Journal of General Internal Medicine

, Volume 28, Issue 1, pp 18–24

Variation in Primary Care Physicians’ Colorectal Cancer Screening Recommendations by Patient Age and Comorbidity


    • VA Health Services Research & Development Center of Excellence on Implementing Evidence-Based PracticeRichard L. Roudebush VA Medical Center
    • Division of General Internal Medicine and Geriatrics, Department of MedicineIndiana University School of Medicine
    • IU Center for Health Services and Outcomes ResearchRegenstrief Institute, Inc.
  • Carrie N. Klabunde
    • Health Services and Economics Branch, Applied Research Program, Division of Cancer Control and Population SciencesNational Cancer Institute
  • Judith Lee Smith
    • Division of Cancer Prevention and ControlCenters for Disease Control and Prevention
  • Gigi Yuan
    • Information Management Services, Inc.
Original Research

DOI: 10.1007/s11606-012-2093-6

Cite this article as:
Haggstrom, D.A., Klabunde, C.N., Smith, J.L. et al. J GEN INTERN MED (2013) 28: 18. doi:10.1007/s11606-012-2093-6



Screening patterns among primary care physicians (PCPs) may be influenced by patient age and comorbidity. Colorectal cancer (CRC) screening has little benefit among patients with limited life expectancy.


To characterize the extent to which PCPs modify their recommendations for CRC screening based upon patients’ increasing age and/or worsening comorbidity


Cross-sectional, nationally representative survey.


The study comprised primary care physicians (n = 1,266) including general internal medicine, family practice, and obstetrics-gynecology physicians.


Physician CRC screening recommendations among patients of varying age and comorbidity were measured based upon clinical vignettes. Independent variables in adjusted models included physician and practice characteristics.


For an 80-year-old patient with unresectable non-small cell lung cancer (NSCLC), 25 % of PCPs recommended CRC screening. For an 80-year-old patient with ischemic cardiomyopathy (New York Heart Association, Class II), 71 % of PCPs recommended CRC screening. PCPs were more likely to recommend fecal occult blood testing than colonoscopy as the preferred screening modality for a healthy 80-year-old, compared to healthy 50- or 65-year-old patients (19 % vs. 5 % vs. 2 % p < 0.001). For an 80-year-old with unresectable NSCLC, PCPs who were an obstetrics-gynecology physician were more likely to recommend CRC screening, while those with a full electronic medical record were less likely to recommend screening.


PCPs consider comorbidity when screening older patients for CRC and may change the screening modality from colonoscopy to FOBT. However, a sizable proportion of PCPs would recommend screening for patients with advanced cancer who would not benefit. Understanding the mechanisms underlying these patterns will facilitate the design of future medical education and policy interventions to reduce unnecessary care.


cancer screeninghealth servicescolorectal cancerprimary care physicians


Colorectal cancer (CRC) screening is recommended by multiple organizations for patients age 50 and older.1,2 However, patients with advanced age and high severity of illness are less likely to benefit from early detection.3 Most clinical guidelines account for clinical judgment and patient choice in the recommendation of whether and how to screen.

The majority of cancer screening tests are initiated within primary care physician (PCP) offices. From a nationwide survey that administered clinical vignettes to PCPs, we set out to answer the question of how CRC screening decisions are influenced by a patient’s age and comorbidity. We also tested what physician and practice characteristics are associated with CRC screening recommendations.



We surveyed a nationally representative sample of PCPs between September 2006 and May 2007 in the National Survey of Primary Care Physicians’ Recommendations and Practices for Breast, Cervical, Colorectal, and Lung Cancer Screening. The American Medical Association's Physician Masterfile, which contains demographic and practice information on all allopathic and nearly all osteopathic physicians in the U.S., was used as the sampling frame. Eligible respondents were office-based family physicians, general practitioners, general internists, and obstetrician/gynecologists aged ≤75 years.

A systematic, stratified random sample of 2,576 physicians was selected from the Masterfile, with the four specialty types used as sampling strata. Sampling was designed to provide national estimates of physicians' cancer screening recommendations for the four specialties combined with a margin of error of ±3 % at a 95 % CI. The sample was selected after sorting the sampling-frame database by physicians' age, gender, urban versus rural practice location, and U.S. census region. Physicians were then excluded who had requested on the Masterfile that they not be contacted (4 %), for whom current contact information could not be obtained (7 %), or who were reclassified as ineligible after a screening telephone call. Physicians were reclassified as ineligible if they reported not providing primary care, providing clinical care for <1 day per week, practicing in a federal healthcare facility, or practicing outside the U.S. Physicians were also ineligible if retired or deceased.


A total of 1,975 physicians were mailed a packet including a questionnaire on colorectal and lung cancer screening; support letter from the American Academy of Family Physicians, the American Society of General Internal Medicine, or the American College of Obstetricians and Gynecologists, depending on the physician's specialty; and $50 honorarium check. Three physicians completed a telephone interview. The survey’s absolute response rate was 69.3 %; the cooperation rate, which excludes physicians lacking valid contact information, was 75.0 %. Full survey administration details have been reported elsewhere.4

Clinical Vignettes

We assessed CRC screening recommendations with nine clinical vignettes that varied the patient's age and comorbidity (Text box 1). Clinical vignettes are a useful method for understanding how physicians respond to specific clinical situations 5,6 and have been employed by other investigators to study CRC screening decisions.7,8

Text Box 1. Clinical vignettes

Age was varied across three categories: age 50, 65, and 80. Comorbidity was varied across the following categories: “healthy” patient (no comorbidity), patient “with ischemic cardiomyopathy who experiences dyspnea with ordinary activity (NY Heart Association Class II) treated with appropriate medication” (congestive heart failure (CHF), moderate comorbidity), and patient “with unresectable non-small cell lung cancer” (NSCLC, severe comorbidity). For each vignette, physicians were asked to indicate which of the following they would recommend: both fecal occult blood test (FOBT) and flexible sigmoidoscopy, FOBT only, flexible sigmoidoscopy only, both FOBT and colonoscopy, colonoscopy only, or no screening.

The order of the vignettes progressed from less to more severe comorbidity, then within comorbidity categories, progressed from younger to older patients; the order of the vignettes or response choices was not varied between surveys. Regarding consistency of the vignette items among our respondents, the value of the Cronbach's alpha for the nine items was 0.73, with the outcome of interest dichotomized as 'recommends screening' vs. 'Does not recommend screening'. This result indicates that there was good consistency (vs. randomness), in physicians' responses to these items. An open-ended “other” response was also possible from which the category CT colonography (virtual colonoscopy) was coded as a separate category, but a rare one (0.5 % of all vignettes). Otherwise, the unstructured data was recoded into one of the available categories when appropriate.

Physician/Practice Characteristic Measurements

Physician sociodemographic characteristics were assessed. Physician attitudes included the influence of (1) patient preferences and (2) guidelines from the United State Preventive Services Task Force (USPSTF) or American Cancer Society (ACS) upon their CRC screening recommendations. Measured by physician self-report, practice characteristics included type of practice, ownership, location, percentage of uninsured patients, type of medical record system, use of physician and patient computer reminder systems, and receipt of reports on CRC screening rates. The complete survey instrument is available online.9

Data Analysis

Descriptive statistics characterized the physicians surveyed. The proportions of physicians who recommended different types of CRC screening tests were described, and stratified by patient age and comorbidity. Categorical comparisons of different CRC screening recommendations for patients with varying levels of comorbidity, or between different test choices, were performed with chi-squared tests. Logistic regression models were used to test the association between physician or practice characteristics and a positive recommendation for CRC screening (of any type). The models focused on responses to the vignettes describing 80-year-old patients of varying comorbidity because we wanted to assess PCPs’ screening recommendations among the elderly. A sample weight that adjusted for undercoverage, nonresponse to telephone-screener calls, and survey nonresponse was assigned to each respondent. Sample weights were applied in the statistical analyses to obtain national estimates. SAS version 9.1.3 was used to obtain unweighted sample sizes. SUDAAN version 9.0.1, which accounted for the complex survey design, was used to apply the sampling weights in calculating percentages and 95 % CIs in the descriptive statistics and for regression models.


Physician/Practice Characteristics

Most physicians were either general internists (37 %) or family/general practice physicians (45 %); fewer were obstetrics/gynecology physicians (18 %). For medical record systems, 18 % of physicians reported a full electronic medical record (EMR), 26 % either a partial EMR or transitioning to an EMR, and 56 % a paper chart. Table 1 describes additional physician and practice characteristics, including physician sociodemographics and attitudes.
Table 1

Physician and Practice Characteristics


Number (unweighted)

Percentage (weighted)




251 (20.1 %)


385 (30.7 %)


398 (31.9 %)


232 (17.4 %)



400 (31.2 %)


  White, Non-Hispanic

925 (72.1 %)

  Black, Non-Hispanic

46 (3.9 %)


66 (5.5 %)


184 (14.7 %)

  Other (AIAN, Hawaiian/Pacific Island, Multiple, Other)

45 (3.8 %)


  General Internal Medicine

415 (36.9 %)

  Family Practice

481 (41.6 %)

  General Practice

66 (3.6 %)


304 (17.9 %)

 Board certified


999 (80.2 %)

 International Medical Graduate


265 (21.7 %)

 Medical school affiliation


441 (35.1 %)

 Patient Preferences for CRC screening

  Very Influential

323 (26.2 %)

  Somewhat Influential

659 (52.5 %)

  Not Influential/Not Applicable/Not Familiar with/Missing

284 (21.2 %)

 U.S. Preventive Services Task Force guidelines

  Very Influential

799 (63.7 %)

  Somewhat Influential

354 (28.2 %)

  Not Influential

22 (1.7 %)

  Not Applicable/Not Familiar with/Missing

91 (6.4 %)

 American Cancer Society guidelines

  Very Influential

859 (67.3 %)

  Somewhat Influential

357 (28.6 %)

  Not Influential

18 (1.5 %)

  Not Applicable/Not Familiar with/Missing

32 (2.7 %)


  Solo Practice

328 (26 %)

  Practice Size (2-100+), Single specialty

614 (48.1 %)

  Practice Size (2-100+), Multi-specialty

292 (23.3 %)

 Rural/urban location


1032 (81.8 %)

  Large rural city/town

133 (10.2 %)

  Small/isolated small rural town

101 (8 %)

 Geographic region


246 (20.1 %)


318 (23.1 %)


420 (33.2 %)


264 (22.6 %)

 Main primary care practice location

  Full owner, part owner, or employee of a physician owned practice

827 (64.5 %)

  Employee of a large medical group, health care system, group or staff model HMO

197 (16.5 %)

  Employee of a university or non-university hospital or clinic or other practice

235 (18.4 %)

 Percentage of patients uninsured

  0-5 %

759 (60.3 %)

  6-25 %

366 (28.8 %)

  26 %+

82 (6.4 %)

 Number of patients seen in main primary care practice location in a typical week

  25 or fewer

62 (5 %)


155 (12 %)


224 (17.3 %)


404 (31.9 %)


245 (20 %)

  126 or more

154 (12.2 %)

 Type of medical record system used in the practice

  Paper charts

709 (55.7 %)

  Transitioning from paper to EMR

196 (15.5 %)

  Partial EMR

133 (10.3 %)

  Full EMR

218 (17.7 %)

 Physician computer reminders: Computer prompt or computer-generated flow sheet


156 (13.3 %)

 Patient computer reminders:Reminder by email


10 (0.8 %)

 Receive reports on rates of CRC screening


146 (11.9 %)

*percentages may not sum to 100 % because of item non-response

†CRC colorectal cancer, AIAN American Indian or Alaska Native, EMR electronic medical record

Age/Comorbidity/Screening Recommendation

The recommendations for CRC screening tests based on clinical vignettes, stratified by age and comorbidity, are visualized in Figure 1. Among patients age 50, physicians were very likely to recommend screening if the patient was healthy (99.6 %) or had CHF (98.4 %). Recommendations were similar for patients age 65, with physicians very likely to recommend screening if the patient was healthy (100 %) or had CHF (97 %). Among patients age 80, physician recommendations differed more markedly between patients who were healthy and those who had CHF (90 % vs. 71 %; p < 0.001). Among patients with unresectable NSCLC, physician recommendations for screening declined with advancing age: age 50 (42 %), age 65 (38 %), and age 80 (25 %); p < 0.001 for trend. Across all age groups, physicians were less likely to recommend any screening among patients with unresectable NSCLC (34 %) than among those who were healthy (96 % vs. 34 %, p < 0.001) or had CHF (89 % vs. 34 %, p < 0.001).
Figure 1.

Recommendation for CRC screening test stratified by patient age and comorbidity. FOBT fecal occult blood test; flex sig flexible sigmoidoscopy; CHF congestive heart failure; NYHA New York Heart Association; NSCLC non-small cell lung cancer.

Age/Comorbidity/Choice of Test

Among physicians who recommended any type of screening test, we examined the association between the patient’s age and the type of screening test, including FOBT alone, colonoscopy (with or without FOBT), and all other tests. Among patients who were healthy, physicians less often recommended FOBT alone than colonoscopy (with or without FOBT) for patients who were age 50 (5 % vs. 85 %), age 65 (2 % vs. 91 %), or age 80 (19 % vs. 73 %), with a chi-square p value of <0.001. Among patients who had CHF, physicians still less often recommended FOBT alone than colonoscopy for patients who were age 50 (16 % vs. 66 %) or age 65 (19 % vs. 64 %), although physicians more often recommended FOBT alone than colonoscopy for patients age 80 (54 % vs. 32 %), p < 0.001. Among patients with unresectable NSCLC, physicians recommended FOBT alone with the same frequency or more often than colonoscopy for patients who were age 50 (46 % vs. 46 %), age 65 (50 % vs. 41 %), and age 80 (65 % vs. 28 %), p < 0.001.

Physician/Practice Characteristics Associated with Screening Recommendation

Several physician and practice characteristics were associated with a positive recommendation for CRC screening (of any type) for 80-year-old patients who were healthy, had CHF, or NSCLC (table available online). In adjusted models for an 80-year-old patient with NSCLC, physicians who were OB-GYN or female were more likely to recommend CRC screening, while patients who were board-certified or had a full EMR were less likely to screen.


Using data from a large, nationally representative survey of PCPs, we showed that physician screening decisions were sensitive to both patient age and comorbidity. Almost all PCPs recommended some type of screening for patients age 50 or age 65 who were either healthy or had CHF. However, the proportion of PCPs who recommended screening declined among 80-year-old patients across all clinical categories.

USPSTF guidelines recommend against screening patients with limited life expectancy,1 and the 5-year survival for a population with unresectable NSCLC (Stage III or greater) is 7 % or less.10 PCPs recommended CRC screening among patients with limited life expectancy (unresectable NSCLC) less often than for patients with less severe illness (CHF). Nonetheless, 25 % of PCPs still recommended CRC screening for an 80-year-old patient with unresectable NSCLC; among physicians who recommended screening in this case, most (65 %) recommended FOBT only.

Other studies, largely conducted in VA settings, have reported how often PCPs recommend screening for patients with severe illness. A claims-based VA study reported high rates of CRC screening among patients with metastatic tumors or severe comorbidity.11 In a survey of 183 VA PCPs, 10 % of PCPs recommended screening for an 80-year-old patient with an “active malignancy outside the colon”, and 7 % recommended screening among patients “with limited life expectancy”.8 Compared to the VA survey, our study suggests an even higher percentage of PCPs (≥25 %) recommend CRC screening among patients who are unlikely to benefit. Our results are more broadly generalizable to the U.S. health care system because our survey population represents a nationally representative sample of practicing PCPs. Ultimately, recommendations to perform CRC screening among patients of any age with unresectable NSCLC can be described as having no demonstrable clinical benefit,12 or even being harmful. Harms related to unnecessary cancer screening may include false positives, anxiety, and procedural risks, for example, if a positive FOBT were to lead to colonoscopy. Harms from therapy may result from the treatment of pseudodisease, i.e., cancers that would never cause problems during a person’s lifetime even if left untreated.13

PCPs with a full electronic medical record (EMR) were less likely to screen an 80-year-old patient with NSCLC than those with less developed EMR capability; this finding suggests that health information technology, in addition to increasing guideline-concordant care,14 may have the potential to reduce unnecessary care. PCPs identified as OB-GYN physicians were significantly more likely to recommend screening among patients with moderate or severe comorbidity; OB-GYN physicians are also more likely to recommend inappropriate Pap tests.15 More research to understand the mechanisms of how such factors influence the over-recommendation of CRC screening may inform future medical education and policy interventions.16

It is unclear why physicians would recommend CRC screening among patients with unresectable NSCLC when there is no clear clinical benefit. Quality measures that promote CRC screening, without regard to comorbidity, may contribute to overscreening.17 Clinical considerations relevant to the patient–physician relationship may also influence such recommendations, specifically, PCPs may not wish to withdraw hope and so continue to offer screening to patients with limited life expectancy.18 This explanation is consistent with low-risk FOBT (compared to colonoscopy) being the predominant screening modality recommended by PCPs for patients with unresectable NSCLC. Such non-invasive screening may have the unintended consequence of delaying clear, informed discussions regarding prognosis. Further research is needed to better understand how physicians can develop more comfort in stopping screening when there is no clinical benefit, as well as quantifying the impact of overscreening upon patient outcomes through the exposure to procedural risks. This work should be done while balancing the need to make certain that healthy patients with the highest likelihood of benefit from screening still have access to services that reduce mortality.

Limitations of our study include the lack of direct measurement of health care utilization; nonetheless, responses to clinical vignettes are often consistent with clinical practice.5,6 Another limitation is that our physician survey was fielded in 2006–7, prior to the change in the USPSTF guidelines (2008) which recommended against routine CRC screening in adults 76 to 85 years of age. With the change in guidelines, physicians may be less likely to recommend screening among healthy 80-year-old patients.

Most physicians appear to shift their recommendations in an appropriate manner in response to varying patient age and comorbidity. Yet there were a significant proportion of outliers who recommend CRC screening among patients with limited life expectancy for whom screening tests have no benefit and possible harm. Health care systems should be organized to provide a realistic amount of time19 for physicians to have meaningful discussions with patients and their families about comorbidity, prognosis, and the utility of screening. While the majority of Americans trust information about medical topics provided by a doctor or health care professional,20 improvement in medical education and professionalism curricula may help to promote honest, compassionate physician communication around issues of unnecessary care16 and policy changes to reduce over-utilization.21


Funding support for this study was provided by the National Cancer Institute (contract number N02-PC-51308), the Agency for Healthcare Research and Quality (inter-agency agreement numbers Y3-PC-5019-01 and Y3-PC-5019-02), and the Centers for Disease Control and Prevention (inter-agency agreement number Y3-PC-6017-01). Dr. Haggstrom is the recipient of VA HSR&D Career Development Award CD207016-2.

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the National Cancer Institute, Department of Veterans Affairs, or the Centers for Disease Control and Prevention.

Conflict of Interest

The authors declare that they do not have a conflict of interest.

Supplementary material

11606_2012_2093_MOESM1_ESM.doc (79 kb)
ESM 1(DOC 79 kb)

Copyright information

© Society of General Internal Medicine 2012