We surveyed a national probability sample of physicians in obstetrics and gynecology and family medicine and advance practice nurses. These clinicians provide the majority of primary reproductive health care to women.9
The sampling frame was limited to nurses and physicians working in the United States. Physicians were identified using the American Medical Association’s Physician Masterfile, a comprehensive database of nearly 1 million physicians and residents that includes members and nonmembers of the American Medical Association. The database is updated weekly. Stratified probability samples of 600 eligible physicians (ie, not retired, not in residency, spend majority of time in direct patient care) specializing in family medicine and 600 physicians specializing in obstetrics and gynecology or gynecology alone were drawn. Nurses were identified using a national database of nurse practitioners and advanced practice nurses (Verispan), which contains more than 143,000 advanced practice nurses and is updated monthly. Stratified samples of 600 advanced practice nurses specializing in obstetrics and gynecology, family planning, or women’s health and 600 advanced practice nurses specializing in family medicine were drawn to allow comparisons within and across these clinician subgroups. In total, 1,200 physicians and 1,200 advanced practice nurses were selected with replacement using a random number generator to select participants from each clinician-type strata. Duplicate names in the sample (n=29) were excluded from the mailing.
Clinicians were first sent a letter introducing the study, followed by a survey and cover letter, postagepaid return envelope, and $20 in cash mailed by U.S. Priority Mail. A reminder postcard was mailed 1 week later, and a second copy of the survey and postagepaid return envelope was sent by first class mail 3 weeks later if the original survey had not yet been returned. If mail was returned as undeliverable, research staff used online state nursing and medical boards, directories, and search engines to locate current information. After the final mailing, recruitment efforts were made by phone. Data were collected from September 2008 through February 2009. Of the 2,371 clinicians surveyed, 350 were ineligible: 123 nonrespondents were identified as ineligible in the process of telephone follow-up, and 227 respondents were not eligible based on screening questions in the survey (ie, do not provide direct patient care or do not see family-planning or STI patients). The response rate was calculated by subtracting ineligible clinicians and adjusting for an estimated proportion of eligible participants among unknown respondents (). There were 1,196 eligible respondents. The analytic sample for the study (N= 1,196) excludes 32 clinicians with missing data on the dependent variable. The study protocol was approved by the Committee on Human Research at the University of California, San Francisco.
Fig. 1 Mailed and returned surveys. We assumed that the proportion of providers eligible was the same for providers of known and unknown eligibility and calculated the response rate accordingly. The American Association for Public Opinion Research. Standard (more ...)
An original questionnaire with closed-ended questions was developed for the survey, informed by 31 qualitative interviews with providers conducted during an earlier phase of the study12,13
and using items validated in other studies.
The dependent variable was based on a single question asking clinicians, “Do you require a pelvic examination when prescribing oral contraceptives to your patients?” with the response options “never,” “sometimes,” “usually,” and “always.” A dichotomous variable was constructed to compare clinicians who always require the pelvic examination to those with other practices.
To define practice setting, clinicians were asked where they work the majority of the time. To assess the patient population served, clinicians were asked to indicate approximately how many of their patients are Medicaid recipients, uninsured, Hispanic or Latino, African American, and teenage females. The response categories were “none,” “some,” “half,”“most,” or “all,” and these were dichotomized to compare those with half or more to those with less than half of patients in the category. In addition, a variable describing patient populations as half or more minority (Hispanic or Latino, African American) was created by combining responses from the two individual items.
Population density was based on an item from the survey asking whether the location of the respondent’s clinical practice was a city, a suburban community or large town, a small town, or a rural area. The item was coded “urban,” “suburban,” or “rural,” where rural included those in a small town. Region was defined using standard U.S. census bureau classifications for states. Clinician characteristics measured included age, sex, race or ethnicity, and whether they had received training in family planning. Finally, we examined two patient scenarios that were presented to clinicians, a 16-year-old and a 24-year-old, to ascertain the clinical tests they would provide in office visits.
Stratification design were applied in all analyses to account for disproportionate sampling. The characteristics of the clinicians in our sample and their practice settings were described and differences tested with the design-based Pearson χ2 test for overall categorical differences and the adjusted Wald test for mean differences. A logistic regression model testing the independent effect of clinical specialty on the likelihood of requiring a pelvic examination, adjusted for clinician and practice setting characteristics then was computed. Finally, we evaluated how the clinical screening and prevention activities recommended by clinicians differed depending on their pelvic examination requirements. For the hypothetical patient cases, we examined how preventive screening practices related to cervical cancer and STI prevention co-vary with pelvic examination practices, using the design-based Pearson χ2 test. Stata 11.1 was used for all analyses.