The study of health care disparities is complicated by the difficulty of controlling for all factors which are related both to the predictors of race/ethnicity and SES and to the relevant outcomes. The use of videos of standardized patients provides one of the best methods for minimizing confounding and varying only the sociodemographic factors of interest. Although this method has been used in several published studies,4 5 25–27
none of these prior studies applied this method to reproductive health decision making or investigated the interaction between patient race/ethnicity and SES.
Our stratified multivariate results indicate that providers are more likely to recommend IUC to Black and Latina women than to White women, but only when these women are of low SES. They are less likely to recommend IUC to a low SES woman than to a high SES woman, but only when the woman is White. Moreover, these patterns only seem to hold among women with no perceived risk factors for IUC.
The results of this study do not lend themselves to easy interpretation. While the contrary and interacting ways in which race/ethnicity and SES influence provider recommendations is one source of complexity, this pattern is itself of interest, as it underscores the importance of considering race/ethnicity and SES, both alone and in combination, in the study of health care disparities. In addition, the preference-sensitive nature of contraceptive decision-making does not allow for conclusions about whether one group is receiving better care. Regardless of the etiology or the interpretation of these differences, however, the finding of variation in recommendations to individuals who differ on no clinically relevant variables, but only on their race/ethnicity and SES, deserves attention. While contraceptive recommendations will and should vary by patient, and IUC may be the most appropriate method for many patients, it is important to consider whether patient’s race/ethnicity and SES should be a factor in these recommendations.
Possible explanations for the variation in recommendations for IUC by sociodemographic group include the presence of conscious or subconscious biases regarding the use of this method in certain groups. Alternatively, these results could indicate that providers are making assumptions about the appropriateness of IUC based on the patient’s SES or race not due to bias, but rather as a result of an overly broad application of probabilistic reasoning.28
This phenomenon - termed “statistical discrimination” – occurs when epidemiologic evidence or clinical experience is used to guide treatment decisions for patients within specific sociodemographic groups without the appropriate use of individualizing information.28
With respect to SES, one possible explanation for our findings is that providers perceived low SES patients to be at higher risk of an STI than high SES patients. As prior studies have documented that providers believe that providing IUCs to women at risk of STIs can lead to complications such as infertility,17 18
the lower likelihood of recommending this method to low SES Whites compared to high SES Whites could be related to this perception. As there is evidence that in fact there is an epidemiologic association between SES and risk for STIs,29
this can be interpreted as an example of statistical discrimination, in that knowledge of this association appears to be influencing providers’ decision-making despite the lack of any differences in the histories of the low SES and high SES White patient. The finding that there was no significant differences in recommendations between low and high SES Latina and Black patients could indicate that providers consider the perceived risk of infertility differently in White than in minority patients.
The findings that low SES Black and Latina women were more likely to have IUC recommended could be interpreted as indicating that providers have a bias towards use of this method in minority populations. Alternatively, this could result from the use of race-and ethnicity-based assumptions about the acceptability or appropriateness of IUC which are unrelated to bias. Regardless of the underlying causes of these differences, the increased odds of recommending IUC to Black and Latina women is of concern given the historical relationship of efforts to promote contraception with attempts to limit the fertility of minority and poor women in the United States.30 31
Providers recommending highly effective pregnancy prevention methods in a differential manner by race/ethnicity could be perceived negatively by communities and individuals aware of these issues.32
These findings are consistent with one previous study of physician behaviors, in which physicians provided with clinical vignettes were more willing to sterilize Black women than White women.13
In addition, several studies of patient experiences of contraceptive counseling have found that Black and Latina women more frequently report being encouraged to limit their family size and use contraceptive methods than Whites.11 12 33
The finding that being of higher SES eliminated the effect of race/ethnicity on recommendations for the levonorgestrel IUC may indicate that being of high SES equalizes perceptions across racial/ethnic categories. Similarly, the attenuation of differences in recommendations by both race/ethnicity and SES by the presence of gynecologic characteristics historically considered to be risk factors related to use of IUC may indicate that these risk factors are of greater importance to providers than their differing perceptions of patients by sociodemographic characteristics.
Limitations of our study include the difficulty of ensuring blinding of study participants to our interest in measuring disparities, given the previous publication of reports of studies with similar methodologies. While we had no indication from participants that this awareness existed, we anticipate that, if present, it would result in a conservative bias. The use of a convenience sample obtained at meetings of professional medical societies could affect the generalizability of our results. There is also the possibility that providers would be less likely to recommend IUC to low SES women because of concern about insurance coverage. We consider this unlikely due to the information provided to the participants that the patient had insurance which would pay for all the methods. An additional limitation is our inability to explore the effect of racial/ethnic concordance between providers and patients on recommendations for IUC due to the limited number of non-White health care providers in our sample. Finally, the use of an experimental design with videos of standardized patients may not accurately capture the recommendations of providers in their actual clinical practice.
Our results suggest that providers of contraception, policy makers and advocates should be vigilant and aware of the potential effect of patient sociodemographic characteristics on contraceptive recommendations. On one hand, our study suggests that there may be barriers to providing IUC to low SES White women which should be specifically addressed by these efforts. From another perspective, our results draw attention to the need for historical and cultural sensitivity in the promotion of this highly effective contraceptive method among minorities. Future research could address whether and how much provider recommendations contribute to disparities in contraception use, as well as determine what types of interventions may alleviate differences in recommendations. These studies should build on research in other fields which suggest that enhancing provider awareness of the presence and effect of stereotyping and promoting an emphasis on patient-centered care may be of use in decreasing health care disparities.9 34