A nationally representative sample of PCPs participated in a survey between September 2006 and May 2007, sponsored by the National Cancer Institute (NCI) in collaboration with the Agency for Healthcare Research and Quality (AHRQ) and the Centers for Disease Control and Prevention (CDC). Eligible physicians included office-based family physicians, general internists, general practitioners, and obstetrician/gynecologists (OB/GYNs) aged 75 or younger. A sample of PCPs was selected from the American Medical Association’s Physician Masterfile using the four specialties as sampling strata. Prior to selection, the sampling frame database was sorted by age, gender, US Census region, and urban-rural practice location within each stratum. Detailed survey content and sampling strategy are reported elsewhere.12,13
The study was reviewed by the institutional review boards of the NCI and CDC and determined to be exempt.
In September 2006, 1,975 PCPs were sent a questionnaire on colorectal and lung cancer screening via express mail, and several methods were used to encourage participation.14
These methods included letters of endorsement from national physician organizations, postage-paid return envelopes, a $50 honorarium, two additional mailings, up to three follow-up telephone reminders to non-respondents, and the option to complete the survey by telephone. Additional details have been published elsewhere.12
The dependent measure of interest was physicians’ reports of changes in the volume of screening tests they ordered, performed, or supervised in the past 3 years for FOBT, sigmoidoscopy, and colonoscopy. Response categories were on a 5-point scale (increased substantially, increased somewhat, stayed the same, decreased somewhat, or decreased substantially).
Guided by the social ecological perspective—which recognizes the influence of multi-level factors on behavior—and the Theory of Planned Behavior,15,16
we assessed the relationship of selected factors with test ordering reports. The Theory of Planned Behavior highlights the potential influence on behavior of a physician’s attitudes, perceived norms, and perceived challenges.
Physicians’ attitudes toward colonoscopy screening included agreement with three statements: “it is the best of the available CRC screening tests,” “it is readily available for my patients,” and “I worry that I might be sued if I do not offer screening colonoscopy to my patients.” Responses to these statements were measured on a 4-point scale, from “strongly agree” to” strongly disagree.” Additionally, we created a summary measure of the strength of agreement across the three items: strongly agree with none of the three items, strongly agree with one of the three items, strongly agree with two, or strongly agree with all three items.
Three items related to reported influence of perceived norms. Respondents rated local collegial norms, patient preferences, and national recommendations (USPSTF, ACS, and published clinical evidence) on a scale of “very influential,” “somewhat influential,” and “not influential” to their practice of CRC screening. Respondents were also given a “not applicable/not familiar” choice.
With respect to perceived challenges, they were asked how influential two items were in their recommendations for CRC screening: the cost of screening tests for patients with no third party coverage and availability of reimbursement by third party payers, including Medicare and Medicaid.
Additional measures included individual physician characteristics, that is specialty, age, sex, race, and medical school affiliation. Practice context measures included size (number of physicians), geographic location (urban, large rural city/town, small/isolated rural town), and the practice’s panel of patients, including percent of uninsured patients. Practice systems support measures included whether CRC screening guidelines had been implemented, whether CRC screening reminders were provided to the physician and/or patients, and type of medical records system used (paper, in transition, or electronic).
We used descriptive statistics to examine reported changes in ordering by test type, reports of agreement with attitudes concerning colonoscopy, and the distribution of factors potentially associated with physicians’ colonoscopy test ordering. Further, we used chi-square statistics to assess the bivariate associations of these measures with the reported change in colonoscopy ordering. Variables with an association at p
0.10 were retained for multivariate models. Polytomous logistic regression models assessed factors associated with changes in colonoscopy ordering. We estimated two models: one with the three attitude items included separately as covariates and one with a summary attitude measure as a single covariate. We used the statistical program SUDAAN version 9.1 to account for the complex survey design and to incorporate survey weights to obtain national estimates. Analyses were conducted in 2010–2011.