcomes in this study was also included as a potential confounder for adjustment^[21]. Interaction effects with significant covariates were tested via an overall test with the 3-level print group variable. Some categories for education, marital status, and language were combined based on unadjusted results and practical implications. For interaction effects, stratified models were used to assess the relationship of SC rates to print group within levels of the interaction variable.

Due to the novelty of the culturally-targeted print materials and the population (understudied, low-income Hispanics), power was based on our own and prior research^[33]. The latter research indicated that, compared to a standard condition, a targeted message increased the intention of Hispanics to discuss kidney disease with their physician, with a moderate effect size (d = 0.55) is consideredmoderate. Given the assumption of a moderate effect size on completion of SC, and using our own research which found patient navigation was associated with a screening rate of 40%, we planned for a moderate increase of 20% in screening completion. Under these conditions, the comparison of patient navigation (n = 73) to culturally-targeted print materials (n = 141) would yield power = 0.90 (alpha = 0.05) using an arcsine transformation. The third arm, standard CDC print materials did not affect the power for comparison of culturally-targeted to navigation only arms.

Results

Demographics.

Participants’ average age was 60, 54% had less than a high school education, 56% lived in households with annual income under $10,000, 42% spoke only Spanish, and 90% were receiving public insurance (See demographics in Table 1).

Table 1.

Participant baseline characteristics (N=304)

Characteristic	N	Navigation Only, % (n = 66)	CDC Standard Print, % (n = 116)	Cultural Targeted Print, % (n = 122)
Age
50 – 65	217	64	72	75
> 65	87	35	28	25
Gender
Male	115	30	43	37
Female	189	70	57	63
Nativity
Born in US	93	33	35	25
Born in Puerto Rico	105	32	30	40
Born outside US	105	35	34	35
Not reported	1	0	1	0
Education
<6 years	55	29	16	15
6–12 years	108	27	35	40
High School	81	26	24	30
Some College or higher	58	17	24	16
Not reported	2	2	1	0
Annual household income
<$10,000/year	172	58	53	60
≥$10,000/year	132	42	47	40
Family History of CRC Prior to Age 60
Yes	14	6	3	6
No	290	94	97	94
Any relative with a cancer diagnosis
Yes	162	59	45	58
No	142	41	55	42
Previous cancer diagnosis
Yes	32	11	8	13
No	271	89	92	86
Unknown	1	2	0	0
Language
Only Spanish	128	33	41	48
Spanish>English	29	9	10	9
Both Equally	68	30	22	18
English>Spanish	37	14	12	11
Only English	42	14	14	14
Insurance
Private	31	2	15	11
Public	273	98	85	89
Marital Status
Married	100	29	35	33
Divorced	37	14	12	11
Single	138	47	44	45
Widowed	29	11	8	11
Fatalism, Mean	289	0.32	0.27	0.29
Fear, Mean	304	1.76	1.98	1.20
Worry, Mean	304	2.46	2.44	2.52
Pro Con–DB, Mean	304	2.18	2.25	2.16
Perceived Risk, Mean	304	1.75	1.76	1.83
Acculturation, Mean	304	2.69	2.59	2.53
Perceived Curability, Mean	304	4.21	4.36	4.33
Perceived Preventability, Mean	304	3.95	4.12	4.07

Differences in navigated versus not navigated.

Of 344 consented patients, 40 did not receive patient navigation services and were thus ineligible, resulting in 304 patients for the main analysis. Among those navigated, 82% completed SC within one year of referral; 8% of those that could not be navigated completed SC within the same time frame (²(1) = 95; p < .001). None of the non-navigated patients had a personal or family history of colorectal cancer, compared to 11% of navigated patients with a personal cancer history (P = .036) and 4% with a family colorectal cancer history over the age of 60 (P = .062). Navigated and non-navigated patients did not differ on other measured characteristics, though there were differences in gender distribution by randomization arm within the non-navigated patients. Specifically, of 50 men randomized to CDC print materials, all 50 (100%) were navigated, compared to 83% navigation among the men randomized to the other two groups (P = .002).

Differences between print groups.

Patients were randomized to one of three groups: culturally-targeted print materials (n=122); standard CDC print materials (n=116), or no print materials (n=66). All three groups received patient navigation. The groups differed slightly on insurance type; only 2% were privately insured in the patient navigation group versus 15% and 11% for standard CDC print materials and culturally-targeted print materials, respectively (P = .019). There were no other observed differences between groups (Table 1).

Unadjusted relationship of baseline characteristics to SC rates.

Completers (N=248) and non-completers (N=56) differed by income group, nativity, and perceived risk (Figure 2). Patients with an annual household income less than $10,000 had an 87% rate, higher than the 75% rate for patients with household incomes over $10,000 (P = .011). Overall, US- and Puerto Rican-born patients had a lower rate (77%) than those born elsewhere (90%). Completers had slightly higher perceived risk (mean = 1.83) than non-completers (mean = 1.61). Rates did not differ by group; 82% of the culturally-targeted print materials, 81% of the standard CDC print materials group, and 82% of the patient navigation group completed SC. Unadjusted associations of baseline characteristics with SC rates are in Table 2.

Figure 2.

Unadjusted SC Rates by print group, nativity, household annual income, and perceived risk. Optimism and perceived risk are dichotomized using median splits. *P<.05

Table 2.

Unadjusted odds ratios (OR) between baseline characteristics and SC Rates

Characteristic	OR	95% CI
Group, Cultural print1	1.06	(0.6, 2.0)
Group, Navigation only1	1.05	(0.5, 2.0)
Age	0.97	(0.9, 1.0)
Gender, Female	1.18	(0.7, 2.1)
Nativity*, Puerto Rico	0.98	(0.5, 1.9)
Nativity*, outside US	2.49	(1.1, 5.5)
Education, 6–12 years	0.60	(0.2, 1.5)
Education, High School	0.55	(0.2, 1.4)
Education, Some College	0.70	(0.2, 2.0)
Income*, ≥$10,000/year	0.46	(0.3, 0.8)
Family History of CRC	0.55	(0.2, 1.8)
Any relative with cancer	0.76	(0.4, 1.4)
Previous cancer diagnosis	1.25	(0.5, 3.4)
Language, Only Spanish	1.36	(0.6, 2.9)
Language, Spanish>English	0.74	(0.3, 2.1)
Language, English>Spanish	0.86	(0.3, 2.3)
Language, Only English	0.87	(0.3, 2.2)
Insurance, Private	1.20	(0.4, 3.3)
Marital status, Divorced	0.46	(0.2, 1.2)
Marital status, Single	0.64	(0.3, 1.3)
Marital status, Widowed	0.39	(0.1, 1.1)
Fatalism	1.06	(0.5, 2.5)
Fear	0.93	(0.7, 1.2)
Worry	1.37	(1.0, 1.9)
Pro Con–DB	0.77	(0.6, 1.0)
Perceived Risk*	1.81	(1.1, 3.0)
Acculturation	0.86	(0.7, 1.1)
Perceived Curability	0.79	(0.5, 1.3)
Perceived Preventability	0.90	(0.6, 1.3)

¹The reference groups are as follows: for group is CDC standard print, nativity is birth in the US, for education is less than 6 years, income is <$10,000/ year, language is both equally, insurance is public, marital status is married.

^*P<.05

Adjusted relationship of baseline characteristics and print group to SC.

A fully-adjusted model was fitted to assess the effects of print group (e.g. CDC standard or CT), income, country of birth, perceived risk, and LOT-R on SC. Lower income (adjusted odds ratio [aOR], 0.48; 95% CI, 0.26–0.90) and foreign birth (aOR, 2.69; 95% CI, 1.18–6.13) remained associated with SC after adjustment. LOT-R also showed a moderate inverse association (aOR, 0.91; 95% CI, 0.84–0.99). The association of perceived risk was not significant after adjustment. Adjusted association of all important baseline characteristics with SC are in Figure 3. Although the primary analysis per protocol was to exclude patients who never received navigation, a sensitivity analysis including the non-navigated participants to assess impact of the inclusion criterion found comparable results, though the adjusted LOT-R association was no longer significant (P = .066).

Figure 3.

Adjusted odds ratios (OR) with confidence intervals for model of SC rates. Optimism (LOT-R) was also included as a covariate for adjustment due to findings being presented previously. **P<.01; *P<.05

Print group, overall, was not associated with SC after adjustment. However, an interaction did exist between nativity and print group (P = .022), but not the acculturation measure; the interaction with nativity was also sustained in the sensitivity analysis. A subset analysis of participants born in Puerto Rico revealed that, even after adjustment for LOT-R and income, culturally-targeted print materials was associated with lower SC than standard CDC print materials (P = .032; aOR, 0.75; 95% CI, .07-.85). Specifically, among Puerto Ricans, SC rates were 86%, 67%, and 86%, for CDC standard, CT, and no print materials, respectively.

Discussion

This randomized clinical trial examined SC for 304 Latino patients randomized to three arms: patient navigation, patient navigation with standard CDC print materials, and patient navigation with culturally-targeted print materials. Results indicated no support for the study’s primary hypothesis: SC did not differ among study arms. Adding culturally-targeted print materials to patient navigation did not increase SC rates.

Patient navigation was associated with SC rates in this sample of Latinos. The high SC rate (82%) among navigated patients in this study (n=304) compared to the average rate among Latinos (48%), regardless of type of print materials, points to the effectiveness of patient navigation in promoting SC as well as lack of effect of adding print educational materials. Prior research has identified patient navigation’s positive impact on CRC screening; the present study confirmed navigation’s efficacy in promoting SC^{[13, 34, 35]}. There was minimal variation in SC for participants by language preference, highlighting the value of bilingual navigation resources. Non-navigated participants were not navigated because they were lost to follow-up or withdrew from study (n=40). The low SC rate for non-navigated participants (8%) likely self-selects for those most averse to SC with the greatest barriers to overcome. The disparity between navigated and non-navigated groups highlights the need to develop interventions to facilitate participation in patient navigation.

Although there were no differences in the study groups regarding SC, socio-demographic or psychometric factors that could predict SC were investigated. Overall, completers and non-completers differed on income in the unadjusted model. Eighty-seven percent of participants with an income below $10,000 completed SC, while 75% of participants with income over $10,000 completed SC. The reasons behind this counter-intuitive discrepancy are unclear. Lower-income participants may face greater barriers to healthcare and been more apt to take advantage of the study, thus enhancing their experience compared to higher-income participants (this analysis excludes 4 participants without insurance; type of insurance did not influence SC rates). They may also be more cost-conscious and SC in the study context may have provided assurance that no unexpected medical bills would arise. There were no other demographic differences found between completers and non-completers.

In a fully-adjusted model, income remained a SC predictor despite the limited range of participants’ income. After adjustment for LOT-R and income, print group was associated with SC among Puerto Ricans (P = .032), such that culturally-targeted print materials were associated with lower SC rates as compared to standard CDC print materials (OR = 0.75;95% CI = .07-.85). These findings differ from cancer prevention and control studies that examined the effect of cultural targeting interventions in diverse populations which found that compared to a standard approaches, cultural targeting resulted in more positive evaluations and greater acceptability of cancer prevention intervention^[36].

Strengths and Limitations

Strengths of this prospective, study design included focusing on underserved Latinos, measuring SC rates from medical records rather than participants” self-report, and considering many SES and behavioral variables that might impact screening behavior. We acknowledge several limitations: First, results may not be generalizable to Latino populations from other geographic regions whose cultural backgrounds vary from that of the study’s majority Puerto Rican and Dominican populations in East Harlem. Second, results may not be generalizable to other healthcare delivery systems; the Mount Sinai Health System has resources to offer consistent physician referrals for SC and patient navigation, whereas many patients in the US face greater screening barriers and lower healthcare access. All participants in this study had insurance, thus, results may not apply to those uninsured. Lastly, a non-navigated group (n = 40) was not intentionally part of the study, but rather a result of participants being lost to follow up. The SC rate comparison between navigated and non-navigated participants may overstate the actual disparity between the two groups if the non-navigated group is particularly biased against SC. Interestingly, among men there were differences in navigation rates by randomization arm, though the causes of failure to navigate are considered random or administrative and so this finding may be overstated.

An important note is that during this study, efforts to promote CRC screening in New York were increased and the positive impact of navigation on SC is within the context of this larger regional effort^[37]. In addition, willingness to take part of this study indicates an openness to patient navigation for SC.

Future Directions

Individuals receiving patient navigation in this study had SC rates exceeding national averages. Evidence suggests patient navigation is an effective tool for promoting SC. Future work examining the efficacy of navigation among uninsured patients is warranted. The withdrawn/lost-to-follow-up of 40 (12%) individuals from the study demonstrates that a one-size-fits-all approach to patient navigation is not effective. Future studies may benefit by addressing innovations in navigation such as in-person contact along with telephone calls or differing educational approaches.