Ip, Edward H. PhD; Marshall, Sarah; Vitolins, Mara RD, DrPH; Crandall, Sonia J. PhD, MS; Davis, Stephen MA; Miller, David MD, MS; Kronner, Donna; Vaden, Karen; Spangler, John MD, MS
The prevalence of obesity in the United States has increased dramatically over the past few decades. Since 2010, the prevalence of obesity has exceeded 20% of the population in all states, and 12 states (concentrated in the South) have obesity rates in excess of 30%.1 Given the negative implications of obesity for an individual’s health, including a predisposition to heart disease, diabetes, some cancers, and other conditions,2 physicians must continue to work collaboratively with patients who are obese toward the goal of healthy weight loss.
Regrettably, researchers have shown that physicians are not exempt from the biased views regarding people with obesity —a disorder that is so prevalent in the United States.3–5 Such views undermine the patient–doctor relationship and contribute to obesity stigma, which, rather than motivating patients to lose weight, contributes to a variety of physical and psychological problems.6,7 Research has demonstrated that even those doctors who specialize in the treatment of obesity hold negative attitudes toward people who are obese, stereotyping such patients as lazy on an implicit association test.8 Hebl and Xu9 found that physicians held more negative views of people who are obese compared with their views of people of normal weight on 12 of 13 indices, including the desire to help the patient and estimated duration of appointments. Huizinga and colleagues10 found that increasing body mass index was significantly associated with a lack of respect for patients who are overweight.
Such negative inclinations also exist among medical students. Although Persky and Eccleston11 and Wigton and McGaghie12 did not find differences in students’ clinical decision making, they did note biases in beliefs, attitudes, and interactions on the basis of a patient’s size alone. They noted that such biases could cause patients to delay both follow-up and preventive services. In turn, many people who are obese report difficulty in pursuing medical care for a variety of reasons, such as lack of appropriate-sized equipment and embarrassment about their weight.13–15
Medical students constitute an enticing target for efforts to reduce obesity bias and obesity stigma in health care. Even though research has shown that such bias is well established and affects patient care, many medical schools have not addressed the problem directly.16,17 In response to this need, our research team has recently developed and implemented both a series of educational modules and a program that involves standardized patients who are obese—which, together, are intended to challenge students’ attitudes and beliefs about obesity. We hypothesize that a more favorable disposition toward patients who are obese could undergird improved counseling and rapport-building skills, perhaps leading to improved clinical outcomes.
To determine the efficacy of these and similar interventions,17–19 a valid and reliable means of assessing medical students’ beliefs and attitudes toward obese patients is needed. In this report, we both describe the development of such an instrument, the Nutrition, Exercise, and Weight Management (NEW) Attitudes Scale, and provide evidence of its validity and reliability.
We developed and tested the NEW Attitudes Scale over many months (August 2008 through March 2011) and many stages (each described in detail below). We conducted a literature search to seek prior scales, and, finding none to fit our needs exactly, we established a team of experts who developed potential items for a novel scale. We sought feedback regarding the items from third-year medical students. Next, we solicited the input of medical education experts to judge and weight the items in a Thurstone scaling study. We tested the external validity of the weighted-item scale with second- and fourth-year medical students. We examined its internal consistency with a new group of third-year students who interacted with a standardized patient, and we examined reliability with a new group of fourth-year medical students who completed the scale twice over two weeks without experiencing any intervention. Finally, we determined test dimensionality through a factor analysis.
In August 2008, we conducted a literature search in PubMed and Medline to identify validated measures of obesity bias. From the relevant abstracts and/or articles we uncovered, we identified several scales that assess beliefs and attitudes about obese people, including the Anti-Fat Attitudes Questionnaire (AFA), the Anti-Fat Attitudes Scale (AFAS), the Anti-Fat Attitudes Test, the Attitudes Toward Obese Persons Scale, the Beliefs About Obese Persons Scale (BAOP), and the Fat Phobia Scale.20–24
None of these instruments target medical students as the specific audience or emphasize patient care as a focus. The most relevant instruments are perhaps the AFA, BAOP, and AFAS. However, for the purpose of the current study, these instruments have limitations. As Morrison and O’Connor observed,22 the AFA contains some ambiguous items (e.g., “I worry about becoming fat”), which carry the risk of conflating prejudice with concern for one’s own health. The eight-item BAOP focuses on beliefs about what causes obesity and lacks the breadth of covering attitudes about weight management and counseling, which are important to this study. The five-item AFAS focuses strictly on personal dislike of people who are fat and lacks a specific target population. All of these scales also do not contain questions about attitudes within the context of a physician interacting with patients, rendering them unsuitable for our purposes.
Item development and refinement
After carefully reviewing these known instruments, we decided to design a novel tool, the NEW Attitudes Scale. We carefully built content validity into the NEW Attitudes Scale through a team development approach during the item-generation process (January to April 2009). Content experts who participated in the process included medical educators, nutritionists, physicians, and psychometricians (E.I., M.V., S.C., S.D., D.M., D.K., J.S.). The team extensively reviewed the items to ensure that the items adequately covered the desired domains of the targeted construct. The team generated new items and modified some from the existing instruments to form an initial pool of items. The team divided the items into the domains of nutrition, exercise, and weight management for further discussion and refinement. We combined revised items from each domain for use in focus groups.
After we developed and grouped by domain our initial items, we conducted six focus groups from May 2009 to October 2009. Each group consisted of six to nine third-year medical students completing their family medicine clerkship at Wake Forest School of Medicine (WFSM), the site of our study. The family medicine clerkship coordinator (K.V.) solicited random volunteers by e-mail from the six clerkship rotations (10 students per rotation) to participate in the focus groups. The students completed the scale before the focus group; during the focus group they provided feedback on the items in the pool and considered other items for inclusion in the NEW Attitudes Scale. Students received small monetary compensation and lunch in exchange for their participation. Their participation and comments were independent of assessment and their standing in the program. Each focus group’s facilitator, a rotating study team member, recorded comments and took notes on the discussion; after the focus groups, we analyzed the notes to identify salient themes. On the basis of the results of the focus group discussion and our team review, we created a final list of 31 items.
Thurstone scaling study
To increase the accuracy of the NEW Attitudes Scale, we performed a Thurstone scaling study (August–October 2010) to weight the final list of 31 items. Originally developed for achieving smaller bias and lower error variance in measuring attitudes about religion, the Thurstone scale has since been applied to a variety of subjects.25–27 For example, a group of researchers recently used the Thurstone scale to identify desirable characteristics for police officers.28 Unlike Likert-type scales, which give equal weight to each item (typically in the form of a statement), the Thurstone scale uses differential weights for individual items/statements. Briefly, the Thurstone scaling approach solicits ratings on individual items from a panel of “experts” about the magnitude and direction of an attitudinal item. In this case, we requested experts in the field of medical education to serve as judges. We asked them to determine the positivity or negativity of each item, using a scale from 1 (most unfavorable) to 11 (most favorable). The judges did not directly respond to the items but merely rated them for use by future survey respondents. We felt that accounting for the differential discriminatory power of individual items would improve the psychometric behavior of the NEW Attitudes Scale. For example, a future survey respondent who endorses a very negative statement about obese patients would be viewed as more biased than one who endorses a mildly negative statement, rather than as equally biased.
We recruited our Thurstone judges (predominately faculty at medical schools) from both clinical and research backgrounds using the listservs of the Groups on Educational Affairs (of the Association of American Medical Colleges), which have divided physicians into four regions: Central, Northeastern, Southern, and Western United States. We also identified internal judges within WFSM and invited them to participate. All judges held advanced degrees. We did not offer the judges incentives; their participation was voluntary and anonymous, and their participation and ratings had no effect on their evaluations, promotion, tenure, or standing within the Group on Educational Affairs or WFSM.
We determined the weight for each NEW Attitudes Scale item from the statistic of the distributions of ratings (on a scale of 1–11). We selected a robust statistic—the median of the distribution, which is not as sensitive to outlier values as the mean—to form the raw weight. To get the actual question weightings, we subtracted 6, the midpoint of the scale, from the median ratings. For example, a question that received a median rating of “1” on the Thurstone scale would receive an actual weight of −5 on the survey instrument, and a question rated an “8” would receive a weight of 2. We used the weights in the final instrument as multipliers and a 5-point scale to capture students’ responses to each item: −2 (strongly disagree), −1, 0, 1, or 2 (strongly agree) points. By centering the scale at 0, positive scores would suggest positive attitudes, and vice versa. Thus, if an item had a weight of −5, and a student strongly disagreed (score of −2), the item would be scored as −5 × −2 = 10 points; if a student strongly agreed with that negative item, it would be scored −5 × 2 = −10 points.
Measuring validity, reliability, internal consistency, and dimensionality
After the careful development, selection, refinement, and weighting of the 31 items, we administered the finalized NEW Attitudes Scale to a convenience sample of second- and fourth-year medical students at WFSM whom the project manager (D.K.) recruited to take the survey (these students had not participated in the focus groups). Participation was voluntary, confidential, and independent of assessment or promotion. Students were given a small monetary reimbursement for participation. In August 2010, students completed not only the NEW Attitudes survey but also two previously validated measures of obesity bias: the AFA and the BAOP.20,23 To assess construct validity, we correlated the NEW Attitudes Scale with the other two scales. We expected the correlations to be moderate, demonstrating concurrent validity.
We further assessed construct validity by measuring medical students’ attitudes toward obese patients before and after an intervention intended to reduce stigma and bias. Specifically, third-year medical students completing their family and community medicine clerkship at WFSM in January through March 2011 participated in a simulated outpatient visit with a standardized patient instructor who was obese (these students had neither participated in the focus groups nor completed the NEW Attitudes, AFA, or BAOP instruments). The standardized patient presented to the clinic, complaining of a five-pound weight gain and concerned about her risk of cancer. After the encounter, the standardized patient instructor provided feedback meant to increase the students’ confidence in caring for patients who are obese and highlighting the challenges of weight loss. If the NEW Attitudes Scale is appropriate and accurate and if the intervention works, the scale should detect a positive change in attitudes after the intervention.
We assessed reliability in February through March 2010 by testing a small volunteer group of fourth-year medical students in a different class year and retesting them two weeks later to determine whether their scores changed. These students had not participated in any of the prior stages of the development or testing of the NEW Attitudes Scale, and their participation was voluntary. Each participating student received a small incentive of $10, and their participation and scale results were independent of assessment and standing. The team chose a two-week retesting to avoid the possibility of students’ attitudes actually changing as a result of new experiences, new learning, or some other influence. We assessed internal consistency using Cronbach alpha.
Finally, we conducted a factor analysis of the responses to test dimensionality of the scale. We first categorized the item as either positive (Thurstone weight > 0) or negative (weight ≤ 0) items, and reverse-coded the negative item responses. We used a scree-plot to determine the number of factors and then applied factor analysis, based on varimax rotation, to delineate orthogonal factors.
We accounted for missing values (for all analyses: Thurstone weighting, external validity testing, postintervention testing, and reliability retesting) by one of two ways: discarding surveys in which an excessive amount of information was missing or, if only a few answers were omitted, imputing missing data with a participant’s mean response to the other items. We discarded surveys if a participant did not answer at least 80% of the items.
Statistical analysis and ethical approval
We used SAS version 9.2 (SAS Inc., Cary, North Carolina) to complete all analyses: weighting, correlations, Cronbach alpha for reliability, and the factor analysis. The WFSM institutional review board approved this protocol after expedited review.
Our initial PubMed/Medline literature search resulted in 75 abstracts or articles on attitudes/measures regarding obesity. We examined five of these carefully, adapting some of the items described for our NEW Attitudes Scale.20–24
Development of the NEW Attitudes Scale
We developed an initial item pool of 50 questions in the content domains of nutrition, exercise, and weight management. Focus group feedback helped us refine and select 31 items for inclusion in a revised instrument. Reasons for which students suggested excluding questions included ambiguity and inability to discriminate well between respondents who might endorse different attitudes. For example, we excluded “I am likely to provide care for a significant number of overweight/obese patients in my practice” because students may have answered “yes”—not because of any preference for such people but because they know that these patients make up a large portion of the U.S. population. We updated some questions to reflect a level of training appropriate to medical students. Ultimately, 7 items pertained to nutrition, 6 to exercise, and 11 to weight management. Although another 7 questions did not fall into a specific domain, we deemed these relevant to the scale and decided to retain them.
Thurstone scaling study
A sample of 201 expert judges rated each of 31 questions using a Thurstone scale ranging from 1 (most unfavorable) to 11 (most favorable). Table 1 shows the characteristics of the judges. Missing values were low, constituting only 0.9% of responses.
Actual ratings ranged from 1 to 8. In Table 2, we report the actual items and weights of items (judges’ median score minus 6). Of note, several statements of attitude deemed to be neutral in nature, such as “Patients understand the connection between exercise and cancer,” received a raw median score of 6. Thus, neutral items received a weight of 0 and, subsequently, do not affect scoring on the NEW Attitudes Scale. Although they do not affect the scoring, we deemed them important to include because they were representative of the targeted construct and could be used to inform future studies. Other items received decisively positive or negative ratings; for example, “I do feel a bit disgusted when treating a patient who is obese” (Item 23) received a median raw rating of 1 (weight of −5). Some variations were wide, whereas others were narrow; for example, the item “It is usually sufficient to give a person brief, clear advice about weight management” had a variance of 5.30, whereas Item 23 has a variance of 0.8.
Validity, reliability, and dimensionality
A sample of 111 second-year and fourth-year medical students completed the NEW Attitudes Scale alongside the AFA and BAOP scales; 66 respondents (59%) were female, and 55 (50%) were second-year students. In Table 2 we report the number and percentage of respondents who agreed with (either 1 or 2 on the actual scale), felt neutral toward (0), disagreed with (−1 or −2), or did not provide a response for all items. The percentage of missing values was low at 0.6%. The technique for handling missing values, described in the Method section, should be sufficient to impute values, and we do not expect results to differ with more advanced missing-value methodology. Cronbach alpha for the sample of 111 medical students was 0.63, suggesting moderate internal consistency.
Table 3 shows the mean scores of students, stratified by gender and year, for all three obesity scales, including the NEW Attitudes Scale. The mean score on the NEW Attitudes Scale was 24.4. Out of a possible score of −118 to +118, the lowest score was −37 and the highest score was 76. Men and women respondents did not differ significantly. However, second-year students were more positive than fourth-year students (P = .01).
The Pearson correlations between the NEW Attitudes Scale and the AFA and BAOP were, respectively, −0.47 and 0.23. We expected a negative value for AFA because a higher score on AFA indicates a more negative attitude toward obese people (for both NEW Attitudes Scale and BAOP, higher scores indicate more positive attitudes toward obese people). Thus, the NEW Attitudes Scale was moderately correlated with both AFA and BAOP. Figure 1 shows the distribution of scores on the NEW Attitudes Scale.
We administered the NEW Attitudes Scale to a total of 103 third-year medical students both before a simulated encounter with an obese standardized patient instructor and (as described in the Method section) after receiving feedback in order to determine whether the scale could detect a change in attitudes. The mean score before the encounter was 25.5 and 32.2 on retesting, a difference of 27% (P < .001, df = 102).
The test–retest reliability (correlation coefficient) of the 24 fourth-year medical students tested at baseline and again two weeks later was 0.89.
We identified three factors that respectively account for 15%, 11%, and 8% of variance. The factor analysis revealed the following dimensions: (1) antifat (including, for example, item 23: “I do feel a bit disgusted when treating a patient who is obese”), (2) self-efficacy and propensity to provide counseling to patients who are obese (including, for example, item 12: “I have a personal desire to counsel patients about exercise”), and (3) belief about how others understand obesity (including, for example, item 8: “Patients understand the connection between nutrition and cancer”). Table 2 indicates the factor on which each item is most loaded.
Discussion and Conclusions
Given both the scale of the obesity epidemic and the positive role physicians can play in healthy weight loss, the availability of an instrument with substantial evidence for reliability and validity that can assess medical students’ attitudes toward obese patients is both timely and significant. As medical educators develop and deploy efforts to change medical students’ attitudes (e.g., new curricula for obesity counseling), to be able to accurately gauge the success of such interventions is imperative. The NEW Attitudes Scale represents a scientifically validated tool, with the unique advantage of a clinical focus, for capturing medical students’ attitudes toward patients who are obese.
Some of the strengths of the NEW Attitudes Scale include the steps in its development. First, the use of a Thurstone scale increases the accuracy of the NEW Attitudes Scale by weighting individual questions. Specifically, giving an individual weight to each question reveals a finer level of detail regarding participants’ attitudes. Second, the scale shows strong evidence of face validity, content validity, concurrent validity, and criterion validity, as well as satisfactory internal structure. Third, the scale demonstrates a high level of reliability, and its internal consistency is also satisfactory.
Given the average student score of 24.4 on the NEW Attitudes Scale, WFSM students appear to have, collectively, a mildly positive attitude regarding the care of patients who are obese. We found this finding surprising given the previous studies that identified negative attitudes toward patients who are obese in medical students.11,12 However, the authors of those studies did not report a difference in approach to management. Perhaps the more positive score can be accounted for by the clinical focus of the NEW Attitudes Scale. Students’ desire to provide excellent care may override personal bias. Also of interest is the finding that scores of the men did not differ significantly from those of the women across all three instruments, whereas fourth-year students had more negative attitudes than second-year students on both the NEW Attitudes Scale (0.48 standard deviation difference) and AFA (0.63 standard deviation difference). One might expect that attitudes would improve with increased experience caring for obese patients, but providers may become more skeptical of patients achieving weight loss over time.
The present study does have some limitations. We included only third-year medical students at WFSM in our focus groups and in our pre–post standardized patient instructor study, we included only fourth-year WFSM students in our test–retest reliability study, and we included only second- and fourth-year WFSM students in the survey, limiting generalizability. Also, although the instrument captured an expected improvement in scores after students completed the standardized patient instructor encounter, efficacy of the intervention has not yet been determined. The detected effect may be short-term and not sustained. Finally, we have not been able to collect data for assessing longer-term predictive validity. We expect that attitudinal scores are likely to predict differences in approach to and/or management of patients who are obese in actual practice, but this needs to be investigated in future studies.
In conclusion, the NEW Attitudes Scale demonstrated good validity and reliability and may be used in future studies to measure medical students’ attitudes toward patients who are obese.
Funding/Support: NCI grant number R25CA117887-01A2.
Ethical approval: The Wake Forest School of Medicine institutional review board granted expedited review.
Other disclosures: None.
2. Haslam DW, James WP. Obesity. Lancet. 2005;366:1197–1209
3. Andreyeva T, Puhl RM, Brownell KD. Changes in perceived weight discrimination among Americans, 1995–1996 through 2004–2006. Obesity (Silver Spring). 2008;16:1129–1134
4. Puhl RM, Andreyeva T, Brownell KD. Perceptions of weight discrimination: Prevalence and comparison to race and gender discrimination in America. Int J Obes (Lond). 2008;32:992–1000
5. Puhl RM, Heuer CA. The stigma of obesity: A review and update. Obesity (Silver Spring). 2009;17:941–964
6. Puhl RM, Moss-Racusin CA, Schwartz MB. Internalization of weight bias: Implications for binge eating and emotional well-being. Obesity (Silver Spring). 2007;15:19–23
7. Puhl RM, Heuer CA. Obesity stigma: Important considerations for public health. Am J Public Health. 2010;100:1019–1028
8. Teachman BA, Brownell KD. Implicit anti-fat bias among health professionals: Is anyone immune? Int J Obes Relat Metab Disord. 2001;25:1525–1531
9. Hebl MR, Xu J. Weighing the care: Physicians’ reactions to the size of a patient. Int J Obes Relat Metab Disord. 2001;25:1246–1252
10. Huizinga MM, Cooper LA, Bleich SN, Clark JM, Beach MC. Physician respect for patients with obesity. J Gen Intern Med. 2009;24:1236–1239
11. Persky S, Eccleston CP. Medical student bias and care recommendations for an obese versus non-obese virtual patient. Int J Obes (Lond). 2011;35:728–735
12. Wigton RS, McGaghie WC. The effect of obesity on medical students’ approach to patients with abdominal pain. J Gen Intern Med. 2001;16:262–265
13. Merrill E, Grassley J. Women’s stories of their experiences as overweight patients. J Adv Nurs. 2008;64:139–146
14. Olson CL, Schumaker HD, Yawn BP. Overweight women delay medical care. Arch Fam Med. 1994;3:888–892
15. Amy NK, Aalborg A, Lyons P, Keranen L. Barriers to routine gynecological cancer screening for white and African-American obese women. Int J Obes (Lond). 2006;30:147–155
16. Banasiak M, Murr MM. Medical school curricula do not address obesity as a disease. Obes Surg. 2001;11:677–679
17. Roberts DH, Kane EM, Jones DB, et al. Teaching medical students about obesity: A pilot program to address an unmet need through longitudinal relationships with bariatric surgery patients. Surg Innov. 2011;18:176–183
18. Endevelt R, Shahar DR, Henkin Y. Development and implementation of a nutrition education program for medical students: A new challenge. Educ Health (Abingdon). 2006;19:321–330
19. Wiese HJ, Wilson JF, Jones RA, Neises M. Obesity stigma reduction in medical students. Int J Obes Relat Metab Disord. 1992;16:859–868
20. Crandall CS. Prejudice against fat people: Ideology and self-interest. J Pers Soc Psychol. 1994;66:882–894
21. Lewis RJ, Cash TF, Jacobi L, Bubb-Lewis C. Prejudice toward fat people: The development and validation of the antifat attitudes test. Obes Res. 1997;5:297–307
22. Morrison TG, O’Connor WE. Psychometric properties of a scale measuring negative attitudes toward overweight individuals. J Soc Psychol. 1999;139:436–445
23. Allison DB, Basile VC, Yuker HE. The measurement of attitudes toward and beliefs about obese persons. Int J Eat Dis. 1991;10:599–607
24. Bacon JG, Scheltema KE, Robinson BE. Fat phobia scale revisited: The short form. Int J Obes Relat Metab Disord. 2001;25:252–257
25. Thurstone LL. Attitudes can be measured. Am J Soc. 1928;33:529–554
26. Thurstone LL, Chave EJ The Measurement of Attitude: A Psychophysical Method and Some Experiments With a Scale for Measuring Attitude Toward the Church. 1929 Chicago, Ill University of Chicago Press
27. Krabbe PF. Thurstone scaling as a measurement method to quantify subjective health outcomes. Med Care. 2008;46:357–365
28. Guffey JE, Larson JG, Zimmerman L, Shook B. The development of a Thurstone scale for identifying desirable police officer traits. J Police Crim Psychol. 2007;22:1–9