Preferences of Community Physicians for Cancer Screening Guidelines

Early-Detection Guidelines

Establishing guidelines for early detection of cancer has been a major activity of both voluntary and government agencies. The American Cancer Society has a long-standing policy of publishing guidelines for cancer-related check-ups; the most recent guidelines were published in 1985 [9]. Following a directive from the Assistant Secretary for Health [10-12], the National Cancer Institute established objectives for reducing mortality caused by cancer. The Early Detection Branch of the Division of Cancer Prevention and Control at the National Cancer Institute developed and disseminated early-detection guidelines in 1987 [12], which paralleled those published by the American Cancer Society.

The Canadian Task Force on Periodic Health Examination and the United States Preventive Services Task Force also published guidelines [6, 8, 13], which are generally more conservative than those of the National Cancer Institute and the American Cancer Society. For example, the American Cancer Society and the National Cancer Institute advocate periodic mammography every 1 to 2 years for women aged 40 to 49 years and annual mammography for women older than 50 years. Neither of the task forces currently recommends mammography for women younger than 50 years [9, 12, 16-18]. The National Cancer Institute and the American Cancer Society advocate fecal occult blood tests, flexible sigmoidoscopy for persons older than 50 years, and digital rectal examinations, whereas the task forces make no recommendations on these procedures [8, 9, 12, 13]. Generally speaking, the American College of Physicians and the American Academy of Family Practice use the guidelines established in the Guide to Clinical Preventive Services, the “Blue Book” [6-8]. The American College of Surgeons and the American College of Gynecologists generally follow the American Cancer Society guidelines [6-9]. However, as shown below, these are not consistent patterns.

Methods

Study Design

Data were collected through a national survey of physicians in four specialties (family practice, internal medicine, general surgery, and gynecology) who did not list oncology or hematology as subspecialties. The survey was conducted as part of the evaluation of the National Cancer Institute's Community Clinical Oncology Program [19]. Two samples of physicians were selected: community physicians who had admitting privileges at hospitals associated with the program (the evaluation sample) and a national sample of community physicians derived from the American Medical Association's listing of physicians in all 50 states and Puerto Rico (the comparison sample). Both samples were disproportionate, stratified, and randomly selected.

Interviews were conducted between 2 October 1989 and 31 March 1990. Eligibility was ascertained by telephone, and those eligible were interviewed by mail or by telephone, depending on their preference. After screening, the total number of eligible physicians included 2655 from the evaluation sample and 2451 from the comparison sample. Interviews were completed with 3436 physicians: 1898 (71%) from the evaluation sample and 1538 (63%) from the comparison sample. Completion rates by specialty were 74% for surgeons, 73% for gynecologists, 63% for family practitioners, and 55% for general internists [19].

Defining Preference for Screening Interval

For each screening procedure, respondents were asked to select from a list the screening interval they would recommend for patients with no known risk for the relevant cancer. The questions had a common format. The question for clinical breast examination read, “How often would you recommend breast palpation by a physician for women who have no known risk factors?” Response choices for each question were “every 6 months or less,” “every 7 to 11 months,” “every 12 months,” “between 1 to 2 years,” “every 2 years or more,” and “other (specify).” For each question, consensus was defined as the modal response from all physicians. Because the existing guidelines for mammography specify different screening intervals for patients aged 40 through 49 years than for those 50 years and older, separate questions were asked to establish consensus on the appropriate interval for women in each age category.

Factors associated with the modal response were compared with those associated with other screening intervals. Highly unlikely responses (outliers) were deleted (generally 1% to 3% of all responses). In most cases, this procedure resulted in two clearly defined alternatives. However, there were exceptions. The variation from the mode for clinical breast palpation did not cluster, so the comparison was between annual palpation (the mode) and any other preference. For sigmoidoscopy, two comparisons were tested. The mode was “every 2 years or less frequently.” It was compared with two secondary modal points: “annual/between 1 and 2 years” and “never”

Physician and Practice Characteristics

Two sets of variables were used to examine physician preference: physician specialization and physician and practice characteristics. Surgeons were chosen as the reference category for specialization. Thus, family practice, general internal medicine, and gynecology were compared with surgery.

We assessed whether cancer control initiatives in the Community Clinical Oncology Program influenced community physicians to be more sensitive to guidelines for early detection than other community physicians. Because samples were selected from different frames, we combined the two samples and controlled for the sample frame. This simultaneously assessed the question of program influence and controlled for selection bias.

It was hypothesized that recent medical school graduates and those not in practice by themselves might have the most access to current information about recommended screening intervals. Preferences among those who graduated from medical school more than 20 years ago were compared with preferences among more recent graduates. Similarly, preferences of those in solo practice were compared with those of physicians in group or other practice settings.

A series of questions assessed belief in the effectiveness of screening. Respondents were asked to rate the effectiveness of early detection in improving the likelihood of surviving cancers of the colon, rectum, breast, lung, prostate, bladder, and skin. Respondents were not asked about their assessment of the effectiveness of early detection of gynecologic cancers. Instead, a summary variable of overall assessment of screening effectiveness, based on the physicians' ratings of the other seven items, was used.

Logit analysis with maximum-likelihood estimation [20] was used to model the main effects. In each case, the residual chi-square was not significant, suggesting no need to explore interaction effects.

Results

To assess the recommendations of various professional groups and of the U.S. Preventive Services Task Force, the American Cancer Society, and the National Cancer Institute, the summary provided by Hayward and colleagues [8] was used. The recommendations by the American Cancer Society and the National Cancer Institute are summarized in Table 1. We rate the extent to which other groups agreed with these recommendations, using four categories: “high” (all specialty groups, including the American Academy of Family Practice, the American College of Physicians, the American College of Surgeons, the American College of Gynecologists, and the U. S. Preventive Services Task Force, agree with the recommendations of the American Cancer Society and the National Cancer Institute), “moderate” (all agree except the U.S. Preventive Services Task Force), “weak” (the American College of Physicians, the American College of Gynecologists, the American Cancer Society, and the National Cancer Institute agree); and “very weak” (only the American Cancer Society and the National Cancer Institute agree). For the Papanicolaou smear, the consensus on interval is rated “high,” but there are differences regarding the age at which screening is recommended. The American College of Physicians and the U. S. Preventive Services Task Force recommend cessation of screening after the age of 65 years, and the American College of Physicians recommends starting screening at the age of 20 years rather than at the age of 18 years.

All published guidelines are consistent in their recommendation regarding the clinical breast examination, mammography for women aged 50 years and older, the chest roentgenogram (which is not recommended by any group), and the Papanicolaou smear; the guidelines vary in recommendations regarding age. Thus, these four procedures are rated “high” in consensus among published guidelines. Recommendations vary by group for all other procedures. Only the American Cancer Society and the National Cancer Institute make recommendations regarding all eight procedures.

The consensus or modal choice of screening interval by responding physicians matched or encompassed the recommendations by the American Cancer Society and National Cancer Institute for every procedure but the chest roentgenogram. However, the modal value varied widely by procedure from 83% for the fecal occult blood test to 51% for mammography for women aged 40 through 49 years (Table 1). Moreover, a consistent relation does not appear to exist between the extent of agreement among the published guidelines and consensus among responding physicians. The extent of overall consensus across specialties was also highly variable (Table 1).

A consensus of 70% or more with the overall mode across specialization was found for the clinical breast examination, the fecal occult blood test, the Papanicolaou smear, and the pelvic examination (Table 1). Among these four procedures, the most consistent consensus was found for the fecal occult blood test, for which the range across physician specialty was 80% to 85%. The next highest consensus was found for the clinical breast examination, for which the interspecialty range was 77% to 87%, although the intraspecialty variation was higher. The consensus choice for the appropriate screening interval for the Papanicolaou smear and pelvic examination was shared by at least 70% of physicians in each specialty group, but the range was wide. Of the procedures in this group, annual clinical breast examination and the Papanicolaou smear are recommended by all published guidelines. “Weak” agreement was found among published guidelines recommending the annual fecal occult blood test. The annual or less frequent pelvic examination had a “very weak” rating.

Three procedures—mammography for women aged 50 years and older, digital rectal examination, and chest roentgenogram—achieved 60% or higher consensus across physician specialties. There was 17% variation by specialty in the degree of acceptance of the consensus choice for annual mammography for women aged 50 years and older, and all guidelines recommend annual mammography for these women. The 62% to 79% consensus on annual mammography for the older group of women reflects the preference of a substantial minority in all specialties for less frequent screening (gynecologists were most likely to prefer the consensus). Variation from the modal preference for annual digital rectal examinations indicated a general preference for less frequent screening except among the gynecologists, who strongly preferred the annual interval. The rating for the digital rectal examination was “very weak.” The chest roentgenogram was the third procedure for which a 60% or higher consensus was found. No published guidelines recommend this procedure. Nevertheless, two thirds of family practitioners, general internists, and gynecologists and 82% of the surgeons agreed that the procedure should be done at least every 2 years when the patient has no significant risk for lung cancer.

The least consensus was found among respondents on the appropriate screening intervals for mammography in women aged 40 through 49 years and for flexible sigmoidoscopy in persons aged 50 years and older. Biennial screening intervals for mammography in women aged 40 through 49 years was endorsed most consistently by general surgeons and general internists; family practitioners and gynecologists were about equally divided between the consensus choice and a less frequent interval (every 2 years or less often). Mammography for women aged 40 through 49 years received a “moderate” rating for agreement among published guidelines. A higher degree of consensus that flexible sigmoidoscopy should be done at intervals of 2 years or less frequently for persons older than 50 years was observed among family practitioners and general internists than among general surgeons or gynecologists. Agreement among written guidelines was rated “weak.”

The relations between physician and practice characteristics and screening interval preference are summarized in Table 2. Except for sigmoidoscopy, the reference category is the consensus response. In the case of sigmoidoscopy, the modal category (every 2 years or less frequently) was compared with “annual/between 1 and 2 years” and “never.”

The first comparisons were made between general surgeons (as the reference category) and the other specializations. Internists were similar to surgeons in that they preferred screening intervals that were more frequent than those preferred by either gynecologists or family practitioners. When they disagreed, internists favored less frequent mammography for both age groups, favored less frequent sigmoidoscopy, and were less likely to ever favor screening with chest roentgenogram. Family practitioners were more likely than general surgeons to endorse the consensus for pelvic examinations, Papanicolaou smears, and sigmoidoscopy and to favor less frequent intervals for mammography in both age groups. Gynecologists differed substantially from general surgeons in all 10 comparisons. Five of the comparisons were particularly relevant to the health of women. Gynecologists were more likely than surgeons to endorse the consensus category for Papanicolaou smears and pelvic examinations and were less likely to favor doing mammography annually or every 1 to 2 years for women aged 40 through 49 years. Gynecologists were also more likely to endorse annual fecal occult blood tests, annual digital rectal examinations, and the consensus for sigmoidoscopy and less likely to ever recommend chest roentgenogram for patients at normal risk.

Physicians who believed in the effectiveness of early detection in reducing mortality consistently favored the consensus category for mammography (both age groups), fecal occult blood test, digital rectal examination, Papanicolaou smear, and pelvic examination. Those with the strongest belief in the effectiveness of early detection of lung cancer also endorsed the most frequent screening interval for the chest roentgenogram (Table 2).

Physicians who graduated from medical school 20 years or more before the survey was conducted favored less frequent screening for six procedures: clinical breast palpation, mammography for women aged 50 years and older, fecal occult blood test, digital rectal examination, Papanicolaou smear, and pelvic examination. However, they were more likely than recent graduates to endorse screening with the chest roentgenogram. Being in solo practice was associated with preference for more frequent sigmoidoscopy and with more consistent support for the consensus for use of the chest roentgenogram.

Discussion

Overall, community physicians favored the screening intervals recommended by the American Cancer Society and the National Cancer Institute. For all but two screening procedures—mammography for women aged 40 through 49 years and flexible sigmoidoscopy—the mode exceeded 60%. In the case of mammography for younger women, the lack of consensus reflected differences about the appropriate interval and not whether screening should be done. The “never” response was significant only for flexible sigmoidoscopy and chest roentgenogram, exceeding 10% in each case.

Variability across physician specialty was present in the extent to which physicians endorsed the modal (consensus) judgment for each procedure; it was least for fecal occult blood tests and varied for other procedures. For every procedure but the chest roentgenogram, the consensus matched or encompassed the American Cancer Society and National Cancer Institute recommendations. Variability by specialization was not more common for procedures for which variability was present among published guidelines, and variation among specialization was not associated with the position of the published guidelines endorsed by the professional groups of the responders.

Surgeons tended to favor shorter screening intervals than did either family practitioners or general internists, although family practitioners were more likely than surgeons to agree with the recommended interval for Papanicolaou smears and pelvic examinations. Gynecologists more frequently preferred the modal judgment about screening intervals for cancers occurring in women than did surgeons, except they preferred less frequent mammography for women aged 40 through 49 years.

Physician specialty and practice setting were also associated with screening interval preference. Older physicians tended to favor more conservative screening. They consistently favored chest roentgenogram for screening, which the published guidelines universally agree is not effective. Those who believed screening was “very effective” consistently supported the consensus. Those in solo practice also tended to favor more frequent use of flexible sigmoidoscopy and chest roentgenogram compared with those in other forms of practice.

The degree of acceptance of chest roentgenogram is interesting and suggests that community physicians as a group may not be attuned to the most current screening interval guidelines [9, 21]. Although all published guidelines recommend against chest roentgenogram, more than 60% of surveyed physicians specified a screening interval, a result consistent with other recent surveys [9, 22, 23].

In general, the literature on physician beliefs about screening is based on small samples drawn from local or statewide populations. Questions do not focus on screening interval but typically use general wording to ask about practice (for example, whether the physician “ever” or “routinely” does specific procedures) [9, 21-32] rather than asking about specific beliefs, as we did. The absence of national samples and variations in wording make comparisons difficult. Nevertheless, it appears that physicians provide cancer screening at frequencies well below those recommended by the American Cancer Society or the National Cancer Institute. Although according to these surveys, the guidelines for breast palpation, breast self-examination, and Papanicolaou smear appear to be widely accepted and followed [9, 21-23, 25, 28-30, 32], physicians do not routinely recommend mammography [9, 19, 22, 24, 26, 27, 31] or do pelvic examinations [19], digital rectal examinations, stool guaiac tests, or sigmoidoscopy [9, 19, 22, 23, 28] in patients with no known risk factors for cancer. Many continue to use the chest roentgenogram to screen for lung cancer [9, 22, 23], despite unanimous recommendations against this. However, this may be the first analysis that has specifically examined the relation between consistency in published guidelines and preference for screening interval, which is the subject of the guidelines.

Physician and practice characteristics are well established predictors of practice patterns [33-35]. Other studies [33, 34, 36] have found that years in practice, type of practice, and specialty are related to physician practice patterns. For example, several studies have found that younger physicians prescribe more ancillary services than do their older colleagues [35, 36]. Moreover, those who practice alone are less likely to be aware of new information on patient care [34]. Our findings seem to be consistent with this literature.

We did not examine practice behavior but looked instead at physician beliefs about the most appropriate screening intervals. Economic considerations, perceived test reliability, patient acceptance, and demands or limitations in the practice setting all affect actual practice, regardless of belief. What was unique about our study was the focus on norms or physician beliefs about appropriate screening intervals, regardless of practice, and the factors associated with those beliefs.

These data were collected for an evaluation of a National Cancer Institute program to introduce clinical and cancer control research into community settings. The survey was done to determine whether community physicians in participating hospitals knew more about cancer control than did a general sample of community physicians. The results indicate no differences between the two samples, and they were combined.

This analysis used the largest sample of community physicians reported in the literature. Thus, the estimates of physician preference are likely to be stable. Response rates were less than desirable for some physician specialties, consistent with other physician surveys. Comparisons between early responders and those responding after many attempts were made to contact them (a group that may be similar to nonresponders) indicated no correlation between time of response and screening interval preference.

The significance of the results are as follows: First, community physicians accept the idea of screening, even when screening is not recommended; second, physicians tend to accept the most heavily publicized guidelines even when they differ from recommendations by their own professional societies (this suggests that changes in screening practices, even when new evidence questions the cost benefits associated with a practice, will be difficult [35]); third, physician specialization and characteristics do influence screening interval preference.