Abstract
Risk perception may be an important motivator of health-related behaviors. To develop effective risk communication messages, it is important to understand both the patterns of association between perceived risk and health-related behaviors as well as the correlates of risk perception. Very little is known about whether correlates of risk perception are similar in cross-sectional data compared with prospective data. Furthermore, there are scant data on consistency of correlates of risk perception across groups who vary in objective medical risk. If correlates differ, it would underscore the need to tailor intervention messages based on subgroup characteristics as well as increase awareness of the limitations of basing intervention messages only on cross-sectional data.
We analyzed data on a subset of 5042 employees who participated in The Next Step Trial, a randomized health promotion trial to encourage colorectal cancer screening and dietary change. We restricted our analysis to only those automotive workers who were white, male, and did not have colorectal cancer (4477/5042) and who returned surveys both at baseline (2,684/4,477) and at year 2 of follow-up (1955/2684). Initial analyses detected interactions between a history of polyps and several of the other covariates. Therefore, univariate and multivariable analyses were conducted separately for men with and without a personal history of colorectal polyps. Within each of the four subgroups (those with or without polyps in the baseline or follow-up analyses), we examined associations between perceived risk measured at baseline(cross-sectional analyses) and at year 2 of follow-up (prospective analyses) in relation to intervention group status, demographic,medical history, psychosocial, and worksite characteristics measured at baseline. To assess the predictive ability of the models, we computed sensitivity and specificity as measures of each model’s ability to correctly classify men into their respective subgroup.
Although there was no association between perceived risk and intervention group status in the four subgroups analyzed, we included intervention group status as a covariate in all analyses. At baseline(cross-sectional analyses) among men with and without a history of polyps, perceived risk was positively associated with family history of colorectal polyps or cancer, family support for screening, and worry about being diagnosed with colorectal cancer. In addition, for men without polyps, perceived risk was positively associated with being a current smoker. At year 2 of follow-up (prospective analyses) for men with and without polyps, perceived risk at year 2 was positively associated with family history and baseline perceived risk and was negatively associated with having a normal screening examination or no examinations during the trial. In addition, for men with polyps,perceived risk was positively associated with belief in the salience and coherence of screening and with intention to be screened and was negatively associated with access to screening at the worksite. Specificity was higher than sensitivity in three of four subgroups and was >65% in all subgroups.
Except for family history, messages to influence perceived risk would emphasize different factors, depending on whether associations were based on baseline or follow-up data and depending on whether men reported a personal history of polyps. For example, although intervention messages using baseline data would emphasize the same factors for men with or without polyps, messages based on follow-up data would emphasize psychosocial characteristics, such as salience and coherence of screening and intention for men with a history of polyps but not for men without. Our findings support the need to delineate subgroups in the study population to target and tailor health-related messages based on respondent characteristics. Our findings also underscore the need to base health-related messages on prospective data as well as cross-sectional data to better address health-related beliefs and behaviors.
Introduction
Perceived risk or susceptibility is a central construct in a number of theories or models of health behavior [e.g., the health belief model (1), the precaution adoption process (2, 3), the self-regulation model of health behavior (4, 5, 6), the transactional model of stress and coping (7, 8), and protection motivation theory (9). Risk perception derives from threat appraisal, which is considered to be an important motivating factor in health behaviors. Threat appraisal is based on beliefs about disease risk and disease severity (10). As defined by Weinstein and Klein (11),perceived risk is one’s belief about the likelihood of personal harm.
Because risk perception may be an important motivator of a number of health-related behaviors, it is important to understand both the patterns of association between perceived risk and specific health-related behaviors as well as the correlates of risk perception. Correlates of perceived risk may differ depending on whether the outcome, i.e., perceived risk, is measured using baseline(cross-sectional) or follow-up (prospective) data. For example,cross-sectional data may inform our understanding of factors associated with prior screening behavior but may be misleading if the goal is to develop interventions to encourage future screening, particularly repeat screening. Likewise, correlates of a behavior may vary for subgroups with different medical or screening histories, such as a personal history of colorectal polyps or prior experience with screening. In a similar vein, correlates of perceived risk may vary depending on how data were collected and on personal characteristics.
There is increasing interest in developing and evaluating educational messages to change risk perceptions (12). An understanding of the correlates of perceived risk can inform the development of risk communication messages that encourage adoption of health behaviors. In a recent literature review on risk perception and risk communication for cancer screening behaviors, Vernon (12) noted that all but two studies3(13) of correlates of perceived risk for cancer used a cross-sectional study design (14, 15, 16, 17, 18, 19, 20, 21, 22, 23). Five studies3 (14, 15, 17, 23) were of persons at increased risk based on a family history of cancer. Very few studies examined psychological or psychosocial correlates of perceived risk,and only two (15, 20) examined correlates of perceived risk for colorectal cancer. One (15) compared siblings of colorectal cancer patients with siblings of general surgical patients,and the other (20) studied older, predominantly African-American, clinic users. Data on the consistency of correlates of perceived risk across different population subgroups would inform our efforts to develop tailored intervention messages.
In this report, we used a subset of data from a large cohort of current and former automotive employees participating in The Next Step Trial (24) to examine baseline (cross-sectional) and year 2 follow-up (prospective) associations between perceived risk of colorectal cancer or polyps measured at baseline and at year 2 of follow-up in relation to intervention group status, demographic,medical history, psychosocial, and worksite characteristics measured at baseline. The Next Step Trial was a worksite-based health promotion trial of interventions designed to encourage colorectal cancer screening and dietary change (24). The cohort of employees in the trial was at increased risk for colorectal cancer mortality (25) and had been offered periodic screening between 1980 and 1985 through a company-sponsored program. At the time the trial was initiated, screening participation was low.
We addressed the following questions: (a) What factors measured at baseline are associated with perceived risk, also measured at baseline, in men with and without a history of polyps(cross-sectional analyses)? (b) What factors, measured at baseline, are associated with perceived risk measured at year 2 of follow-up in men with and without a history of polyps (prospective analyses)? (c) Are correlates of perceived risk similar in baseline (cross-sectional) and follow-up (prospective) analyses? and(d) What is the predictive ability (i.e.,sensitivity and specificity) of the models of correlates of perceived risk?
Patients and Methods
Study Population.
The study population for this report consisted of a subset of The Next Step Trial cohort. The cohort was comprised of white male employees who were eligible to participate in the trial, responded to both the baseline and year 2 surveys, and did not have colorectal cancer at baseline or develop it during the study period. The baseline and year 2 surveys assessed demographic, medical history, preventive health behaviors, and beliefs about colorectal cancer and screening.
Of 5042 employees in the total cohort at baseline, 4550 (90%) were white men, of whom 4477 were without a history of colorectal cancer prior to or during the trial. Of the 4477, 2684 (60%) returned a baseline survey, 2041 (46%) returned a survey at year 2, and 1955(44%) returned surveys both at baseline and year 2 follow-up. Men with colorectal polyps were included in the trial, but recommendations for follow-up examinations may have differed, e.g., colonoscopy instead of FOBT,4digital rectal examination, or flexible sigmoidoscopy.
Details about the design, implementation, and baseline findings, as well as the primary outcome results from the trial, are published elsewhere (24, 26, 27). Briefly, 28 worksites were randomized to receive either an educational intervention or usual care. Employees at all worksites were offered colorectal cancer screening(i.e., digital rectal examination, FOBT, and/or flexible sigmoidoscopy) on work time. Each worksite developed its own process for offering screening. In addition, employees at the intervention worksites received a mailed invitation to the screening program and an educational booklet tailored to the employee’s screening history and individual screening recommendations followed by a telephone call to reinforce messages from the booklet. The booklet was based on behavior change theories and models (1, 28, 29, 30).
Measures of the Variables.
The dependent variables for this analysis were perceived risk of developing colorectal cancer or polyps as measured on the baseline survey in 1993 and on the year 2 follow-up survey in 1995. Perceived risk was measured using a 3-item, Likert-style scale with four response categories from strongly agree (4) to strongly disagree (1). Items were: I believe that the chance I might develop colorectal cancer is high; I think it is very likely that I will develop colorectal cancer or polyps; and I believe that the chance that I will develop colorectal polyps is high. Cronbach’s coefficient α was 0.79 in the study population. Because scale scores were skewed and some cell sizes were small when the entire range of the scale was used, we dichotomized perceived risk as the belief that one is at risk (scale score ≥3) or not at risk (scale score <3) of developing colorectal polyps or cancer.
Most of the independent variables examined in relation to perceived risk were measured on the baseline survey. Demographic and medical history variables were abstracted from employment records and were augmented by the surveys. In addition to screening history in the 2 years prior to baseline, screening examination status between baseline and year 2 of follow-up was ascertained. Scales were used to measure some constructs, including belief in the salience and coherence of colorectal cancer screening (four items), intention to be screened (two items), perceived self-efficacy related to screening (four items), and worries or fears about being diagnosed with colorectal cancer (two items). Single items were used to measure other variables shown in Tables 1,2,3. All psychosocial variables were measured using the same 4-point scale as for perceived risk. Some items were reverse coded so that high scores corresponded to more of the variable being measured, e.g., more worry, greater intention, more discomfort. Scale scores were standardized by dividing the total score by the number of items in a scale. Because these data were skewed and some cells were sparse when the entire range of the scale was used, all items and scales were dichotomized at ≥3 (strongly agree or agree) or <3(disagree or strongly disagree). Scale development and validation are described in detail elsewhere (31). Because of issues of skewness and kutosis, we also categorized age and education (32).
We surveyed medical personnel at each of the 28 worksites to ascertain characteristics related to delivery of the screening program, i.e., whether screening was offered all year or only during certain months, whether screening examinations (digital rectal examination and flexible sigmoidoscopy) were offered at the worksite or off-site, whether FOBTs were distributed by mail or by other means, and whether the plant medical staff notified employees by mail or telephone to schedule appointments for screening or whether responsibility to schedule an appointment was left to the employee.
Statistical Analyses.
Because this was a randomized controlled trial, intervention or control group status was included in all multivariable models, although it was not statistically significant in univariate analyses. By including the intervention term in the models, we adjusted for even modest intervention effects before drawing conclusions about other factors associated with perceived risk. All analyses were done using SUDAAN statistical software to adjust for the effects of cluster sampling(worksite), taking correlations within worksite into account (33). We first examined the characteristics of respondents and nonrespondents to the baseline and year 2 surveys using P ≤ 0.05 to evaluate the statistical significance of these comparisons. We also conducted preliminary univariate analyses that included respondents to either of the surveys to examine associations between perceived risk measured at baseline(n = 2684 in baseline analyses) or at year 2 follow-up(n = 2041 in prospective analyses) and the independent variables. Results of those analyses were similar to results for the more restricted subset of respondents who answered both surveys(n = 1955). Therefore, analyses were based on 1955 survey respondents.
As noted above, personal history of colorectal polyps might affect risk perception; therefore, we tested for two-way statistical interactions between personal history of polyps and all other independent variables. Because there were a number of interactions, we stratified all analyses by personal history of polyps measured at baseline. Univariate comparisons of perceived risk in relation to the independent variables were tested for statistical significance using Waldχ 2 tests. To assess the relative importance of the independent variables measured at baseline in relation to perceived risk measured at baseline (cross-sectional analyses) and perceived risk measured at year 2 of follow-up (prospective analyses), we conducted multivariable analyses by fitting logistic regression models separately for each time period and for men with and without a history of colorectal polyps. Independent variables with a P ≤0.20 in univariate analyses were included in step one of the logistic regression analyses. In subsequent steps, covariates with P > 0.05 in the prior step were removed until all covariates in the model were statistically significant at P < 0.05. Variables in the models were adjusted for the effects of other variables in the model. Odds ratios and 95%confidence intervals were used to summarize the analyses for the final multivariable models. Categorization of the variables for the multivariable analysis is shown in Tables 1,2,3. The reference groups for calculating the odds ratios are indicated in Table 4.
As one approach to validating our models, we assessed predictive ability. Showing association is not sufficient evidence of prediction;therefore, we assessed the predictive ability of a model by using the coefficients from the model to calculate the predicted probability of having a value ≥3 for each study participant. We then categorized the individuals as <3 or ≥3 using a cutoff value for the probability equal to the observed proportion ≥3. We calculated the sensitivity and specificity of each model and assessed the predictive ability of the models by comparing the classification based on the model to the observed classifications as reported by the participants.
Results
Comparison of Respondent/Nonrespondent Characteristics.
Compared with nonrespondents, respondents to the baseline survey were more likely to be older, married, currently working, not a current smoker, and to have more years of formal education. They also were more likely to have a family history of colorectal cancer or polyps, to have a personal history of colorectal polyps, and to have undergone colorectal cancer screening within the past 2 years. Patterns were similar for the year 2 survey, except that there were no differences in educational attainment between respondents and nonrespondents at year 2 of follow-up. Similar to baseline survey respondents, year 2 survey respondents were more likely than nonrespondents to have been screened between the baseline and year 2 surveys.
Additional data were available from the baseline survey for respondents and nonrespondents to the year 2 survey. Compared with nonrespondents at year 2 of follow-up, respondents were more likely to intend to be screened, to believe in the salience and coherence of screening, to have high self-efficacy related to screening, to believe in the efficacy of screening (marginally significant), to express concern about screening-related discomfort, to be receptive to what family members wanted them to do regarding screening, and to report that family members supported their being screened.
Baseline (Cross-Sectional) Analyses.
In baseline univariate analyses, we observed generally similar patterns of association for men with and without a history of polyps (Tables 1,2,3). There were four exceptions. For men with a history of polyps,age was inversely associated with perceived risk, and education showed a curvilinear pattern of association (Table 1). Among men with a history of polyps, the percentage who scored ≥3 on the perceived risk scale was higher for men with less than a high school education and for those with more than a high school education compared with men who graduated from high school (Table 1). For men with polyps, perceived risk was positively associated with plant notification that FOBT was due and was negatively associated with offering screening at the worksite (Table 3). Those four associations were not observed in the group without polyps.
In baseline multivariable analyses, with the exception of smoking status, the same variables were positively associated with perceived risk for men with and without a personal history of polyps, i.e., family history, support for screening from family members, and fear or worry about being diagnosed with colorectal cancer(Table 4).
Year-2 Follow-Up (Prospective) Analyses.
In general, the univariate results of the year 2 follow-up analyses were similar for men with and without a personal history of polyps(Tables 1,2,3). There were four exceptions. In men without a history of polyps, perceived risk was positively associated with being a current smoker (Table 1). It also was positively associated with belief in the efficacy of screening and with perceived self-efficacy related to screening (Table 2). Although these patterns were similar for men with polyps,associations were not statistically significant. Among men with polyps,perceived risk was negatively associated with offering screening at the worksite (Table 3).
In multivariable analyses, family history and baseline perceived risk were positively associated, and having no examinations or normal examinations was negatively associated with perceived risk at year 2 follow-up in men with and without polyps (Table 4). In addition, for men with a history of polyps, salience and coherence and intention to be screened were positively associated and whether the plant offered examinations on site was negatively associated with perceived risk.
Comparison of Baseline (Cross-Sectional) and 2-Year Follow-Up(Prospective) Analyses.
In all univariate comparisons, both baseline and year 2 follow-up analyses, the percentages of men who scored ≥3 on the perceived risk scale were greater for men with a history of polyps compared with men without a history of polyps. The percentage of men who scored ≥3 on the perceived risk scale decreased between baseline and year 2 of follow-up in virtually all univariate comparisons. In multivariable analyses, family history was the only variable that was consistently associated with perceived risk in all analyses. Differences between baseline and year 2 follow-up analyses were more pronounced for men with a history of polyps, where fewer psychosocial variables were associated with perceived risk in baseline compared with year 2 follow-up analyses (Table 4). This pattern was not observed for men without a history of polyps.
Sensitivity and Specificity Analyses.
The sensitivity, i.e., the percentage of men who scored ≥3 on the perceived risk scale and were correctly predicted by the model as scoring ≥3, was higher in year 2 of follow-up than in baseline analyses for men with and without a history of polyps (Table 5). For each type of analysis, i.e., baseline and year 2 follow-up, the estimates were similar for men with and without a history of polyps.
With the exception of the year 2 follow-up analyses for men without a history of polyps, specificity, i.e., the percentage of men who scored <3 on the perceived risk scale and were correctly predicted as reporting <3, was similar across subgroups of men with and without a history of polyps (Table 5).
Discussion
Only two other published studies (15, 20) examined correlates of perceived risk for colorectal cancer, and both used a cross-sectional study design. There was very little overlap in the independent variables examined in our study and theirs.
Our finding that family history of colorectal cancer was consistently positively associated with perceived risk was similar to findings reported in the breast cancer literature. We did not directly assess participants’ knowledge of colorectal cancer risk factors in any of the surveys. However, the patterns in the data showed that men with a family history of colorectal cancer or a personal history of polyps were more likely than men without those risk factors to perceive themselves at risk, indicating an awareness of their objective medical risk of developing the disease. Although family history is an important factor to emphasize in developing intervention messages, it is relevant only for a minority of the general population. Therefore, there is a need to identify other factors that can be addressed in interventions.
Our focus on correlates of perceived risk assumes that those associations are important because perceived risk is related either directly or indirectly to behavior, in this case participation in colorectal cancer screening. In a review of the literature on colorectal cancer screening adherence, Vernon (34) found that two (35, 36) of eight studies reported a positive association between perceived risk and completion of FOBT, whereas six (37, 38, 39, 40, 41, 42) reported no association. In contrast, studies of perceived risk of breast cancer showed a fairly strong and consistent pattern with mammography screening (43). Although only three studies examined the association between perceived risk and sigmoidoscopy completion (36, 44, 45), all found a positive association. In an analysis of our data on the correlates of coverage (i.e., completion of at least one screening examination during the study period) and compliance (i.e.,completion of all recommended screening examinations during the study period), perceived risk was associated with both outcomes in univariate but not in multivariable analysis.5It may be the case that the effects of perceived risk are mediated through another variable, such as intention. Cross-sectional analysis of our baseline data (46) showed that perceived risk was independently associated with intention to have colorectal cancer screening (odds ratio, 2.1; 95% confidence interval, 1.7–2.6). Although the magnitude of the association lessened, baseline perceived risk also was associated with intention measured at year 1 of follow-up(odds ratio, 1.5; 95% confidence interval, 1.1–2.0) but not with intention at year 2 of follow-up (47).
The patterns between perceived risk and examination status between baseline and year 2 were noteworthy. Compared with men who had abnormal findings on examination, perceived risk was low among men, both with and without polyps, who were not examined during the study period and was similar to that for men whose exams were normal. For men whose exams were normal, it is logical that perception of risk could decrease, at least in the short term, because screening with sigmoidoscopy or colonoscopy may reduce the risk of developing cancer. Thus, participation in the screening program may have decreased feelings of susceptibility to developing colorectal cancer or polyps.
The group of men who were not examined during the study period, and who perceived their risk to be low, represent a subgroup that may require more intensive intervention efforts. Nonparticipants in a study of colorectal cancer genetic counseling and testing (48)reported lower perceived ability to cope with mutation-positive test results and were more likely to report the presence of depressive symptoms, especially among women. Psychological distress also has been associated with reduced adherence to preventive behaviors for colorectal cancer (49). A possible explanation is that these persons wished to avoid unfavorable health information (48). These findings support the need to include other psychosocial variables, not measured in our study, such as psychological distress and coping ability, that may affect how information about risk is processed and may result in failure to adopt or sustain health-promoting behaviors.
No association was detected between perceived risk and intervention group status at baseline or year-2 follow-up, although there was a modest effect of the intervention on screening compliance (27). The absence of an association between perceived risk and intervention group status in the year 2 follow-up analysis is of interest because men in the intervention group received educational materials that conveyed personal medical risk information, including their risk factors for colorectal cancer and prior screening history. Although men in the control group were not given personalized risk information, they were told that, as an occupational group, they were at increased risk of developing colorectal cancer. Men in the control group also may have received information about their personal risk from their employer or union, although if provided, such information would not have been delivered in a consistent or standardized manner. The modest impact of the intervention on colorectal cancer screening adherence as reported for the trial as a whole (27), the lack of an association between the intervention and perceived risk at year 2 of follow-up, the perception of low risk in those not screened during the trial, and the decreased proportion of men scoring ≥3 on the perceived risk scale at year 2 of follow-up compared with baseline are consistent with the interpretation that the intervention was not effective at increasing perceived risk. Alternatively, the impact of the intervention on perceived risk may have been diluted because men at all worksites were told that they were at increased risk of colorectal cancer and were offered several opportunities to be screened prior to the intervention. Under these circumstances, simply providing information about risk factors may be insufficient to increase risk perception. An intervention with more emphasis on the psychosocial correlates of perceived risk may have the potential to increase risk perception, but this possibility would need to be tested in another prospective trial.
In our data, specificity was higher than sensitivity in all but one subgroup and was 65% or greater in all subgroups. To our knowledge,other investigators have not reported the sensitivity and specificity of their models, and thus we have no basis for comparison. In the context of conducting an intervention to increase perceived risk,particularly among persons at increased risk, it is preferable to have high specificity in order to have a better chance of correctly identifying those who perceive themselves to be at low risk. It would be informative to validate our models in other populations. The inclusion of other variables, not studied here, might increase sensitivity and specificity.
At present, the body of empiric research on cancer risk communication is relatively small (12, 50). As noted by Harding and Eiser (51), to understand and influence a person’s behavioral decisions on health-related issues, specific behaviors must be studied separately in relation to perceived risk. Moreover, as Slovic (52, 53) and Fischhoff (54) pointed out, perceptions of risk are determined not only by quantitative estimates of risk but also by qualitative characteristics of a particular risk. Just as the importance of perceived risk may vary in relation to different health-related behaviors, various factors may assume differential importance in relation to risk perception,depending on the health behavior being assessed. For instance, risks may be perceived differently by persons with different medical risk, e.g., personal history of polyps or family history of cancer. Perception of risk also may differ for cancers that can be prevented through early detection of premalignant lesions, such as cervical and colorectal cancers, compared with those, such as breast cancer, where early detection confers a survival benefit but does not prevent the disease. We found some support for this view in the follow-up analyses that showed differences in the correlates of perceived risk for men with and without polyps.
There are several limitations that should be considered in interpreting our results. The baseline and subsequent response rates, although similar to response rates for other worksite interventions that used mailed questionnaires (55, 56), were relatively low. Respondents and nonrespondents to the baseline and year 2 surveys differed in terms of some demographic, medical history (including previous colorectal cancer screening behavior), and psychosocial characteristics; however, there was no difference in perceived risk between respondents and nonrespondents to the year 2 survey. In general, the pattern of response was similar to that observed by Lerman and Shemer (57), who reported that respondents to health promotion programs tended to be people already committed to healthy lifestyles. Other considerations in generalizing our findings are that the study population included only white males and that the entire cohort was at increased risk for colorectal cancer mortality. However,the patterns of association for men with a history of polyps may generalize to other groups of men who are at increased risk because of family history of colorectal cancer or polyps or to a personal history of inflammatory bowel disease. Most studies of colorectal cancer screening adherence were conducted prior to the dissemination of colorectal cancer screening guidelines (58, 59, 60), when public awareness about colorectal cancer risk and the benefits of prevention and early detection was low. However, because of the promotion of colorectal cancer screening by the employer and union prior to the trial (24), there may have been increased awareness in our study population compared with the general public. The rates of colorectal cancer screening adherence observed in our trial support this view (27).
In conclusion, although there are limitations, our study is one of the few to compare correlates of perceived risk for colorectal cancer using both baseline and follow-up data and to evaluate whether correlates were similar for men with and without a personal history of colorectal polyps. Our findings, that correlates of perceived risk differed for baseline and follow-up analyses and for men with and without polyps,have implications for intervention development. Except for family history, messages to influence perceived risk would emphasize different factors, depending on whether associations were based on baseline(cross-sectional) or follow-up (prospective) data and depending on whether men reported a personal history of polyps. For example,messages based on cross-sectional data would emphasize the same factors for men, regardless of their history of polyps, whereas messages based on follow-up data would emphasize psychosocial characteristics, such as salience and coherence of screening and intention for men with a history of polyps but not for men without. Our findings support the need to delineate subgroups in the study population to target and tailor health-related messages based on respondent characteristics. Our findings also underscore the need to base health-related messages on prospective data as well as on cross-sectional data to better address health-related beliefs and behaviors.
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.
This work was supported by Grant CA52605 from the National Cancer Institute.
E. A. Fries, K. S. White, D. J. Bowen, S. Talpin, and D. E. Montaño. A prospective study of accuracy of risk perceptions for breast cancer and mammography use. Unpublished data.
The abbreviation used is: FOBT, fecal occult blood test.
R. E. Myers, S. W. Vernon, and B. C. Tilley,unpublished data.
Percentage of men who reported a perceived risk scale score ≥3 at baseline (cross-sectional analyses) and at year 2 of follow-up(prospective analyses) by personal history of colorectal polyps and background characteristics measured at baseline
Independent variables . | Men with a history of polyps (n = 612) . | . | . | Men without a history of polyps (n = 1343) . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | n a . | Baseline . | Year 2 follow-up . | n a . | Baseline . | Year 2 follow-up . | ||||
Intervention group | ||||||||||
Intervention | 249 | 52.2 | 41.4 | 676 | 20.4 | 16.3 | ||||
Control | 363 | 52.1 | 43.5 | 667 | 21.0 | 14.5 | ||||
Age (yr) | ||||||||||
<50 | 58 | 63.8 | 58.6 | 400 | 21.8 | 18.0 | ||||
50–59 | 169 | 55.0 | 45.0 | 383 | 22.2 | 17.0 | ||||
≥60 | 385 | 49.1b | 39.2b | 560 | 18.9 | 12.5b | ||||
Education (yr) | ||||||||||
<12 | 72 | 51.4 | 40.3 | 119 | 16.8 | 17.7 | ||||
12 | 179 | 46.4 | 39.1 | 425 | 20.7 | 13.7 | ||||
>12 | 360 | 55.0b | 45.0 | 787 | 21.2 | 15.9 | ||||
Married | ||||||||||
Yes | 553 | 51.9 | 42.9 | 1212 | 20.8 | 15.8 | ||||
No | 59 | 54.2 | 40.7 | 127 | 19.7 | 11.8 | ||||
Retired | ||||||||||
Yes | 433 | 50.8 | 40.2 | 673 | 20.7 | 14.3 | ||||
No | 179 | 55.3 | 48.6b | 670 | 20.8 | 16.6b | ||||
Family history of CRC or polypsc | ||||||||||
Yes | 203 | 60.6 | 53.7 | 327 | 28.8 | 22.9 | ||||
No | 409 | 47.9b | 37.2b | 1016 | 18.1b | 13.0b | ||||
Screened within 2 years prior to baseline | ||||||||||
Yes | 534 | 53.0 | 44.0 | 1030 | 23.3 | 17.1 | ||||
No | 78 | 46.2b | 33.3b | 313 | 12.1b | 9.9b | ||||
Exam status between baseline and year 2 | ||||||||||
No exams | 81 | 34.6 | 332 | 9.3 | ||||||
Normal exam | 351 | 35.3 | 931 | 15.2 | ||||||
Abnormal exam(s) | 180 | 60.6b | 80 | 43.8b | ||||||
Current smoker | ||||||||||
Yes | 118 | 56.8 | 45.8 | 182 | 27.5 | 20.3 | ||||
No | 490 | 50.8b | 41.8 | 1159 | 19.7b | 14.7b |
Independent variables . | Men with a history of polyps (n = 612) . | . | . | Men without a history of polyps (n = 1343) . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | n a . | Baseline . | Year 2 follow-up . | n a . | Baseline . | Year 2 follow-up . | ||||
Intervention group | ||||||||||
Intervention | 249 | 52.2 | 41.4 | 676 | 20.4 | 16.3 | ||||
Control | 363 | 52.1 | 43.5 | 667 | 21.0 | 14.5 | ||||
Age (yr) | ||||||||||
<50 | 58 | 63.8 | 58.6 | 400 | 21.8 | 18.0 | ||||
50–59 | 169 | 55.0 | 45.0 | 383 | 22.2 | 17.0 | ||||
≥60 | 385 | 49.1b | 39.2b | 560 | 18.9 | 12.5b | ||||
Education (yr) | ||||||||||
<12 | 72 | 51.4 | 40.3 | 119 | 16.8 | 17.7 | ||||
12 | 179 | 46.4 | 39.1 | 425 | 20.7 | 13.7 | ||||
>12 | 360 | 55.0b | 45.0 | 787 | 21.2 | 15.9 | ||||
Married | ||||||||||
Yes | 553 | 51.9 | 42.9 | 1212 | 20.8 | 15.8 | ||||
No | 59 | 54.2 | 40.7 | 127 | 19.7 | 11.8 | ||||
Retired | ||||||||||
Yes | 433 | 50.8 | 40.2 | 673 | 20.7 | 14.3 | ||||
No | 179 | 55.3 | 48.6b | 670 | 20.8 | 16.6b | ||||
Family history of CRC or polypsc | ||||||||||
Yes | 203 | 60.6 | 53.7 | 327 | 28.8 | 22.9 | ||||
No | 409 | 47.9b | 37.2b | 1016 | 18.1b | 13.0b | ||||
Screened within 2 years prior to baseline | ||||||||||
Yes | 534 | 53.0 | 44.0 | 1030 | 23.3 | 17.1 | ||||
No | 78 | 46.2b | 33.3b | 313 | 12.1b | 9.9b | ||||
Exam status between baseline and year 2 | ||||||||||
No exams | 81 | 34.6 | 332 | 9.3 | ||||||
Normal exam | 351 | 35.3 | 931 | 15.2 | ||||||
Abnormal exam(s) | 180 | 60.6b | 80 | 43.8b | ||||||
Current smoker | ||||||||||
Yes | 118 | 56.8 | 45.8 | 182 | 27.5 | 20.3 | ||||
No | 490 | 50.8b | 41.8 | 1159 | 19.7b | 14.7b |
Varies from the subgroup totals because of missing values on some of the independent variables.
P < 0.2, chosen to identify possible variables for inclusion in the multivariable model. Test statistics were computed to compare the distribution of each independent variable for those at high and low perceived risk within the specified subgroup. P were based on χ2tests, adjusted for clustering within worksites using SUDAAN (34).
CRC, colorectal cancer.
Percentage of men who reported a perceived risk scale score ≥3 at baseline (cross-sectional analyses) and at year 2 of follow-up(prospective analyses) by personal history of colorectal polyps and psychosocial factors measured at baseline
Independent variables . | Men with a history of polyps (n = 612) . | . | . | Men without a history of polyps (n = 1343) . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | n a . | Baseline . | Year 2 follow-up . | n a . | Baseline . | Year 2 follow-up . | ||||
Baseline perceived risk | ||||||||||
≥3 | 293 | 61.1 | 1065 | 42.1 | ||||||
<3 | 319 | 22.5b | 278 | 8.5b | ||||||
Intention to do CRCSc | ||||||||||
≥3 | 535 | 53.8 | 45.2 | 1045 | 23.5 | 17.0 | ||||
<3 | 67 | 38.8b | 20.9b | 265 | 9.1b | 8.7b | ||||
Belief in salience and coherence of CRCS | ||||||||||
≥3 | 584 | 53.4 | 44.2 | 1173 | 22.9 | 16.5 | ||||
<3 | 28 | 25.0b | 10.7b | 169 | 5.3b | 7.7b | ||||
Belief in efficacy of CRCS | ||||||||||
≥3 | 599 | 52.9 | 42.9 | 1293 | 21.1 | 15.9 | ||||
<3 | 13 | 15.4b | 30.8 | 44 | 4.6b | 2.3b | ||||
Belief polyp removal prevents CRCS | ||||||||||
≥3 | 548 | 53.3 | 43.3 | 1169 | 21.6 | 15.7 | ||||
<3 | 62 | 40.3b | 37.1 | 162 | 15.4b | 13.6 | ||||
Belief CRC can be cured | ||||||||||
≥3 | 592 | 52.7 | 43.1 | 1280 | 21.1 | 15.6 | ||||
<3 | 16 | 25.0b | 37.5 | 60 | 11.7b | 11.7 | ||||
Perceived self-efficacy related to CRCS | ||||||||||
≥3 | 454 | 53.7 | 43.8 | 861 | 23.5 | 17.8 | ||||
<3 | 158 | 47.5b | 39.2 | 479 | 15.7b | 11.3b | ||||
Worry or fear of being diagnosed with CRC | ||||||||||
≥3 | 132 | 65.9 | 54.6 | 185 | 38.4 | 24.9 | ||||
<3 | 464 | 48.5b | 39.4b | 1124 | 17.4b | 13.7b | ||||
Concern about CRCS discomfort | ||||||||||
≥3 | 273 | 51.3 | 42.1 | 664 | 21.5 | 14.3 | ||||
<3 | 335 | 52.5 | 43.0 | 669 | 19.9 | 16.6 | ||||
Social support and influence variables | ||||||||||
Receptivity to family member support | ||||||||||
≥3 | 491 | 54.0 | 43.0 | 956 | 22.0 | 15.8 | ||||
<3 | 105 | 44.8b | 41.0 | 333 | 18.0b | 14.7 | ||||
Support for screening from family members | ||||||||||
≥3 | 537 | 54.9 | 43.8 | 1063 | 23.6 | 16.8 | ||||
<3 | 50 | 26.0b | 32.0b | 199 | 8.5b | 8.5b |
Independent variables . | Men with a history of polyps (n = 612) . | . | . | Men without a history of polyps (n = 1343) . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | n a . | Baseline . | Year 2 follow-up . | n a . | Baseline . | Year 2 follow-up . | ||||
Baseline perceived risk | ||||||||||
≥3 | 293 | 61.1 | 1065 | 42.1 | ||||||
<3 | 319 | 22.5b | 278 | 8.5b | ||||||
Intention to do CRCSc | ||||||||||
≥3 | 535 | 53.8 | 45.2 | 1045 | 23.5 | 17.0 | ||||
<3 | 67 | 38.8b | 20.9b | 265 | 9.1b | 8.7b | ||||
Belief in salience and coherence of CRCS | ||||||||||
≥3 | 584 | 53.4 | 44.2 | 1173 | 22.9 | 16.5 | ||||
<3 | 28 | 25.0b | 10.7b | 169 | 5.3b | 7.7b | ||||
Belief in efficacy of CRCS | ||||||||||
≥3 | 599 | 52.9 | 42.9 | 1293 | 21.1 | 15.9 | ||||
<3 | 13 | 15.4b | 30.8 | 44 | 4.6b | 2.3b | ||||
Belief polyp removal prevents CRCS | ||||||||||
≥3 | 548 | 53.3 | 43.3 | 1169 | 21.6 | 15.7 | ||||
<3 | 62 | 40.3b | 37.1 | 162 | 15.4b | 13.6 | ||||
Belief CRC can be cured | ||||||||||
≥3 | 592 | 52.7 | 43.1 | 1280 | 21.1 | 15.6 | ||||
<3 | 16 | 25.0b | 37.5 | 60 | 11.7b | 11.7 | ||||
Perceived self-efficacy related to CRCS | ||||||||||
≥3 | 454 | 53.7 | 43.8 | 861 | 23.5 | 17.8 | ||||
<3 | 158 | 47.5b | 39.2 | 479 | 15.7b | 11.3b | ||||
Worry or fear of being diagnosed with CRC | ||||||||||
≥3 | 132 | 65.9 | 54.6 | 185 | 38.4 | 24.9 | ||||
<3 | 464 | 48.5b | 39.4b | 1124 | 17.4b | 13.7b | ||||
Concern about CRCS discomfort | ||||||||||
≥3 | 273 | 51.3 | 42.1 | 664 | 21.5 | 14.3 | ||||
<3 | 335 | 52.5 | 43.0 | 669 | 19.9 | 16.6 | ||||
Social support and influence variables | ||||||||||
Receptivity to family member support | ||||||||||
≥3 | 491 | 54.0 | 43.0 | 956 | 22.0 | 15.8 | ||||
<3 | 105 | 44.8b | 41.0 | 333 | 18.0b | 14.7 | ||||
Support for screening from family members | ||||||||||
≥3 | 537 | 54.9 | 43.8 | 1063 | 23.6 | 16.8 | ||||
<3 | 50 | 26.0b | 32.0b | 199 | 8.5b | 8.5b |
Varies from the subgroup totals because of missing values on some of the independent variables.
P < 0.2, chosen to identify possible variables for inclusion in the multivariable model. Test statistics were computed to compare the distribution of each independent variable for those at high and low perceived risk within the specified subgroup. P were based on χ2tests, adjusted for clustering within worksites using SUDAAN (34).
CRC, colorectal cancer; CRCS,colorectal cancer screening.
Percentage of men who reported a perceived risk scale score ≥3 at baseline (cross-sectional analyses) and at year 2 of follow-up(prospective analyses) by personal history of colorectal polyps and plant characteristics measured at baseline
Independent variables . | Men with a history of polyps (n = 612) . | . | . | Men without a history of polyps (n = 1343) . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | n a . | Baseline . | Year 2 follow-up . | n a . | Baseline . | Year 2 follow-up . | ||||
Plant characteristics | ||||||||||
CRCb offered all year | ||||||||||
Yes | 281 | 51.6 | 42.4 | 463 | 19.9 | 14.9 | ||||
No | 331 | 52.6 | 42.9 | 880 | 21.1 | 15.7 | ||||
CRC offered at worksite | ||||||||||
Yes | 403 | 50.1 | 40.2 | 710 | 19.9 | 14.7 | ||||
No | 209 | 56.0c | 47.4c | 633 | 21.6 | 16.3 | ||||
FOBTs distributed by mail | ||||||||||
Yes | 444 | 52.3 | 41.9 | 840 | 20.4 | 15.2 | ||||
No | 168 | 51.8 | 44.6 | 503 | 21.3 | 15.7 | ||||
Plant notified employees that FOBT was due | ||||||||||
Yes | 437 | 54.0 | 44.2 | 884 | 19.6 | 14.8 | ||||
No | 175 | 47.4c | 38.9 | 459 | 22.9 | 16.6 | ||||
Plant notified employee that DRE and FS were due | ||||||||||
Yes | 374 | 53.7 | 44.7 | 698 | 20.6 | 15.3 | ||||
No | 238 | 49.6 | 39.5 | 645 | 20.8 | 15.5 |
Independent variables . | Men with a history of polyps (n = 612) . | . | . | Men without a history of polyps (n = 1343) . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | n a . | Baseline . | Year 2 follow-up . | n a . | Baseline . | Year 2 follow-up . | ||||
Plant characteristics | ||||||||||
CRCb offered all year | ||||||||||
Yes | 281 | 51.6 | 42.4 | 463 | 19.9 | 14.9 | ||||
No | 331 | 52.6 | 42.9 | 880 | 21.1 | 15.7 | ||||
CRC offered at worksite | ||||||||||
Yes | 403 | 50.1 | 40.2 | 710 | 19.9 | 14.7 | ||||
No | 209 | 56.0c | 47.4c | 633 | 21.6 | 16.3 | ||||
FOBTs distributed by mail | ||||||||||
Yes | 444 | 52.3 | 41.9 | 840 | 20.4 | 15.2 | ||||
No | 168 | 51.8 | 44.6 | 503 | 21.3 | 15.7 | ||||
Plant notified employees that FOBT was due | ||||||||||
Yes | 437 | 54.0 | 44.2 | 884 | 19.6 | 14.8 | ||||
No | 175 | 47.4c | 38.9 | 459 | 22.9 | 16.6 | ||||
Plant notified employee that DRE and FS were due | ||||||||||
Yes | 374 | 53.7 | 44.7 | 698 | 20.6 | 15.3 | ||||
No | 238 | 49.6 | 39.5 | 645 | 20.8 | 15.5 |
Varies from the subgroup totals because of missing values on some of the independent variables.
CRC, colorectal cancer; DRE,digital rectal examination; FS, flexible sigmoidoscopy examination.
P < 0.2, chosen to identify possible variables for inclusion in the multivariable model. Test statistics were computed to compare the distribution of each independent variable for those at high and low perceived risk within the specified subgroup. P were based on χ2tests, adjusted for clustering within worksites using SUDAAN (34).
Odds ratios and 95% confidence intervalsa for a perceived risk scale score ≥3 at baseline (cross-sectional analyses) and at year 2 of follow-up (prospective analyses) by personal history of colorectal polyps and independent variables measured at baseline
Independent variables . | Men with a history of polyps . | . | Men without a history of polyps . | . | ||
---|---|---|---|---|---|---|
. | Baseline (n = 573)b . | Year 2 follow-up (n = 602)b . | Baseline (n = 1235)b . | Year 2 follow-up (n = 1343)b . | ||
Intervention group (Yes/No) | 1.0 (0.7–1.4) | 0.8 (0.5–1.1) | 0.9 (0.6–1.4) | 1.2 (0.8–1.8) | ||
Family history of CRCc or polyps (Yes/No) | 1.7 (1.3–2.3) | 1.7 (1.1–2.6) | 1.8 (1.4–2.4) | 1.6 (1.1–2.4) | ||
Exam status between baseline and year 2 | ||||||
No exams | 0.4 (0.2–0.7) | 0.1 (0.07–0.2) | ||||
Normal exams | 0.3 (0.2–0.5) | 0.2 (0.1–0.3) | ||||
Abnormal exams | 1.0 | 1.0 | ||||
Current smoker (Yes/No) | 1.6 (1.1–2.2) | |||||
Baseline perceived risk (≥3, <3) | NA | 5.4 (3.7–7.6) | NA | 8.3 (5.7–12.1) | ||
Support for screening from family members (≥3, <3) | 3.7 (2.2–6.3) | 3.2 (2.0–5.1) | ||||
Fear or worry about being diagnosed with CRC (≥3, <3) | 2.3 (1.5–3.7) | 3.0 (1.9–4.6) | ||||
Belief in salience and coherence of screening (≥3, <3) | 3.9 (1.4–11.1) | |||||
Intention to do CRCS (≥3, <3) | 2.0 (1.0–4.1) | |||||
CRC screening offered at worksite (Yes/No) | 0.6 (0.4–0.9) |
Independent variables . | Men with a history of polyps . | . | Men without a history of polyps . | . | ||
---|---|---|---|---|---|---|
. | Baseline (n = 573)b . | Year 2 follow-up (n = 602)b . | Baseline (n = 1235)b . | Year 2 follow-up (n = 1343)b . | ||
Intervention group (Yes/No) | 1.0 (0.7–1.4) | 0.8 (0.5–1.1) | 0.9 (0.6–1.4) | 1.2 (0.8–1.8) | ||
Family history of CRCc or polyps (Yes/No) | 1.7 (1.3–2.3) | 1.7 (1.1–2.6) | 1.8 (1.4–2.4) | 1.6 (1.1–2.4) | ||
Exam status between baseline and year 2 | ||||||
No exams | 0.4 (0.2–0.7) | 0.1 (0.07–0.2) | ||||
Normal exams | 0.3 (0.2–0.5) | 0.2 (0.1–0.3) | ||||
Abnormal exams | 1.0 | 1.0 | ||||
Current smoker (Yes/No) | 1.6 (1.1–2.2) | |||||
Baseline perceived risk (≥3, <3) | NA | 5.4 (3.7–7.6) | NA | 8.3 (5.7–12.1) | ||
Support for screening from family members (≥3, <3) | 3.7 (2.2–6.3) | 3.2 (2.0–5.1) | ||||
Fear or worry about being diagnosed with CRC (≥3, <3) | 2.3 (1.5–3.7) | 3.0 (1.9–4.6) | ||||
Belief in salience and coherence of screening (≥3, <3) | 3.9 (1.4–11.1) | |||||
Intention to do CRCS (≥3, <3) | 2.0 (1.0–4.1) | |||||
CRC screening offered at worksite (Yes/No) | 0.6 (0.4–0.9) |
Computed taking clustering into account using SUDAAN (34).
Varies from subgroup totals because of missing values on some independent variables. Numbers in parentheses, 95% confidence interval.
CRC, colorectal cancer; CRCS,colorectal cancer screening; NA, not applicable.
Prevalence, sensitivity, and specificity, including 95% confidence intervals (95% CI) of the multivariable models for a perceived risk scale score ≥3 at baseline (cross-sectional analyses) and at year 2 of follow-up (prospective analyses) by personal history of colorectal polyps
. | Men with a history of polyps . | . | . | . | . | . | Men without a history of polyps . | . | . | . | . | . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Baseline . | . | . | Year 2 follow-up . | . | . | Baseline . | . | . | Year 2 follow-up . | . | . | ||||||||||
. | n a . | % . | (95% CI) . | n . | % . | (95% CI) . | n . | % . | (95% CI) . | n . | % . | (95% CI) . | ||||||||||
Prevalenceb | 573 | 52.9 | (48.8–57.0) | 602 | 42.5 | (38.6–46.4) | 1235 | 21.1 | (18.8–23.4) | 1343 | 15.4 | (13.5–17.3) | ||||||||||
Sensitivityc | 303 | 54.1 | (48.5–59.7) | 256 | 73.8 | (68.4–79.2) | 261 | 56.5 | (50.5–62.5) | 207 | 68.6 | (62.3–74.9) | ||||||||||
Specificityd | 270 | 65.2 | (59.5–70.9) | 346 | 69.1 | (64.2–74.0) | 974 | 65.9 | (62.9–68.9) | 1136 | 82.5 | (80.3–84.7) |
. | Men with a history of polyps . | . | . | . | . | . | Men without a history of polyps . | . | . | . | . | . | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Baseline . | . | . | Year 2 follow-up . | . | . | Baseline . | . | . | Year 2 follow-up . | . | . | ||||||||||
. | n a . | % . | (95% CI) . | n . | % . | (95% CI) . | n . | % . | (95% CI) . | n . | % . | (95% CI) . | ||||||||||
Prevalenceb | 573 | 52.9 | (48.8–57.0) | 602 | 42.5 | (38.6–46.4) | 1235 | 21.1 | (18.8–23.4) | 1343 | 15.4 | (13.5–17.3) | ||||||||||
Sensitivityc | 303 | 54.1 | (48.5–59.7) | 256 | 73.8 | (68.4–79.2) | 261 | 56.5 | (50.5–62.5) | 207 | 68.6 | (62.3–74.9) | ||||||||||
Specificityd | 270 | 65.2 | (59.5–70.9) | 346 | 69.1 | (64.2–74.0) | 974 | 65.9 | (62.9–68.9) | 1136 | 82.5 | (80.3–84.7) |
Based on the number of respondents included in the multivariable analysis.
Percentage with increased perceived risk.
Number of men correctly predicted with increased perceived risk divided by the number of men observed with increased perceived risk.
Number of men correctly predicted with low perceived risk divided by the number of men observed with low perceived risk.
Acknowledgments
We acknowledge the support of the following persons: Colette Miesse for assistance with the literature review and bibliography;Donna Mott and Toni Chociemski for assistance with computer programming; Mary Ann Kozlowski for data management support; and the interviewing and abstracting team, headed by Lois Lamerato, for support throughout the trial. We thank Karen Glanz for helpful comments on an earlier version of the manuscript.