Introduction
Previous studies have shown that patients with cleft lip and/or palate may be stigmatized in society. The objective of this study was to use an implicit association test to evaluate the subconscious biases of non–health care providers and orthodontists against patients with a repaired cleft lip (CL).
Methods
Respondents participated in an implicit association test. Pictures of patients with CL and controls were shown to participants, along with terms representing positive and negative attributes. Participants were prompted to match pictures to the attributes. The software algorithm detected whether the participants were more likely to associate CL with positive or negative terms than controls. Demographic information was collected to measure the association between some sociodemographic factors and implicit biases.
Results
Of 130 valid participants, 52 were orthodontists and 78 were non–health care providers. The entire sample displayed a significant implicit bias against CL ( P <0.001). Overall, orthodontists tended to exhibit slightly higher levels of implicit biases against CL than non–health care providers, but the difference was not significant when controlling for sociodemographic factors ( P = 0.34). Females showed significantly lower implicit biases against CL than males ( P = 0.046). Spearman correlations showed that older people and those who reported a more conservative political affiliation tended to show slightly higher levels of implicit biases against CL ( P <0.007).
Conclusions
Orthodontists and non–health care providers showed moderate but significant levels of implicit biases against patients with clefts. Males, older age groups, and patients with a more conservative political affiliation tended to exhibit slightly higher levels of biases than females, younger people, and those with a more liberal political affiliation.
Highlights
- •
Implicit association tests may help identify subconscious biases toward stigmatized groups.
- •
Orthodontists and lay-people tend to show moderate levels of implicit bias against patients with a cleft lip.
- •
Levels of bias may vary on the basis of the age, gender, and political affiliation.
Cleft lip and/or palate (CL/P) is the most common craniofacial condition, with an incidence rate of about 1 in 700 births in the United States. Although the majority of the affected patients undergo multiple surgical procedures to close the CL and optimize the esthetics, most of them still endure a visible facial difference throughout their lives.
The association between appearance and social stigma is established in social sciences research. Seeman and Goffman defined stigma as “The phenomenon whereby an patient with an attribute which is deeply discredited by his/her society is rejected as a result of the attribute.” Previous studies have shown that patients with CL/P may be stigmatized in society. Eye-tracking studies have found that the lip and nose regions among patients with CL draw more attention than unaffected controls. Multiple studies have shown this cohort to face various forms of stigmas, being considered outcasts, referred to with animal terms, and being rated significantly less attractive and less friendly than their nonaffected peers. Patients with CL/P are less likely to be involved in conversations, and they reported encountering daily social challenges, including stares, comments, and questions about their facial differences. , Similarly, affected patients are more likely to be bullied or teased because of their facial differences. These stigmas may extend to the level of discrimination and may negatively affect their employment prospects. Negative social interactions among this population were found to be associated with lower self-esteem, decreased satisfaction with appearance, and an elevated risk of depression.
Although many studies have explored explicit biases against patients with CL, as mentioned above, our understanding of the implicit (subconscious) biases toward this cohort remain limited. Understanding implicit biases is an important component in understanding overall stereotyping behaviors, as a person may have explicit equalitarian views and yet display subconscious prejudices that may affect how one behaves toward a subgroup of patients.
The implicit association test (IAT) was developed by Greenwald et al in 1998 to provide a tool that researchers may use to explore hidden biases. Because its development, the IAT has become a widely used instrument for assessing implicit biases. IATs measure the speed of response of participants to evaluate the relative strength with which they associate specific target groups (stimuli) with positive or negative attributes. For example, when the IAT is used to measure racial biases, people typically respond more quickly if positive attributes share the same response key with white racial stimuli and negative attributes share the same key with black racial stimuli than vice versa. ,
More recently, researchers have shown interest in evaluating implicit biases among health care providers. , Multiple studies have reported the presence of implicit biases among health care providers, but the evidence is conflicting regarding whether the provider’s implicit bias impacts the treatment decisions. , , However, many studies have concluded that physician’s implicit biases may be associated with poorer interpersonal interactions, anticipating lower compliance from the patient, and a less positive patient’s perception of the encounters with health care providers.
The current study aimed to use an IAT to evaluate if non–health care providers and orthodontists may hold subconscious biases against patients with a surgically repaired cleft lip (CL) and to assess the association between these biases and specific sociodemographic factors.
Material and methods
The protocol of this study was approved by the institutional review board (# STUDY20181087). Recruited participants were stratified into 2 groups: orthodontists and non–health care providers. For this study, everyone who was not an orthodontist or orthodontic resident was considered a non–health care provider. Exclusion criteria: patients with a vision disorder that would inhibit them from seeing details, patients not fluent in the English language, participants aged under 21 years, and patients with a CL and/or palate. We excluded patients with CL/P as those patients only constitute a small percentage of the population, and to avoid introducing a confounding factor, because previous studies have shown patients from a specific minority group to favor their in-group. Furthermore, dental students, general dentists, and physicians were excluded from the study because they can neither be classified as orthodontists or non–health care providers.
Enrollment was voluntary, and participants were recruited from an opportunity sample. Recruitment was conducted over 8 weeks (from April 2019 to June 2019). Non–health care providers were recruited in person, by e-mail, and through the daily University newsletter. Orthodontists were recruited in person at the university’s orthodontic department, and directors of all orthodontic residency programs in the United States were contacted via e-mail and asked to share the link to the study with their residents and colleagues. Moreover, orthodontists in a 32-km radius surrounding the institution were identified from the American Association of Orthodontist’s orthodontist locator Web site, which yielded 38 records, and were subsequently contacted by phone and/or e-mail and asked to participate. All participants were provided a link to the Web site where the test was hosted, and they were asked to take it in a private, quiet place. All data collected were anonymous. A consent form was provided on the Web site before starting the test.
Demographic information such as sex, highest level of education, age, race, political affiliation, education level, and whether the participant was an orthodontist or orthodontic resident were collected. Age was categorized into 5 categories as follows: 21-25, 26-35, 36-45, 46-55, and >55 years. Political affiliation was collected on a scale from 1 to 7 (1 = strongly liberal; 7 = strongly conservative). Furthermore, the sociodemographic survey inquired about the highest level of education and whether the participant was a general dentist, dental student, dental resident, orthodontist, orthodontic resident, physician, medical resident, or none of the above .
The IAT test was designed by Inquisit 5 Web Version (Millisecond, Seattle, Wash). , All the Data were collected on the Web, hosted at the Inquisit Millisecond servers.
A total of 16 nonsmiling frontal photographs of consenting patients from the university’s orthodontic department were used: 8 patients with a surgically repaired CL (5 males, 3 females) and 8 controls (CON), matched for age, gender, and ethnicity. CL photographs were labeled as Target A, whereas CON photographs were labeled as Target B for the purposes of data analysis. Eight positive adjectives (attribute A) and 8 negative adjectives (attribute B) were used to assess whether participants were more likely to associate CL with positive or negative attributes and vice versa. The positive adjectives were good , confident , achiever , successful , healthy , friendly , intelligent , and sociable , whereas the negative adjectives were bad , failure , misfit , terrible , miserable , unsociable , isolated , and undesirable . The choice of the adjectives was based on efforts to include a simple diverse set of attributes that encompass perceived sociability, health, success, and intelligence.
In the current study, we used the 7 Block IAT, as described by Greenwald et al, and designed it using the Picture IAT hosted by Inquisit (Millisecond). The IAT design included 7 blocks in total. Blocks 1, 2, and 5 were used as training blocks where data were not collected. Training blocks are used for calibration and to get the participant acquainted with the test. Blocks 3,4, 6, and 7 are the actual test blocks for which data were collected. When the test begins, the participants will be asked to press either E on the left side of the keyboard or I on the right side of the keyboard as quickly as possible after a random target or attribute is displayed in the middle of the screen. Table I describes the 7 blocks used in the current study. Ultimately, if the attribute is strongly associated with the target, the time taken to respond to the correct answer ( E or I ) will be faster, and if it is not associated with the target, the expected time to respond would be slower. The time it takes to respond is recorded in milliseconds, and the software algorithm generates a final D-score, which ranges from −2 to 2. In the current study, a negative D-score would indicate a preference for CON over CL (higher association of CL to negative attributes and CON to positive attributes), whereas a positive D-score would indicate the opposite. More specifically, −0.1 to −0.34 is considered slight bias, −0.35 to −0.63 is moderate bias, and less than −0.64 is strong bias.
| Block ∗ | No. of trials | Function | Item assigned to left key (E) response | Item assigned to right key (I) response |
|---|---|---|---|---|
| 1 | 20 | Practice | CL | No CL |
| 2 | 20 | Practice | Positive attributes | Negative attributes |
| 3 | 20 | Test | Positive and CL | Negative and no CL |
| 4 | 40 | Test | Positive and CL | Negative and no CL |
| 5 | 20 | Practice | No CL | CL |
| 6 | 40 | Test | Positive and no CL | Negative and CL |
| 7 | 40 | Test | Positive and no CL | Negative and CL |
∗ For all the blocks, the attributes and targets alternate on the screen, and the attributes and targets were randomly shown to participants.
Statistical analysis
Statistical analysis was conducted using SPSS statistical analysis software (version 25; IBM, Armonk, NY). First, we ran the Shapiro-Wilk test to determine the normality of the distribution of the D-scores, age, and political affiliation. Because the D-scores were not normally distributed ( P = 0.039), we used the following nonparametric tests: a 1-sample Wilcoxon sign rank test was used to compare the D-scores of the entire sample to a hypothesized control, in which a hypothesized value of zero (neutral responses or no biases) was used as a control. Mann-Whitney U test was performed for the intergroup comparison to compare the mean D-scores of orthodontists and non–health care providers. Because the demographics of the orthodontist and non–health care provider groups were different, we subsequently ran an analysis of covariance (ANCOVA) to compare the scores of orthodontists and non–health care providers again while controlling for the age, gender, and political affiliation, in an attempt to identify whether the differences would be influenced by the group’s sociodemographic factors.
To assess if certain sociodemographic factors among the entire sample were associated with the levels of implicit bias exhibited, the following tests were conducted: Mann-Whitney U test to compare the mean D-scores of males and females and Spearman’s correlation coefficient to assess the association between D-scores and age group, and D-scores and political affiliation. Spearman’s correlation coefficient was used to measure the correlations because age and political affiliation did not show normal distributions ( P <0.01), and they were recorded on an ordinal scale.
Results
A total of 364 participants logged into our Web link. However, only 144 completed the test. There were 130 valid responses once participants who are general dentists (n = 10) and physicians (n = 4) were excluded. Of the 130 valid participants, 52 were orthodontists and 78 were non–health care providers. The entire sample included 69 females, 60 males, and 1 preferred not to answer. Demographics of the participants are displayed in Table II .
| Demographics |
|---|
| Gender |
| Males (n = 60) |
| Females (n = 69) |
| Prefer not to answer (n = 1) |
| Age group |
| 21-26 (n = 35) |
| 26-35 (n = 52) |
| 36-45 (n = 12) |
| 46-55 (n = 10) |
| >55 (n = 21) |
| Highest level of education |
| High school graduate (n = 2) |
| College student (n = 13) |
| Associate’s degree (n = 4) |
| Bachelor’s degree (n = 22) |
| Master’s degree (n = 30) |
| Other advanced degree (n = 9) |
| Graduate school (n = 14) |
| DMD (n = 32) |
| PhD (n = 4) |
| Race |
| White (n = 91) |
| African American (n = 11) |
| Asian (n = 11) |
| Native American (n = 1) |
| Other (n = 16) |
| Political affiliation |
| Strongly liberal (n = 12) |
| Moderately liberal (n = 39) |
| Slightly liberal (n = 23) |
| Middle (n = 34) |
| Slightly conservative (n = 14) |
| Moderately conservative (n = 6) |
| Strongly conservative (n = 2) |
| Orthodontist status |
| Laypeople (n = 78) |
| Orthodontist or orthodontic resident (n = 52) |
Overall, responses were faster on the IAT when CL was paired with negative attributes and CON was paired with positive attributes, compared with the reversed pairings (mean D-score, −0.39). Approximately 80% of participants showed this pattern.
One-sample Wilcoxon signed rank test showed that the entire sample (n = 130) displayed a moderate, yet statistically significant, implicit bias against patients with a CL (mean D-score = −0.39 ± 0.47; median D-score, −0.44, P <0.001) ( Table III ).
| Group | Mean (SD) | Median | 95% CI | P value | |
|---|---|---|---|---|---|
| Lower | Upper | ||||
| Entire sample ∗ | −0.39 (0.47) | −0.44 | −0.47 | −0.31 | <0.0001 † |
| H 0 | 0 | 0 | |||
Stay updated, free dental videos. Join our Telegram channel
VIDEdental - Online dental courses
