Physician Intention to Recommend Complete Diagnostic Evaluation in Colorectal Cancer Screening¹

In 1999, there will be ≈129,400 new cases of colon and rectum (colorectal) cancer diagnosed and ≈6,600 related deaths in the United States. To reduce colorectal cancer morbidity and mortality in older adults (i.e., men and women 50 or more years of age), the National Cancer Institute and the American Cancer Society support colorectal cancer screening (1, 2). Colorectal cancer screening often involves the use of annual FOBT³ and periodic screening sigmoidoscopy. The American Cancer Society and the United States Preventive Services Taskforce have recommended the use of annual FOBT screening and/or periodic flexible sigmoidoscopy for asymptomatic older adults (2, 3).

It has been estimated that FOBT screening can reduce colorectal cancer mortality by 10–30% (4). This projection is supported by preliminary results published for one randomized trial in the United States that suggest that mortality from colorectal cancer can be reduced through FOBT screening among older adults. Specifically, Mandel et al. (5) have reported that annual FOBT screening along with CDE (either colonoscopy or the combination of double air contrast barium enema X-ray and flexible sigmoidoscopy) for FOBT+ cases was associated with a 33% reduction in a 13-year cumulative mortality rate from colorectal cancer. In this and other major controlled trials (5, 6, 7, 8, 9), screening FOBT+ results have been routinely followed-up with CDE. CDE for FOBT+ cases is encouraged by the American Cancer Society and numerous professional organizations, including the American College of Gastroenterology, American Gastroenterological Association, and the American Society of Colon and Rectal Surgeons (10).

For FOBT screening to be effective in reducing colorectal cancer mortality outside of randomized studies, CDE should be performed routinely for FOBT+ cases. However, several population-based FOBT screening programs have shown that CDE is often not performed for persons who have FOBT+ results (11, 12).⁴ Bralow (13) has reported that only 37–63% of patients who present an abnormal FOBT result at a PCP office will undergo CDE. A report from a community-based study of FOBT screening in Texas (14) also indicates that the majority of individuals who have an abnormal screening FOBT result do not undergo CDE. A study conducted in a large-scale FOBT screening program aimed at older adult members of an independent practice association-model MCO indicated that many FOBT+ patients were not advised to have CDE by a primary care physician. It is important to note that among patients who were recommended for CDE, most were adherent (15).

To identify factors that may affect whether PCPs recommend CDE, we conducted a telephone survey of physicians who have patients that are targeted for an MCO-based FOBT screening program. On the survey, we described different case scenarios in which an FOBT+ patient would contact the PCP and ask what should be done to follow-up the screening test result. The physician was then asked whether s/he would recommend CDE at the time of this initial contact. We also obtained survey data on the physician’s practice, individual background characteristics, and psychosocial variables. Using this information, we conducted univariate and multivariate analyses of CDE intention.

In the mid-1980s, U. S. Healthcare, Inc. (now Aetna US Healthcare, Inc.), one of the country’s largest MCOs, initiated the U.S. Healthcare Check Colorectal Cancer Screening Program. Initially, the screening program was provided to members who were 50 or more years of age and were covered by MCOs in Pennsylvania and Southern New Jersey. Today, the program is national in scope.

In the screening program, a central screening office mails an FOBT kit to MCO members 50 or more years of age on an annual basis. FOBT mailings to the member population occur on a rolling weekly basis throughout the calendar year. Individuals who have not returned FOBTs within 60 days receive a reminder letter that encourages adherence. The kit includes an introductory letter, information about colorectal cancer, three Hemoccult II FOBT slides with instructions, and a postage-paid return envelope. Printed instructions approved by the U. S. Food and Drug Administration regarding diet and medication restrictions during the test period are included in the FOBT kit.

An independent laboratory analyzes the returned FOBT cards. The laboratory then notifies the members’ PCP offices of test results and instructs members to contact their PCPs to obtain the results. The MCO has the capacity to access screening program administrative data (e.g., member age, FOBT return status, test results, and diagnostic medical procedures) for use in evaluating the quality of care that is provided to its members.

In 1994, a random sample of 518 (28%) PCPs was selected from an administrative database that listed 1845 independent primary care practitioners who were affiliated with Aetna U. S. Healthcare’s health care plans in Southeastern Pennsylvania and Southern New Jersey and had MCO-member patients who were eligible for the screening program. Information that was available in the database included physician name, address, telephone number, and medical specialty.

An advance letter was mailed to all PCPs announcing that they would be contacted and asked to participate in a telephone survey. The initial letter was sent under the signature of a senior medical director of the MCO. Physicians for whom surveys could not initially be completed were subsequently mailed a reminder letter to encourage survey participation. Complete survey data were obtained from 363 physicians, reflecting a 70% response rate. Reasons for not completing the survey included the following: subject was unavailable during the survey period (n = 65); subject refused to complete the survey (n = 62); and other (n = 18). Twelve additional physicians did not complete all survey items that were used in data analysis. Using information from the MCO administrative database, we compared survey responders and nonresponders in terms of medical specialty (i.e., Family Medicine and Internal Medicine), number of physicians in the practice, and geographic area in which the practice was located. No statistically significant differences (P < 0.05) were found.

Mathematica Policy Research, Inc. of Princeton, NJ administered a telephone survey that was designed to obtain data on physician intention to recommend CDE for FOBT+ cases detected in the MCOs colorectal cancer screening program. A theory-based explanatory framework, which is referenced in this article as the DEM in cancer screening, was used to structure the collection of survey measures. The DEM, which is shown in Fig. 1, is based on the Health Belief Model (16), Theory of Reasoned Action (17), and Social Cognitive Theory (18). This model includes measures of physician practice characteristics; personal background; beliefs about screening, CDE, and patient behavior; social support and influence related to CDE; intention to do CDE; and CDE performance.

In terms of physician practice characteristics, patient population size was measured as a continuous variable using each respondent’s estimate of the number of adult patients 50 or more years of age in the practice. The physician estimate regarding the number of colorectal cancer patients seen in the past 12 months also was treated as a continuous variable. The percentage of the older adult patients in the practice, as reported by respondents, was dichotomized as <20% versus ≥20%. The percentage of MCO-insured older adult members seen in the past 12 months for follow-up of a screening FOBT+ result was dichotomized as <5% versus >5%. These cut-points were based on anecdotal reports from PCPs about what they would consider to be substantial percentages of patients who, respectively, were MCO members and had a FOBT+ result. Physician practice size was dichotomized in terms of one (solo) versus more than one physician in the practice. This dichotomy was chosen because approximately one-half of the PCPs affiliated with the MCO at the time of the survey were solo practitioners.

Physician gender, race/ethnicity, and board certification status were assessed in standard terms. Physician age was considered as a continuous variable. Number of years in practice was dichotomized as <10 years versus ≥10 years. This cut-point represents the median number of years that respondents were in practice. Practice specialty was categorized as either internal medicine or family practice.

Physician belief in the efficacy of annual FOBT screening for the detection of colorectal cancer was assessed using a single item with a six-level Likert-type response pattern (i.e., strongly disagree = 1, moderately disagree = 2, slightly disagree = 3, slightly agree = 4, moderately agree = 5, and strongly agree = 6). Three single items were used to assess physician belief in the effectiveness of population-based FOBT screening as a method to identify persons with adenomatous polyps, early colorectal cancer, and late colorectal cancer. Answers to these items were recorded using a four-level Likert-type response pattern (i.e., not at all likely = 1, a little likely = 2, somewhat likely = 3, and very likely = 4). Physician belief in the benefit of the MCO’s colorectal cancer screening program to his/her practice was assessed in terms of a “Yes” or “No” response. Physician perceptions about CDE also were measured using survey items with the same response pattern. Responses to other items measuring perceptions about CDE (i.e., patient time involved in having CDE, patient discomfort from CDE, the efficacy of CDE, the likelihood that the benefits of CDE for the patient outweigh the risks associated with follow-up, patient receptivity to a CDE recommendation, the impact of normal CDE findings on the physician-patient relationship, the cost of CDE to the practice, the likelihood that CDE-related costs would be considered in decision-making about diagnostic follow-up, and physician time involved in CDE) were assessed using the six-level Likert-type response set. Items that measured physician stress from uncertainty about recommending CDE were assessed in the same fashion (19).

In relation to social support and influence, we assessed the extent to which respondents thought that recommending CDE was standard practice in the community and believed that they adhered to community practice standards related to CDE. The same six-level response set was used.

We obtained a CDE intention score by asking respondents whether they would recommend CDE in different patient-related situations. Hypothetical cases were presented, describing FOBT+ patients as being older, having a family history of colorectal cancer, having anemia, having a history of hemorrhoids, having only one of three FOBT+ test results reported, or being adherent to dietary restrictions as recommended prior to FOBT screening. Anecdotal reports from previous research conducted with PCPs who were familiar with the MCO’s FOBT screening program, which was described earlier (15), suggested that these situations may influence whether CDE is recommended for a FOBT+ patient. A six-level Likert-type response set was used for each of the items. A priori, we planned to consider these items together as a CDE intention scale. This decision was supported by factor analysis results.

Exploratory factor analysis was done using responses to survey items that assessed physician perceptions related to FOBT screening, CDE, and social influence. This analysis led to the formation of the following three physician perception scales: perceived FOBT effectiveness (three items, α = 0.69); physician and patient time required for CDE (two items, α = 0.67); physician stress from uncertainty about recommending CDE (10 items, α = 0.89). The Cronbach’s α correlation coefficient for each scale indicates an acceptable level of reliability for each scale. These scales were standardized for use in subsequent analyses related to the outcome variable, physician intention to recommend CDE. Items that did not form reliable scales were treated as single-item, independent variables. Confirmatory factor analysis was conducted with six items that were posited as measuring physician CDE intention. These items formed a reliable scale (α = 0.73), which was standardized. Responses to single items were coded as low or high (≤3 versus >3, respectively). Scale scores for perceived FOBT effectiveness, physician and patient time required for CDE, and stress from uncertainty about recommending CDE were coded as low or high (≤3 versus >3, respectively). CDE intention was standardized and categorized as being low, moderate, or high (≤3; >3 and ≤ 5; and > 5, respectively).

Frequency distributions were generated for DEM variables. Univariate analyses also were carried out to assess whether associations exist between each independent variable and outcome. Variables that were significant at P < 0.20 were considered in multivariate analyses. Ordinal regression analysis, assuming the cumulative odds model, was used to identify the covariates that were significantly associated with CDE intention (20). The proportional odds assumption was tested throughout the modeling process and was satisfied. We determined that 30% of study participants were in practices where more than one physician completed the survey. Therefore, effect estimates, standard errors, and Ps were computed using the generalized estimating equations for both univariate and multivariate analyses (21). Main effects and interaction terms involving these main effects that were found to be statistically significant (P < 0.05) were retained in the final model. SAS 6.12 was used to conduct the analyses.

Forty-nine % of survey respondents estimated that there were more than 1000 patients 50 or more years of age in their practices. Most (74%) of the physicians reported that they had seen fewer than 10 colorectal cancer patients during the previous year. Fifty-one % of the physicians indicated that at least 20% of their older adult patients were MCO insured, and 23% reported that at least 5% of their older adult, MCO-insured patients had had a recent screening FOBT+ result. Thirty-seven % were in solo practices.

Inspection of physician background characteristics indicates that 65% of survey respondents were 50 or more years of age, 83% were male, and 84% were white. Seventy-two % of the physicians were trained in family medicine, and 28% were trained in internal medicine. Seventy-three % of respondents were board certified. Seventy-five % of respondents had been in practice for >10 years.

Physician cognitive and psychological representations of FOBT screening generally tended to be favorable. That is, 88% reported that FOBT screening can identify individual patients with colorectal cancer, and >96% of respondents reported that at the population level, FOBT screening is an effective approach to finding persons with early colorectal cancer and large colorectal adenomas. Ninety-four % believed that the screening program is beneficial to their practice. Almost all believed that CDE is efficacious (94%), and that the benefits of CDE for the patient outweigh procedure-related risks (96%). Most (72%) physicians believed that FOBT+ patients would adhere to a referral for CDE. Only 3% believed that negative CDE findings might harm the physician-patient relationship. Respondents expressed concern about patient discomfort (31%), physician and patient time required for CDE (56%), and physician CDE-related costs (58%). More than one-half (55%) of the physicians reported that they felt stress from uncertainty about CDE. In terms of social support and influence, 88% of the respondents indicated that CDE was standard practice. Practically all of the respondents (98%) indicated that they adhered to standard practice related to CDE.

Analyses of individual CDE intention scale items suggest that physicians were likely to say they would recommend CDE (i.e., have an item score of >3) for FOBT+ patients who: are younger (54%), have more than one of three FOBT slides that are positive (66%), do not have a history of hemorrhoids (79%), do not have a normal hemoglobin level (80%), do have a family history of colorectal cancer (79%), and follow dietary restrictions that that are part of the screening process (75%). Overall, physician CDE intention scale scores were distributed as follows: low (22%), moderate (51%), and high (27%).

Univariate analyses, shown in Table 1, indicate that a number of DEM variables were associated (P < 0.20) with CDE intention. In terms of practice characteristics, physicians who saw more MCO patients with an FOBT+ result, had more MCO older adult patients, and were in a multiphysician practice had relatively high CDE intention scores. In relation to personal background variables, physicians who were 50 or more years of age, white, board certified, and had been in practice for >10 years had higher levels of CDE intention. Among the cognitive and psychological representation variables, we found that physician beliefs that CDE is efficacious and the benefits of CDE outweigh related risks for patients were positively associated with CDE intention. Perceived efficacy of FOBT screening, perceived FOBT effectiveness, the belief that CDE causes patient discomfort, the belief that FOBT+ patients tend to be nonadherent, the belief that CDE takes a lot of physician and patient time, and concerns related to CDE costs were negatively associated with intention to recommend CDE. In terms of social support and influence, the belief that CDE is standard practice in the community was positively associated with intention to recommend CDE.

Results of multivariate analyses are displayed in Table 2. Variables that were positively and significantly associated with CDE intention included the following: being board certified (OR, 1.9; CI, 1.1–3.1), being in practice for >10 years (OR, 1.8; CI, 1.1–2.9), believing that CDE is efficacious (OR, 3.5; CI, 1.5–8.0), and believing that CDE is standard practice (OR, 3.6; CI, 1.8–7.1). Variables that were negatively associated with outcome included: physician belief that CDE is expensive (OR, 0.6; CI, 0.4–0.9), and the belief that CDE-related costs may influence decision-making (OR, 0.5; CI, 0.3–0.8). Two interaction terms, both of which involved physician practice size, remained in the final model: practice size with FOBT screening efficacy (P = 0.009), and practice size with the belief that CDE benefits outweigh risks (P = 0.007). To understand the interactions, we combined the interaction terms with the appropriate main effects from the model. We found that the association between CDE intention and perceived FOBT screening efficacy was negative in multiphysician practices (OR, 0.10; CI, 0.01–1.05) and was marginally significant. However, CDE intention and FOBT screening efficacy were not significantly associated among primary care physicians in solo practice (OR, 1.05; CI, 0.41–2.69). The association between CDE intention and the belief that CDE benefits outweigh risks was positive (OR, 10.8; CI, 1.3–91.9). No statistically significant association between the variables was found among solo practitioners (OR, 1.03; CI, 0.18–5.78).

We found that more than three-fourths of surveyed PCPs had a moderate to high level of intention to recommend CDE referral for a screening FOBT+ patient at the first follow-up encounter. This finding may appear to be somewhat surprising in light of reports in the literature, including our own research (11, 12, 13, 14, 15),⁴ which indicate that CDE is frequently not performed for FOBT+ patients. However, it is understandable that the magnitude of physician self-reported intention to make a CDE referral is likely to be somewhat greater than actual CDE performance. To some extent, the difference may be accounted for by the fact that patients can refuse to undergo CDE when it is recommended. In addition, the actions of other practitioners may impact patient behavior. That is, the primary care physician may actually advise a FOBT+ patient to have CDE. However, the specialist to whom the patient has been referred may perform procedures other than CDE. Furthermore, it has been documented that primary care physicians commonly overestimate the extent to which they make cancer screening tests available to their patients (22). These circumstances notwithstanding, the identification of factors that are associated with physician orientation to making a follow-up recommendation can provide useful insights into forces that may limit CDE performance.

Ludke (23), Shortell and Vahovich (24), Deber (25), Jaen et al. (26), and Godin et al. (27) have suggested that practice characteristics, physician background, and patient attributes are likely to influence physician use of preventive health care modalities. Our findings indicate that such factors may be important in relation to CDE in colorectal cancer screening. In terms of physician practice characteristics, the data indicate that physicians who had been in practice for >10 years had a higher level of CDE intention than those in practice for a shorter period of time. This finding could reflect awareness and acceptance of the need for more thorough follow-up of an FOBT+ result, a perceptual frame that may have been established over time on the basis of experience with patients who have FOBT+ results.

Physicians who were board certified were more likely to favor CDE than less-experienced practitioners. Similar findings have been reported in research on physician provision of a recommendation for patients to undergo screening flexible sigmoidoscopy (28). It may be that increased training and experience in clinical practice confers a belief that CDE for FOBT+ patients should be considered a routine part of colorectal cancer screening process. In this regard, we observed that physicians who thought that CDE is efficacious had higher CDE intention scores. It is reasonable to assume that those PCPs who recommend CDE as follow-up believe it has substantial benefits for their patients.

The strength of the positive association between perceived screening efficacy and CDE intention was weaker among physicians in larger practices as compared with solo practices. This finding was unexpected. One possible explanation of this finding may be that physicians in larger practices, as compared with those in smaller practices, tended to see more FOBT+ cases that are not ultimately diagnosed with colorectal cancer. In other words, they may have routinely seen more “false positive” cases. Physicians in larger practices may have had less confidence in the efficacy of screening than physicians in smaller practices and, as a result, may have exhibited a lower level of CDE intention. This explanation must be considered to be tentative, because further research is needed to determine what may account for this finding.

Our finding related to the interaction between practice size and perceived CDE benefit is also intriguing. Specifically, we found a positive association between perceived CDE benefit and intention to recommend CDE to be stronger among physicians in larger practices than those in smaller practices. An explanation for this finding also may relate to the number of FOBT+ patients seen in larger versus smaller practices. It may be that practitioners who are faced with a substantial volume of FOBT+ patients tend to use CDE as a routine approach to follow-up. Reported CDE intention, then, would reflect an established pattern of care that is more characteristic of larger than smaller practices. When considered together, interpretations of the interaction effects may appear to be contradictory. However, it would not be surprising to find that physicians who see relatively large numbers of FOBT+ patients are both supportive of CDE and cognizant of the fact that the FOBT+ result may not lead to a diagnosis of colorectal neoplasia.

Survey results also suggest that practitioner support for recommending CDE may be tempered by concerns related to the cost of diagnostic follow-up. Specifically, CDE intention scores were lower among physicians who viewed CDE as costly and those who thought their decision concerning whether to recommend CDE would be influenced by procedure costs. It has been suggested that utilization of specialty services by primary care physicians who are affiliated with MCOs may vary, depending on how much they are at risk for associated costs (29). Recently, this argument has been called into question (30). It is not known whether PCP financial concerns affect physician CDE recommendation rates. Future research should address this issue.

The matter of social influence outside the practice setting is also intriguing. Although the literature in this area is limited, it has been observed that professional peer support has a strong impact on PCP behavior (31, 32, 33). Presently, there is substantial scientific and professional support for FOBT screening for colorectal neoplasia. This support is based on findings from case control and randomized studies reported in the literature; guidelines promoted by the United State Preventive Services Taskforce, the American Cancer Society, and the National Cancer Institute; and various public health organizations and clinical societies. CDE (usually colonoscopy) is reported as the most commonly used approach for the diagnostic evaluation of patients who have had FOBT+ results in international randomized colorectal cancer screening trials. To date, little attention has been paid to defining CDE as part of routine screening outside the context of these trials. The explicit inclusion of CDE in existing colorectal cancer screening guidelines would help to highlight the need to consider CDE as part of routine care.

Physicians routinely use information about patient demographic background, personal and family history, and medical test results to guide decision making about diagnostic evaluation. In the FOBT screening situation, we observed that when presented with specific scenarios, CDE intention varies. For example, some physicians are less likely to recommend CDE for older FOBT+ patients. By restricting CDE to younger patients, physicians may deny an opportunity for early cancer detection to a segment of the population at high risk. On the other hand, it is important to acknowledge that older patients may be less physically able to tolerate invasive diagnostic procedures such as colonoscopy. Given that we did not specify age categories in the scenarios presented to respondents, there is need for further clarification of CDE intention and patient age. We also found that some practitioners appear to assume a positive correlation between the number of abnormal test results and the presence of neoplastic disease. To our knowledge, this assumption is not supported in the literature.

Findings from this study suggest that there are specific cognitive and psychological correlates of physician CDE intention that are potentially amenable to educational intervention (i.e., perceptions about FOBT screening, CDE efficacy, CDE as standard practice, and concern about CDE-related costs). Physician education about these and other issues related to patient follow-up should be considered in developing and implementing future colorectal cancer screening programs. In addition, prospective studies should be conducted to determine how well the variables identified here, including intention, and other candidate variables serve to predict physician behavior. Research is also needed to identify patient-related factors (e.g., refusal to have CDE) that impact CDE rates. With funding provided by the National Cancer Institute, research to address these questions has been initiated in the context of the Aetna US Healthcare’s colorectal cancer screening program. Findings from this investigation will help to inform the development of approaches that may be used to maximize CDE performance and, as a consequence, contribute to the goal of reducing colorectal cancer mortality.

The extent to which findings reported here can be generalized may be limited by several factors. As mentioned earlier, we found that survey respondents and nonrespondents were comparable in terms of practice specialty, number of physicians in the practice, and state in which the practice was located. However, it may be that these two groups did differ in terms of important attitudinal characteristics that may influence intention to recommend CDE. It is also possible that our observations were biased if the respondents tended to provide what they felt were “right” answers to questions about FOBT+ patient follow-up. There were no anecdotal reports by survey interviewers to suggest that this phenomenon occurred, however. It is also possible that physicians who participated in the telephone survey, as compared with those who did not, were more knowledgeable about FOBT screening and were more favorably disposed to CDE. Furthermore, it should be noted that the FOBT+ patient follow-up scenario that was included on the survey related to a FOBT mail-out screening program. Physician intention to recommend CDE in this context may differ from CDE intention in a situation where tests are given to the patient in a physician-patient encounter. Finally, the survey data used in this study were collected in 1994. Developments in colorectal cancer screening since that time, including new reports in the literature on colorectal cancer screening and the dissemination of practice guidelines, could have modified primary care physician attitudes related to CDE.

We thank Dr. Paul F. Engstrom for his support of this study. In addition, we acknowledge Debra Ivany for critical reading of the manuscript. We also extend our appreciation to Dr. Sue M. Marcus, Dr. Eric Ross, Andrew Balshem, and Peter Preston for assistance in conducting initial data analyses. Finally, we thank Trena Diggs for secretarial support in preparing the manuscript.