Health care providers' attitudes toward people who are obese are consistent with those of the general public. Indeed, research indicates that many physicians perceive individuals who are obese to be highly responsible for causing their condition^{2, 17, 18} and endorse stereotypes that they are lazy and lacking in self control^{2, 19}. A corollary to these beliefs is providers' tendency to perceive obese patients as being unmotivated and non-compliant²⁰. These attitudes and beliefs can shape the clinical interaction making it more negative and uncomfortable for patients who are obese^{21, 22} and decreasing patient-provider rapport. Negative attitudes can also be subtly conveyed in interpersonal behaviors^{23, 24}. The degree to which a provider makes eye contact with a patient, for example, can reflect the level of his or her regard for that patient. These nonverbal behaviors color the tone of the medical encounter and influence patient satisfaction²⁵.

Though provider attitudes and beliefs about patients with obesity can clearly be impactful in the clinical encounter, previous studies have largely been limited to surveys of general attitudes. Each interaction between a provider and a patient occurs in a complex clinical context with a multitude of clinical and interpersonal variations. Thus, there are numerous characteristics of the patient and aspects of the interaction aside from patient weight that could initiate or exacerbate negative attitudes. Many of these variables, however, can be confounded with a patient's weight, making it difficult to pinpoint the source of negative attitudes reported in surveys. For example, patients who are obese tend to have different co-morbidities or different health status than patients who are not obese²⁶. Patients who are obese may also exhibit different interaction styles,²⁷ developed through previous experiences with weight bias and negative treatment. Similarly, providers' expectations about patients who are obese may elicit negative interaction styles from these patients (i.e., self-fulfilling prophecies)²⁸. General surveys do not account for these factors. Thus, there is little evidence on the extent to which bias against and negativity toward obese patients is activated by the patient's weight, as opposed to factors that arise during the medical encounter.

In addition to affecting rapport with patients, providers' attitudes toward obese patients might also influence their decisions about patient care. Indeed, individuals' biased attitudes are often linked to related behaviors²⁹. Only two known studies have attempted to examine the link between provider attitudes and patient care behavior. Thus far, findings are mixed. Wigton and McGaghie found no differences in providers' psychiatric-related recommendations in response to viewing video tapes of actors playing obese or non-obese patients³⁰. Hebl and Xu reported finding differences in providers' test ordering for obese versus non-obese patients based on hypothetical chart review³¹. If provider biases about obese patients do lead to differences in clinical decision making, this could clearly have important implications for patient health. Gaining a clear understanding of the nature of weight bias in the clinical encounter is an important first step toward development of strategies to address it.

It is extremely difficult to examine the effect of a single clinical variable, like patient weight, in a real medical setting. Actual clinician-patient encounters are un-standardized and cannot be manipulated for experimental purposes. Even standardized patients or actors are unable to completely standardize their verbal and nonverbal behavior. There is therefore a need for a highly controlled examination of the effects of patient weight within the context of a medical encounter.

To accomplish this, we employed immersive virtual environment (IVE) technology (i.e., virtual reality) which allows experimental manipulation of clinical variables, while maintaining total patient standardization in a realistic clinical context^{32, 33}. In IVEs, users are immersed in a digitally created environment. This is realized using a combination of graphics software and a carefully designed user interface to create three-dimensional environments that users can navigate in a natural way (e.g., by walking). This technology also allows tracking of users' visual gaze, which can serve as a subtle indicator of bias³⁴.

To examine the effect of patient weight on provider attitudes, beliefs, and behavior, we immersed clinical-level medical students (in their third or fourth years of training) in an IVE interaction with a virtual female patient who was visibly either obese or non-obese. The patient presented with two potentially weight-related complaints, knee pain and shortness of breath, and one non-weight-related complaint, eczema. We chose to focus on a medical student population because they are a likely target of intervention or education efforts aimed at improving clinical interaction with obese patients.

We had three main hypotheses:

The attitude and belief outcomes reported here are important in their own right and can have substantial impact on patients' experiences of the medical encounter. Furthermore, Epstein and Street^{41, 42} posit a model through which negative interpersonal interactions in the clinic can contribute to obese individuals' avoidance of preventive care, as demonstrated in the literature^13-15. In the model, factors like clinician-patient rapport and patient satisfaction are considered ‘proximal outcomes’ of the clinical interaction. These proximal outcomes can influence health directly and also influence ‘intermediate outcomes’ such as a patient's commitment to treatment, which impacts health. The current study demonstrated one concrete example of negative interpersonal behavior in that students exhibited decreased visual contact with the obese patient. Visual contact is a proxy for eye contact, a behavior that is linked to empathy expression by providers. Eye contact is part of a constellation of nonverbal behaviors that reflect providers' regard for patients and that are linked to patient satisfaction²⁵.

We anticipated that differences in beliefs and attitudes would affect students' clinical recommendations for the obese patient. Although there were some individual clinical recommendations that differed between the obese and non-obese versions of the patient in the current study, we did not find pervasive patterns indicative of bias. Though we found no clear evidence of bias, there were some differences in rates of individual recommendations that are of interest. The differences we found between conditions (e.g., for lifestyle-related treatment recommendations) were not inappropriate when the patient's body weight was used as clinical data in the decision-making process. It is worth noting, however, that some of the individual recommendations that differed between the obese and non-obese patient, weight loss and lipid profile in particular, though medically reasonable, may be less germane to short-term workup or symptom relief. Other recommendation differences (minimizing knee stress, heat and cold application, and bronchodilator prescription) might also indicate weight-related assumptions about factors causing the patient's symptoms. Such recommendations, therefore, may be indicative of an understanding that the obese patient is less healthy, less active, or should lose weight generally. Further research should explore beliefs and assumptions behind these subtle differences in recommendation patterns.

We did not find evidence to support our hypothesis that students would rely on lifestyle recommendations to address the obese patient's potentially weight-related symptoms and would thus be less thorough with respect to symptom management and diagnostics. There are several possible reasons for this. In general, diagnosis, treatment, and management of conditions like knee pain and shortness of breath involve protocols^{43, 44} that these students are likely to have recently learned. Behaviors that largely depend upon protocol and habit are less likely to be influenced by one's attitudes and beliefs⁴⁵. Thus, it may be the case that the particular symptoms examined here did not leave students leeway to express their attitudes. Previous research has similarly shown little influence of patient characteristics on management plans for shortness of breath whereas these characteristics influenced recommendations for a different symptom⁴⁶. Alternatively, some students in our sample may not yet have been well-versed in these protocols and thus may have made more recommendations across the board so as not miss anything. More research is warranted to disentangle these issues. Regardless, in this study, the attitudes and beliefs students reported toward obese patients seem to hold more implications for the quality of the clinical encounter than for care recommendations.

In examining these findings it is also of note that participant BMI was not a significant predictor in our analyses. Thus, students who had higher BMIs did not exhibit lower levels of bias. This finding extends the currently mixed literature⁶ on whether providers' weight impacts their attitudes and/or beliefs about patients who are obese.

The current study has several limitations. This study focused on medical students. Although we chose this sample because students are a clear target for potential interventions, they are still in the midst of their clinical training. Thus, the current findings may be less generalizable to practicing physicians and other clinicians. Furthermore, although we were able to measure and report students' nonverbal visual contact during the clinical interaction, we did not include other measures of interpersonal behavior or interaction quality. This is in part because the communication between the students and the virtual patient was constrained to keep it constant between participants and between conditions. In the future, examining additional verbal and nonverbal behaviors could shed more light on how attitudes and beliefs impact interpersonal behavior during a medical encounter. In addition, several of our measures consisted of a single item. More in depth assessments may increase validity of belief and attitude measurement in future studies.

Another limitation was the fact that we did not include assessments of whether and the extent to which participants perceived the patient as being obese. The virtual patient did appear to have a somewhat smaller body type than would be typical of someone with a BMI of 39.9. Based on the fact that the vast majority of students recommended weight loss for the patient, however, it clear that the patient was generally perceived as being overweight or obese. Finally, we did not allow for a true physical examination or interview during the clinical interaction. We provided students with several pieces of clinical information (e.g., blood pressure, smoking status) and included a visual examination period in which students could take a closer look at the patient. However, students were not able to perform any other type of examination that might have informed their diagnostic and treatment recommendations. For this initial study we opted to keep the interaction simple. In the future, however, making the flow of the interaction more similar to reality may increase external validity.

Further exploration of the patient-provider relationship and how obesity stigma plays out in this complex interaction is warranted. Experimental work focusing on the patients' experiences in the interaction will aid in understanding processes at work when they report negative encounters with providers. In turn, explication of these processes will help to identify points of intervention where we might improve patient-provider interactions for patients who are obese.