Colorectal cancer screening among African American church members: A qualitative and quantitative study of patient-provider communication

As part of the formative data collection in early 1998, we conducted six focus groups of African American adults. There were three focus groups of African American adult men and three of African American adult women. The focus groups comprised of a convenience sample drawn from members of African American churches located in central North Carolina. The focus groups were conducted in the churches, lasted approximately 60 minutes, were tape recorded, transcribed, and reviewed for accuracy. A $15 incentive and refreshments were provided for the participants. The focus groups were led by sex-matched African American moderators and explored issues associated with colon cancer, CRC screening (barriers and motivators), nutrition, and physical activity. The information from the focus groups was used to develop the questionnaires and to help guide the intervention strategies used in the culturally appropriate colon cancer prevention program.

The second source of participants for the survey study was the baseline intervention sample for the larger WATCH project. The telephone survey collected information focused on general health status, nutrition, physical activity, patient-provider communication, CRC risk knowledge and screening behaviors. The baseline survey was completed, on average, in approximately 40 minutes and was administered by trained interviewers prior to the intervention. The baseline telephone surveys were conducted between October 1998 and October 1999.

In this study, a history of CRC screening was defined as the self-report of undergoing a fecal occult blood test (FOBT), flexible sigmoidoscopy, or colonoscopy within the recommended time period. Participants were asked whether they had each of the CRC screening tests, and if they responded yes, the participants were asked when they had their last test. The items in the survey included a brief explanation of each screening test. Items were described as follows: FOBT, "which is stool slides"; sigmoidoscopy, "which is a tube inserted in the rectum to look at the colon and the bowel"; and colonoscopy, "which is a tube inserted in the rectum to look at the entire intestine, usually given in a hospital or specialist's office." Responses included, "less than 1 year"; "1–2 years"; "3–5 years"; or "more than 5 years."

CRC screening was considered to be within the recommended time period based on an algorithm using ACS guidelines considering which test and how recent the test was performed (e.g. a person who reported having FOBT within the past year was considered within the recommended time period). Participants were considered to have been screened within the recommended time period if they had a FOBT within the preceding year, and sigmoidscopy within the preceding 5 years. Although current CRC screening guidelines recommend colonoscopy every 10 years for average-risk adults, we examined colonoscopy use in the past 5 years because of the limitations of the survey instrument. Self-report of CRC screening behavior has been demonstrated to be a reliable endpoint for intervention trials [22].

Analyses of data from this study included factor analysis, analysis of variance, and logistic regression and were conducted using SPSS, version 10.1. Logistic regression analyses were performed to evaluate whether the level of perceived patient-provider communication was significantly related to CRC screening behavior in this population. Sociodemographic variables were identified as potential covariates if there was plausible theoretical or empirical evidence that the variable might be associated with the communication variable or with CRC screening. Variables that were significantly associated with communication level (p < 0.05) were retained and tested as covariates in the logistic models. Only the sex of the participant, receiving healthcare at a doctor's office versus a clinic/emergency room, and knowledge of CRC risk were significantly associated with communication and only these three covariates were entered into the initial logistic model. Variables in the model were evaluated by the Wald test and interpreted using odds ratios and confidence intervals. The overall fit of each model was evaluated using the Hosmer-Lemeshow Goodness of Fit test and by examining classification tables [23].

Several major themes associated with CRC emerged from the personal experiences expressed about the medical community during the focus groups. One of the themes discussed by the participants in each focus group was patient-provider communication. Selected comments by focus group members about CRC and patient-provider communication are listed in Table 1. This important theme that emerged during the focus group was addressed in the baseline survey by adding items to the questionnaire that specifically addressed perceived communication with health care providers (Table 2).

Originally, 2480 names were obtained from the 12 church rosters located in five rural counties in North Carolina. Many members were ineligible (not 18 years old, deceased, no longer living in the state, medically incapable, phone number no longer working) or we were unable to contact them by telephone, and 239 members who declined to participate in the WATCH Project. There were 850 church members who participated in the WATCH Project and completed the baseline survey. The adjusted response rate was 66% using a calculation method, suggested by the Council of American Survey Organizations (CASRO) [21], that accounted for individuals whose eligibility and response status were unknown because program staff were never able to contact them.

The participants in this study were the 397 church members who participated in the WATCH Project and were 50 years and older. The characteristics of the 397 participants are shown in Table 3. Participants were mostly female (74%) and African American (98%). The mean age was 63 years (SD = 9.7). About half of the sample was currently married, 25% were widowed, and 14% were divorced. Thirty-seven percent had less than a high school education, 30% had a high school diploma or GED, 16% had some college or trade/beauty school, and 18% had a college degree or post-college education. Household income was answered by only 52% of the participants, and of the responders, 51% reported an income of less than $20,000.

Factor analysis of the five communication items was performed from the baseline survey responses and two factors were identified; one with three items and the other with two items. The second factor was dropped because it had only two items and did not add reliability to the scale. The three communication items about shared decision-making and patient satisfaction demonstrated good reliability (Cronbach's alpha = 0.74) and were summed to calculate a communication score. The communication score was used to categorize the participants into three groups: good, fair, and poor communication with providers. Participants were categorized as having "good" communication if they perceived receiving enough information from their provider, being involved in medical decisions, and thinking that their provider understood their health needs almost all the time or always. Participants who rated all three items 'sometimes', 'rarely', or 'never' scored "poor" on the communication scale, and individuals who rated the items with a mix of the above listed responses were assigned to the "fair" group.

In terms of quality of communication, 75% (298/397) responded positively to all 3 questions and were considered have "good" communication; 10% (42/397) responded positively to none of the 3 questions and were considered to have "poor" communication; and 13% (50/397) had fair results. Participants in the good communication group were more likely to be female (p = 0.031), and were more likely to receive their healthcare at a doctor's office versus a clinic/emergency room/health department (p = 0.028). None of the other sociodemographic factors listed in Table 3 appeared to vary significantly among communication groups. Participants categorized in the good communication group were more likely to report having been screened for CRC in the recommended time period compared to those in the poor communication group (35.9% vs. 16.7%; OR = 2.8, CI 1.2, 6.4, p = 0.013).

Only 45% (175/389) of the participants reported that their providers had recommended CRC screening, and just 31% (120/389) of all participants reported being screened within the recommended time interval. Of the individuals who reported being screened, 65% (78/120) stated that their doctor had recommended CRC screening, compared with 36% (97/269) of those who did not report screening.

Knowledge of CRC was assessed using seven items (Table 4) with a mean correct response of 3.8. If the participants answered at least four out of the seven items correctly, they were categorized as having adequate knowledge about colorectal cancer. The participants were considered to have inadequate CRC knowledge if they answered incorrectly or 'don't know' to ≥ 4 of the 7 items.

Knowledge about CRC was considered adequate (knowledge score > = 4) for 57% (228/397) and inadequate for 43% (197/397). Participants with adequate CRC knowledge were more likely to have completed a CRC screening test within the recommended time period compared to those with inadequate CRC knowledge (21% vs. 10%). Adequate knowledge was associated with a higher level of education (p < 0.001), a higher level of income (p < 0.001), having health insurance (p < 0.001), and having Medicare/ Medicaid as one's health insurance (p < 0.001).

Results of the logistic regression analyses are shown in Table 5. Results were similar when using communication and CRC risk knowledge as continuous exposure variables, and when using a history of CRC screening anytime in the past as the outcome variable (instead of recent screening). For ease of interpretation, we chose to present the categorical analyses and use recent screening as the outcome of interest. After adjustment for the sex of the participant and source of healthcare, quality of communication remained significantly associated with completion of a CRC test.

CRC screening within recommended guidelines by perceived communication and knowledge is listed in Table 6. The poor and fair communication groups were combined because of the small numbers within each category. Adequate knowledge is statistically significant for the good communication group but not for the fair/poor communication group. A test for interaction of communication and knowledge was performed for CRC within recommended guidelines and demonstrated no significant interaction.

Although the results from our cross-sectional study have implications for developing colorectal cancer prevention programs, it must be recognized that prospective longitudinal studies are needed that specifically address the effect of patient-provider communication on CRC screening. Because of its cross-sectional design, we cannot infer causal relationships. In addition, since all participants were recruited from 12 churches, there may be some clustering of exposures and outcomes by physician that may affect the results; we did not measure this effect. Patient-provider communication was measured by using only three self-reported communication items, which may not fully capture the full extent of the patient-provider relationship. In addition, we do not know whether our participants had racially discordant or concordant physicians, and we did not have information regarding the physician's beliefs or practices about CRC screening or the physicians' assessment of the quality of communication with the patients. The outcome measure of having undergone CRC screening was self-reported, and these results were not validated. Finally, because of the relatively small number of participants and the use of convenience sampling, results from this study may not be generalizable.