Disparities in health outcomes by race/ethnicity and socioeconomic status (SES) are well documented in many areas. The role of health care providers in contributing to these disparities is an area of growing research, with multiple studies suggesting that providers treat patients differently depending on patients’ race/ethnicity and SES. These findings are consistent with social psychology research indicating that subconscious stereotyping by social categories is widespread even among those who self-identify as nondiscriminatory.

Previous research on the effect of patient race/ethnicity and SES on providers’ clinical behavior has focused on provider-patient interactions around discrete medical decisions for which there is general consensus about appropriate treatment. Contraceptive decision making, in contrast, involves the consideration of multiple clinically appropriate options, with the best treatment being highly dependent on patients’ personal preferences. In addition, the discussion of sexual behavior and contraception use in a clinical encounter is a culturally and socially complex area of medicine in which providers’ subconscious biases or assumptions might play an important role.

The limited data analyzing potential disparities in providers’ decision making in this context suggest that providers may be susceptible to different influences on their recommendations. From one perspective, several studies have suggested that providers may be especially likely to encourage the use of highly effective contraceptive methods and discourage fertility in minority and low-income populations. In contrast, minority and low-income women in the United States have higher rates of unintended pregnancy and lower use of contraceptive methods than do white and higher-income women. Although system and patient-related factors undoubtedly contribute to these statistics, the presence of these disparities also raises the possibility that clinicians may not, in fact, be promoting effective contraceptive methods among patients from these sociodemographic groups.

The effect of patient characteristics on provider recommendations for intrauterine contraception (IUC) is of particular interest due to this method’s high efficacy, as any tendency toward discouraging the fertility of specific populations could be manifested in a greater likelihood of recommending this method. Alternatively, as many providers are concerned that the use of IUC could result in pelvic infections among women at increased risk for sexually transmitted infections, although well-designed clinical studies have indicated that these concerns are misplaced, clinicians who make race- and class-based assumptions about sexual behaviors may be less likely to consider IUC as an appropriate contraceptive method for poor and minority women. As an expansion in the use of IUC is currently being advocated as a means to decrease unintended pregnancy, it is important to understand whether differences in provider recommendations by race/ethnicity and SES exist and, if so, consider how these differences may affect efforts to promote IUC in different demographic groups.

To determine whether patient race/ethnicity and SES affect provider recommendations for the levonorgestrel IUC, we conducted a study of providers’ recommendations using videos of standardized patients of different race/ethnicities and SES.

Materials and Methods

Standardized patient videos

We produced 18 videos portraying a standardized patient requesting advice about contraception, with the patient varying by race/ethnicity (white, black, or Latina), SES (low- or upper-middle class), and gynecologic history (a woman with a history of a vaginal delivery and no history of sexually transmitted infections [STIs]; a woman with a history of a vaginal delivery and history of pelvic inflammatory disease [PID]; or a nulliparous women with no history of sexually transmitted infections). The low SES patient was portrayed as a housekeeper studying for her GED and the high SES patient as a recent business school graduate working as a bank manager. Both the high SES and the low SES patients were portrayed by the same actor within each racial/ethnic category. The providers were told that the patient was 27 years old, had normal blood pressure, and had recently had a negative test for Gonorrhea and Chlamydia and a normal Papanicolaou test. Each patient indicated that she was in a monogamous relationship and that she did not want to become pregnant for at least a few years. For the purpose of these analyses, the primary gynecologic profile of interest was the woman who had previously had a vaginal delivery and had no history of STIs, as women with this history have historically been perceived as ideal IUC candidates. The standardized patients who were nulliparous or had a history of PID were grouped together as having perceived risk factors for complications related to IUC. Photographs of the standardized patients are shown in Figure 1 , A-F.

Standardized patients

A , Low SES white patient. B , High SES white patient. C , Low SES black patient. D , High SES black patient. E , Low SES Latina patient. F , High SES Latina patient.

SES , socioeconomic status.

Dehlendorf. Recommendation for intrauterine contraception. Am J Obstet Gynecol 2010.

In each video, the patient presented her history as a monologue, with the only variation being the study factors. The scripts used in the videos were pretested with a sample of 15 providers to ensure the maximal level of realism. Standardization of verbal factors, such as inflection and tone, were practiced with the 3 actors. Five health care providers watched all 18 videos to verify overall consistency of nonverbal and verbal content.

Study design

We recruited a convenience sample of health care providers (MDs, DOs, Nurse Practitioners, and Physician Assistants) at meetings of professional societies of family medicine and obstetrics and gynecology. Eligibility criteria consisted of being a practicing health care provider in the United States who had completed training. After observing 1 video, selected using randomly permuted blocks of 18, the providers completed a survey about their contraceptive recommendations for the patient shown, ranking each of 6 methods on a scale of −3 to +3, with −3 indicating “Strongly Recommend Against”, 0 indicating “Neither Recommend for nor Against” and +3 indicating “Strongly Recommend For.” The computerized survey randomized the order in which the contraceptive methods were displayed to avoid any sequence effect. The subjects were informed during the survey that the patient’s health care insurance covered all contraceptive methods. Our outcome of interest was the recommendation regarding the levonorgestrel IUC, as this is the more effective of the 2 IUCs offered in the United States. Providers also answered questions about their perceptions of the patient in the video, indicating whether they felt the patient was more or less likely to experience specific outcomes and whether she was more or less intelligent and knowledgeable than an average woman her age. The providers were not aware of the primary study hypothesis regarding the effect of patient race/ethnicity and SES on provider recommendations for IUC.

Our primary research question was whether the recommendations of providers for IUC differ for African American, Latina and White patients. We based our sample size on a binary outcome of willingness to recommend an IUC. We hypothesized that a difference of 15 percentage points in prevalence of this outcome would be clinically significant in populations where overall prevalence of the outcome is 30%. Our sample of 524 provided 84% power to detect a difference of this magnitude in separate comparisons of African American and Latina women to white women.

Statistical analysis

We performed bivariate analysis of the outcome variable of recommendation for the levonorgestrel IUC using χ ² tests, Fisher’s exact tests, and t tests as appropriate. Multivariate logistic regression was performed using a dichotomized response variable of ≥1 or ≤0. Because of the complicated interplay between social factors in other studies, prespecified analyses included analysis of interactions between the patient characteristics of race/ethnicity, SES, and gynecologic history. For the multivariate model, we prespecified the following provider-level variables to be included in the model: age, sex, race/ethnicity, specialty, and provision of IUC. For all other variables, we used backward selection and included any variables that changed any of the coefficients of interest by >10%. The subjects’ perceptions of the patients were analyzed in the same manner. All analyses were performed using Stata Version 9.2 (Stata Corp, College Station, TX).

The Committee of Human Research at the University of California, San Francisco approved this study, and all subjects provided informed consent before participation.