This paper is among the first to examine the relationship between patient-provider race concordance and obesity care. Despite prior research suggesting greater use of needed medical services among race-concordant pairs (11
), we did not find support for our hypothesis that race concordance would be positively associated with obesity care. With respect to the hypothesis that white and black physicians treat their white and black obese patients differently, we found only one effect that was significant in both the main analysis and in sensitivity analysis: visits by black patients to white doctors were less likely to include exercise counseling as compared to visits by white patients to white doctors. We also found that visits by black obese patients to black doctors were less likely to include weight-reduction counseling than visits by white obese patients to black doctors, although this finding may be more tenuous as it did not reach statistical significance in the sensitivity analysis without the imputed patient race/ethnicity data. These findings suggest that visits by black patients include less exercise counseling than visits by white patients when they see white doctors and less weight-reduction counseling than white patients when they see black doctors. In this sample, patient–physician race concordance does not appear to have a positive impact on weight-related counseling for visits among black obese patients. We also evaluated a model to predict any weight-related counseling and found no differences by concordance status for either black or white patients.
There are several potential reasons why black obese patients might receive less obesity care than white obese patients. Several studies suggest that physicians hold more negative perceptions about likelihood of adherence, engagement in risk behaviors, and social resources available to black patients (37
). The low rates of weight-reduction counseling among black patients seeing black physicians may also reject black physicians' efforts to be culturally sensitive to their black patients. Several studies have shown differences in body image and perceptions of overweight and obesity status between blacks and whites (38
). It could also be the case that physicians may not be aware that blacks, as compared to whites, are disproportionately impacted by obesity (3
). Physicians may lack sensitivity to underlying levels of obesity risk in the adult population. The similarly low rate of weight-related counseling among black and white physicians could additionally reflect inadequate usage of clinical obesity guidelines; in other words physicians may generally not get adequate training or have adequate resources to provide weight-related counseling to their obese patients, regardless of their race. It could also reflect physicians' belief that their advice will have little effect on patient behavior (40
), that patients are not interested or ready for treatment (40
), and that obesity is the responsibility of the patient (45
These results, like previous research (7
), indicate that despite clinical guidelines suggesting that clinicians screen all adult patients for obesity and offer intensive counseling to promote sustained weight loss for obese adults, rates of weight-related counseling are low, regardless of patient race/ethnicity (29
). Although studies fail to document disparities in weight-related counseling, factors that contribute to suboptimal rates of weight-related counseling may differ for black and white patients. We hypothesized that race concordance might be one such factor among black patients, but our study suggests this is not the case. Future work should examine similarities and differences in the contribution of patient, clinician, and health system factors to low levels of weight-related counseling among patients of different ethnic groups.
Our finding of low rates of weight-related counseling may be a reflection of general physician stigma toward obesity that has been documented (34
) as well as lower physician respect toward patients with a higher BMI (47
). Also, consistent with past studies, we found that other factors influence weight-related counseling such as patient's gender, age, and comorbidity risk status as well as characteristics of the clinical encounter (e.g., type of visit, time spent with the doctor) (7
). Like previous studies (10
), our results suggest less weight-related counseling in visits by black patients than in visits by white patients.
Improving rates of weight-related counseling in primary care settings is an important strategy to promote behavior change in obese patients. There is a growing body of evidence suggesting that patients who are told by their physician that they are overweight are more likely to lose weight relative to those who are not told (48
), that patients who are counseled about their weight or weight-related behaviors are more likely to report working on those areas (50
), and that patients who are advised by their physician to modify their behavior are generally more confident and motivated to engage in lifestyle modifications (e.g., dietary changes, increased physical activity) (51
The mechanism through which race concordance has been primarily associated with positive patient–physician interaction is through perceptions of similarity (56
). Patient's preference for health care providers of their same race/ethnicity (59
) may also be in response to well documented unequal treatment in heath care by race/ethnicity (60
). In addition to race/ethnicity, there are several other shared identities between the patient and their physician which may also impact obesity care. Therefore, future research could examine the impact of other shared characteristics, including patient–physician gender concordance on obesity care. Previous research on gender concordance and quality of care has found, for example, that preventive screening rates were higher among women seeing female doctors (61
). Future work could also explore why the patterns we observed between race concordance and weight-related counseling differ by the type of counseling.
There are several limitations to this analysis worth noting. First, that we don't observe an effect in some of the race-concordant groups could be because our sample size, especially for white patients seeing black physicians, is too small. Second, unmeasured patient or physician factors may have affected our findings. Examples include patient familiarity with their physician, patient and physician attitudes toward race/ethnicity and preferred communication styles. Third, the data rely on physician reports (which cannot be objectively verified), and which may lead to an underestimation of weight-related counseling given evidence suggesting that physicians underreport behavioral counseling (62
). However, recent research validating the NAMCS indicates that it is highly specific, suggesting that if physicians report having delivered a service, there is a high likelihood that it was given (63
); even if a physician reporting bias existed, it would be unlikely to differ by race concordance in the patient–physician relationship. Weight-related counseling could also be underestimated if physicians provided weight-related counseling to their obese patients but did not document it in the medical chart as it is typically not reimbursed by insurance companies (64
). However, it is unlikely that underreporting of weight-related counseling would vary by doctor–patient race concordance. Nonetheless, to address this potential limitation, we controlled for patient insurance type in the analysis given that reimbursement for weight-related counseling may differ by type of health insurance. Fourth, patient race ethnicity is based upon the physician's report or abstraction from the medical record by office staff or a federal abstractor. It is unknown to what extent patient race/ethnicity is based on the physician's (or potentially even the office staff's) perception of the patient's race/ethnicity as compared to the physician or office staff having asked the patient about their own perception. A physician's perception of a patient's race/ethnicity may differ from the patient's own perception; however, it is likely that physicians' perceptions of patients' race/ethnicity influence their own communication behaviors and decision-making. Fifth, because the NAMCS is a visit-based survey, obese patients who see their physician more often may have received weight-reduction counseling at a clinical encounter not captured by the survey. Sixth, patients were considered to have received weight-related counseling if it was “ordered or provided” at the clinical encounter. In the case of ordered services, it was not possible to know whether the patient actually received them. It is also not possible to know the quality of counseling services provided or ordered. Seventh, we are unable to determine from the data whether patients had health insurance coverage for weight-reduction counseling. However, we controlled for respondents' health insurance status as a crude proxy for access to obesity-related services. Eighth, we controlled for the length of the clinical visit based on literature which suggested that race concordance is associated with longer clinical encounters (8
); however, weight-related counseling may be associated with longer visit times which could lead to overcontrolling and thus bias the estimates of our outcome variables toward the null. We did run the models excluding length of visit (data not shown) and the estimates did not change significantly. Ninth, we would have ideally included other measures in our risk status variable, such as physical inactivity, but those data were not available in the 2005–2007 NAMCS.
Finally, missing data creates the potential for bias in the analysis. Among the 10,884 visits that met the inclusion criteria for the study or had missing data on inclusion criteria, whether a patient was obese could not be determined for 64% of the visits and physician race and ethnicity was missing for an additional 15% of the visits. In addition, as stated above, patient race/ethnicity was imputed for 23.5% of the sample. To look for bias due to each of these areas of missing data, we conducted the following analyses. First, as described above, we conducted all analyses with and without imputed patient race/ethnicity. Second, to look for bias in cases with missing concordance as a whole, we compared the rate of each type of weight-related counseling in visits with known concordance status to those with missing concordance status. We found no significant relationship between missing concordance status and receipt of any type of weight-related counseling. To assess whether bias might exist in the multivariate models, we repeated our logistic regression models using concordance (missing vs. not missing) as the dependent variable and the weight-related counseling variables as the independent variables of interest. Again, no type of weight-related counseling predicted whether concordance was missing. These analyses were also conducted using only cases with unimputed race, and similar results were found.
In searching for bias among visits with missing BMI, we noted that if observations with missing BMI followed a similar distribution to those with BMI, ~46% of those with missing BMI would be obese. However, we hypothesized that height and body weight might be more likely to be recorded if a patient were obese. We, therefore, examined the percent of each type of weight-related counseling within visits by obese patients, nonobese patients, and for visits among patients with unknown obesity status. Compared to known obese patients, rates of all types of counseling were significantly lower among visits by nonobese patients and among visits by patients with unknown obesity status. Compared to nonobese patients, rates of counseling among visits by patients with unknown obesity status were neither significantly different nor significantly lower. These results suggest that bias from missing BMI is less likely since patients with unknown BMI are more similar to nonobese patients, who were excluded from the analysis. An alternative explanation is that in visits by obese patients where the physician does not measure height and weight, leading to missing BMI in the dataset, the physician may be less likely to provide counseling. If these visits are distributed differentially across patient-doctor race groups, this could lead to bias in the analysis. However, height and weight may be missing from the data set for a variety of reasons, including those related to abstraction of the data in addition to the failure of the physician to obtain height and weight at the visit. We also examined whether missing BMI data differed systematically for our primary dependent variable (doctor–patient race concordance) and found no statistically significant differences.
Although our analyses were unable to uncover bias due to missing data, we recognize that, as in all studies with high rates of missing or imputed data, the potential for bias still exists.
Our data suggest that black obese patients receive less exercise counseling than white obese patients in visits to white physicians and may be less likely than white obese patients to receive weight-reduction counseling in visits to black physicians. Further research is needed to understand how to improve counseling, particularly for black obese patients.