Discussion Questions

• ■
  

  What were the study’s objectives?

  
  
• ■
  

  What was the study’s design?

  
  
• ■
  

  What were the main findings?

  
  
• ■
  

  What were the study’s strengths and limitations?

  
  
• ■
  

  Will this study change your approach when recommending contraceptives?

Introduction

Within the first year of typical use, fewer than 1 woman per hundred becomes pregnant while using intrauterine contraception (IUC); the pregnancy rates are 0.8% with the copper T 380 A and 0.2% with the levonorgestrel-releasing intrauterine system. After the etonogestrel implant (first-year pregnancy rate, 0.05%), IUC is the most effective reversible contraceptive available in the United States. Its success does not depend on the patient’s actions, let alone her race, ethnicity, or socioeconomic status (SES). Such demographic factors are known to affect health care outcomes. Could they also influence physicians’ recommendations for IUC? A new study by Dehlendorf and colleagues explored this possibility.

David L. Eisenberg, MD, MPH and George A. Macones, MD, MSCE, Associate Editor

Study Design

Eisenberg: What were the study’s objectives?

Schaecher: The goal was to determine whether patient race/ethnicity and SES affect the provider’s recommendations for levonorgestrel IUC. Recommendations by health care providers have been found to vary by patient race, ethnicity, and SES in other areas of health care. Because such differences might contribute to disparities in outcome, the researchers investigated the effect of these factors on recommendations for contraception.

Eisenberg: What was the study’s design?

Allsworth: This study used a randomized factorial study design, which is useful for evaluating multiple factors simultaneously. For example, in the classical 2 x 2 study design, a researcher can evaluate the effect of the 2 factors studied as well as the interaction between factors. In the work by Dehlendorf et al, the investigators examined the independent and joint effects of race/ethnicity, SES, and gynecologic history. Race had 3 levels: white, black, or Latina. SES had 2 levels: low or high (upper middle class). Gynecologic history had 2 levels. The first represented women who had a history of vaginal delivery and no history of sexually transmitted infection (STI)—a group considered to have no risk factors. The second represented women who were nulliparous with no history of STI and women who had a history of vaginal delivery and a history of pelvic inflammatory disease; these women have perceived risk factors. All permutations of these 3 factors resulted in 18 possible combinations, and the researchers produced videos depicting a standardized patient characteristic of each. This study design is an efficient way to study these effects and interactions.

Study volunteers were recruited in a randomized manner from 4 regional and national obstetric conferences and randomized in blocks of 18; 1 physician-participant for every permutation. Each member of every 18-subject group then watched a different video. Ultimately, each video was viewed an average of 29 times. Randomization was intended to minimize systematic biases in respondents. This was successful with 1 exception: compared with female providers, male providers assigned to standardized patients with apparent risk factors were more likely to be assigned the black patient and less likely to be assigned the Latina patient.

A final key element of the study design that has already been mentioned is the use of standardized patient videos. This approach allows for strict control of sociodemographic characteristics and clinical reporting. It also reduces confounding by these factors.

Results

Eisenberg: What were the main findings of the study?

Ogutha: The following were the results among “low-risk” women—those with previous nonsurgical vaginal delivery and no history of STI:

• •
  

  When race/ethnicity was examined separately from SES, women of low SES were significantly less likely to have IUC recommended by providers than were women of high SES; black women were significantly more likely than white women to have IUC recommended.

  
  
• •
  

  When results were stratified by patient race/ethnicity and SES, IUC was significantly less likely to be advised for black or white women of low SES compared with black or white women of high SES. No such difference was noted among Latina women.

Analyses performed after adjusting for the provider’s sex, specialty, age, race/ethnicity, IUC use in practice, and frequency of prescribing contraception found that:

• •
  

  White women of low SES were significantly less likely to have IUC recommended than white women of high SES. The difference between the likelihood of an IUC recommendation for black women of low SES and those of high SES was no longer significant.

  
  
• •
  

  Black and Latina women of low SES were more likely to have IUC recommended than were white women of low SES.

  
  
• •
  

  Race did not influence IUC recommendation among women of high SES.

Among women with supposed risk factors; that is, nulliparity or a history of PID:

• •
  

  Black women of low SES were as likely to have a recommendation for IUC as were black women of high SES; the same was true of white women.

  
  
• •
  

  A history of PID or nulliparity eliminated the previously seen effect of race among black and Latina women of low SES.

Analyses carried out after adjusting for the provider’s sex, specialty, age, race/ethnicity, IUC use in practice, and frequency of prescribing contraception did not change the results.

Among providers, patients of low SES were judged to be significantly more likely to have an STI, more likely to have an unintended pregnancy, and less knowledgeable. Black women of high SES were perceived as more intelligent and more likely to follow up with medical care—race factored significantly into providers’ perceptions only when comparing these 2 factors among women in each of the high SES groups.

Eisenberg: What were the strengths and limitations of the study?

Vickery: The authors described 3 limitations of the study: the inability to ensure blinding of subjects to the actual nature of the research, the possibility that they did not capture a representative sample, thus preventing generalization of results, and a lack of nonwhite health care providers. Indeed, all of these are limitations. Recruitment was via convenience samples at professional meetings—1 was a national meeting of obstetrician-gynecologists, and the other was a national meeting of family practitioners. They never described their reasoning for choosing these 2 groups or offer any analysis to compare them. They also used data collected at 2 regional meetings of the American Congress of Obstetricians and Gynecologists but did not identify the regions. Given the small number of minority subjects, this information seems relevant.

Perhaps more important, it seems that it would be easy to anticipate the difficulty of recruiting minority physicians from a convenience sample at a professional meeting. I was surprised that the researchers did not describe further efforts to recruit nonwhite subjects, particularly since they mentioned that their numbers were so small, they could not analyze subject-patient representation pairs by congruency. This suggests they intended to do so. Their design lacks a thoughtful recruitment plan.

The researchers explained the steps they took to standardize the patient representations by prompting the actors to modify tone and inflection, and we can see from the offered photos that they were all dressed identically, depending on which part they were playing. However, this description was inadequate. How is a patient representation standardized? Surely, there must be research from the behavioral sciences that clearly shows what does and does not assist an actor in “controlling” all variables of presentation (facial expression, gestures, voice intonation, and subtle body language) aside from the variable of interest. I would like to have seen a behavioral scientist involved in the project, but if this was not possible, at least a thorough review of that literature. Since they made conclusions about the subconscious prejudices the participants might have revealed, a more thorough examination of what subconscious cues the actors might have been giving seems necessary.

This study is important. The question of personal prejudices that physicians might bring to the dialogue with patients is timely and relevant to many disciplines. The authors are correct in stating that the nature of the counseling relationship and negotiations between patients’ voiced preferences and providers’ practice habits are indeed vulnerable to distortion by prejudices.

The main findings are surprising, at least to me, and they speak to the utility of the multifactorial design. For example, researchers found that among women with no risk factors, race/ethnicity seemed to influence physician recommendations among women of low SES but not among those of high SES. They also found that among nulliparous women and those with a history of PID, race/ethnicity appears to have no effect; physicians’ decisions are driven by the presence of apparent risk factors—although the so-called risk factors used in this study are no longer viewed as absolute barriers to IUC use. Nonetheless, these results might not have been demonstrated with a different design, such as a binary approach.

Eisenberg: I agree. They might have missed 1 of their important findings. The strength was that between 42-86% of their respondents felt that IUC was appropriate for the standardized patient. These figures are much higher than the known published rates of IUC usage among women in the United States. Data from the National Survey of Family Growth indicate that from 2006-2008, 5.5% of women aged 15-44 years who were using contraception were using an intrauterine device (IUD). So this cohort of family practitioners and obstetrician-gynecologists are recommending IUC to a minimum of 42% of the 18 standardized patients. That reassures me that the physicians involved in the study are likely to be recommending IUC more often than has been reported for recent years.

Vickery: I think it may represent selection bias in terms of who is at these conferences, who is getting up-to-date information about contraception, and who is more likely to be open to provision of IUC in a nontraditional patient

Eisenberg: I agree, the potential for selection bias—not knowing how this convenience sample of physicians relates to the greater population of providers—makes it really hard to generalize the findings. However, some people might argue that physicians who attend national conferences may be the experts in the field; the thought leaders. If, for example, you’re getting an estimate of 40% of respondents saying that IUC puts patients at risk, then that’s the best estimate possible.

Schaecher: Just in terms of the rate at which IUC was recommended—it was so much higher than the baseline; what we know the rate of IUC use is. Did the investigators provide any information or hint that they were interested in IUC use?

Eisenberg: The authors noted that the subjects did not know the actual intent of the study. The participating providers were not aware of the primary study hypothesis regarding the effects of patient race, ethnicity, and SES on their recommendations, so there was no discussion of that and that is 1 of the weaknesses of this study. You’re measuring providers’ recommendations for IUC based on their perceptions of SES and race/ethnicity, but they needed to ask the individual provider for his or her beliefs about IUC. For example, what’s your take on it as a contraceptive method in general? And so, it’s hard to draw definitive conclusions without having that context. It would have been great if either before or after watching the videos, the providers had been asked whether they were providing IUC in their practices and what percentage of patients with no perceived risk factors receive a recommendation for IUC during contraceptive counseling.

If the providers in this cohort were recommending IUC to patients with no evident risk factors 100% of the time, then you could say race and ethnicity and SES of these standardized patients may have had an impact on their recommendations if they chose not to recommend IUC in certain circumstances. But if you’re talking to a provider who never provides IUC or someone who only recommends it to 10% of the patient population, then withholding a recommendation based on these standardized patient videos may not be as consequential as it otherwise would be. Not having that context is difficult.

Ogutha : I think they probably do have that data because they adjusted for whether someone used an IUC in the practice and the frequency of using it. They don’t report that information but it is 1 of the variables they controlled in their multivariable analysis. Hopefully, the data they are reporting was obtained after controlling for the differences there. But, I think that would still be good information to have. Is there a difference in people who recommend IUC all the time or insert IUDs frequently vs those who don’t? We are missing that piece.

Allsworth : Excellent point. I think they hoped randomization would balance out the groups so the subgroups of physicians that are more or less likely to provide IUC would be balanced across the groups. But you only know what was measured and asked of the clinicians and not some of the questions you have raised here today.

Eisenberg : In Table 1, there was another element that is of concern to me. It pointed out that 74% of the providers involved do perform IUC insertion, and 96% of the clinicians were MDs or DOs. What I thought was interesting—and Dr Vickery brought this up in terms of the incongruence between the providers’ race/ethnicity and the standardized patient race/ethnicity—was that 77% of the participants were white. That clearly may affect the issue at hand, and it complicates our interpretation of this information and how it might affect our practice.