This is a secondary analysis of a study designed to investigate whether and how clinicians’ recommendations for intrauterine contraception differ by patients’ race and ethnicity, socioeconomic status and gynecologic history. Between September 2007 and May 2008, a convenience sample of health care providers was recruited in the exhibit halls of four medical society meetings: two regional and one national meeting of the American College of Obstetricians and Gynecologists (ACOG), and a national meeting of the American Academy of Family Physicians (AAFP). Each participant was randomly assigned to view one of 18 videos of a patient seeking contraceptive advice; in each video, a woman indicated that she was in a monogamous relationship, that she did not want to become pregnant for at least a few years and that her insurance covered all methods. The script, the patient’s appearance and speech patterns, and other verbal and nonverbal factors were standardized across videos, with three exceptions: The videos varied by the patient’s race and ethnicity (white, black or Hispanic), socioeconomic status (low or high socioeconomic status) and gynecologic history (the woman either had had a vaginal delivery and had no history of STDs; had had a vaginal delivery and had pelvic inflammatory disease; or was nulliparous and had no history of STDs). A panel of five health care providers previewed all 18 and confirmed that the only substantial variations were by the three study factors. Results of analyses examining differences in provider recommendations by patient characteristics have been published.8,9
After viewing one of the videos, each participant completed a computerized survey that included the following question: “Assuming that all methods were covered by the patient’s insurance and were provided in your practice, and that the patient had no strong preference, please indicate for each method what your recommendation for this patient would be.” For each of the six most commonly used reversible nonbarrier methods (the pill, injectable, patch, ring, copper IUD and levonorgestrel IUD), the clinician provided a rating ranging from −3 to 3; a rating of −3 indicated “strongly recommend against,” 0 indicated “neither recommend for nor against” and 3 indicated “strongly recommend for.” The order in which the methods were listed was randomly selected for each participant to avoid any sequence effect.
Participants also provided information on their demographic and practice characteristics. These included sex, race and ethnicity (white, black, Hispanic, Asian/Pacific Islander or other), age (35 or younger, 36–45, 46–55, or 56 or older), specialty, professional degree, and board certification (yes or no). Respondents also reported how often they prescribed contraceptives (frequently, occasionally, rarely or never), whether they inserted IUDs (yes or no), and their practice type (academic, private, HMO, or family planning clinic or community health center) and region (Northeast, Midwest, South or West).
To focus our analysis on providers who were most involved in prescribing contraceptives, we limited our sample to clinicians specializing in obstetrics and gynecology or family medicine, as these specialties provide the majority of contraceptive care in the United States.10
For the same reason, we excluded providers who indicated that they rarely or never prescribed contraceptives. In addition, we excluded nurse practitioners and physician assistants; because they constituted only 4% of our sample, our analysis would not have had sufficient statistical power to enable us to draw meaningful conclusions about these providers’ contraceptive recommendations. In total, we excluded 21 nurse practitioners or physician assistants, seven physicians who listed a specialty other than family medicine or obstetrics and gynecology, and 28 family physicians or obstetrician-gynecologists who rarely or never prescribed contraceptives. Our final sample consisted of 468 physicians.
We used chi-square tests to identify overall group differences in associations between physician characteristics and contraceptive recommendations. In these analyses, we classified a participant as recommending a method if he or she gave the method a score of 1 or higher on the −3 to 3 scale.
Next, we used multivariate logistic regression models to assess associations between physician characteristics and recommendation of each method. In these models, we included as covariates the three varying patient characteristics (race and ethnicity, socioeconomic status and gynecologic history), as well as all interactions between patient characteristics with a p value of less than .10. In the multivariate analyses, we considered a physician to have recommended a method if his or her score for that method was higher than the physician’s mean score for all six methods. This approach avoids potential confounding by the association of physician characteristics with a general propensity to recommend contraceptive methods, and focuses the analysis on whether the provider considers a specific method more or less appropriate than others.
To assess whether the use of standardized patients with varying characteristics affects the generalizability of our findings, we performed sensitivity analyses examining the relationship of patient characteristics with overall frequency of recommending each method, as well as interactions between patient and provider characteristics. All analyses were performed using STATA 9.2.
The Committee on Human Research at the University of California, San Francisco, approved this study. All participants provided informed consent using a computerized consent form. They received a food item with a value of approximately $5 for their participation.