COLORISM UNVEILED: EXAMINING HOW SKIN COLOR DISCRIMINATION IS ASSOCIATED WITH ACADEMIC ACHIEVEMENT, MENTAL HEALTH, AND SUBSTANCE USE AMONG LATINX ADOLESCENTS

We examined associations between colorism and developmental outcomes among Latinx adolescents. Colorism was operationalized as adolescents’ experiences with discrimination based on skin color. Developmental outcomes included academic achievement, mental health, and substance use. Participants were Latinx adolescents (N = 703; Mage = 15.93, SDage = 1.21; 47.56% female). Anonymous cross-sectional survey data were analyzed. Results indicated that colorism was negatively associated with academic achievement and self-esteem, and positively associated with depressive symptoms and substance use. Results showed that colorism is an important correlate to academic achievement and several indicators of health and well-being for Latinx adolescents, who are a marginalized population

With a focus on Latin America, Keith and Monroe (2016) described how colorism evolved when European colonizers in Latin America and the United States privileged lighter-skinned individuals over darker-skinned individuals within racial/ethnic minoritized groups.These scholars argued that colorism leads to skin color stratification and results in a structural inequality based on skin color that favors individuals with lighter skin.Other scholars have discussed how colorism is a system that has resulted in disparities in education, income, and housing in the United States (Hunter, 2007), as well as educational and income inequality in Latin America (Telles, 2014).Separately, Charles (2021) highlighted how skin-lightening products are being used as a tool for social mobility, thereby perpetuating colorism on a global scale.In sum, colorism has been examined by scholars drawing from psychology, sociology, and educational fields.In this paper, we have drawn from these multiple perspectives and defined colorism as adolescents' interpersonal experience of discrimination based on skin color.

Research with Adults
Research on colorism has focused primarily on adults.In a national study in the United States, about half of Latinx adults reported that skin color shaped their experiences in daily life and that colorism was a major problem (Noe-Bustamante et al., 2021).Studies have shown that adults with darker skin color reported higher levels of colorism than those with lighter skin color (Araujo-Dawson, 2015;Hargrove, 2019;Monk, 2015;Noe-Bustamante et al., 2021;Perreira & Telles, 2014;Santana, 2018;Uzogara, 2019).For example, a study with individuals from Bolivia, Colombia, Ecuador, Guatemala, Mexico, and Peru showed that those with darker skin tones reported higher levels of colorism than their counterparts (Canache et al., 2014).
Colorism has been associated with racial/ethnic identity and economic position.A study with Cuban, Dominican, and Puerto Rican adults in the United States showed that individuals with lighter skin were less likely to identify as Latinx than those with darker skin (Araujo-Dawson, 2015).Colorism has also been associated with income among immigrants.For example, Hersch (2018) showed how colorism was related to wage differences among a national sample of predominantly Latinx immigrants in the United States.In contrast, Roth et al. (2022) found that among a national sample of adults in Mexico, those with more education were more likely to identify as mestizo, a mixed-race person of Indigenous and Spanish descent, than White, regardless of skin color.
Colorism is pervasive in Latin America, as anti-blackness and anti-indigenous beliefs permeate Latin American countries.Telles (2014) conducted a seminal study about ethnoracial identity, skin color, and inequality in Latin America's four largest nations: Brazil, Colombia, Mexico, and Peru.The findings revealed that skin color, rather than ethno-racial identity was the primary social identity through which discrimination occurred in Latin America.Specifically, the results showed that respondents with lighter skin color had higher socioeconomic status than those with darker skin.However, those who self-identified as mestizo had higher socioeconomic status than those who identified as White.Overall, the study highlighted how individuals with darker skin, including those with African and Indigenous ancestry, experienced socioeconomic disadvantages.
Other research has included African Americans and has demonstrated associations between colorism and developmental outcomes.For example, a qualitative study of African American women reported that colorism impacted how they viewed the world and themselves, such as being color-conscious (Wilder & Cain, 2011).The study revealed that African American women experienced colorist messages as children from their maternal family members (mothers, aunties, grandmothers).Such messages included that dark skin was unattractive compared to lighter skin.They were also advised to refrain from engaging in activities that would darken their skin, such as being in the sun.In support of these findings, Breland-Noble and Hall (2013) conducted a review on colorism, which they defined as internalized skin-tone bias.The review showed that darker skin was associated with higher symptoms of depression, lower self-esteem, lower body satisfaction, higher incarceration rates, lower access to prenatal care, higher risk for smoking behavior during pregnancy, and lower childbirth weight of African and Latinx adolescent girls.However, researchers also found that individuals with darker skin reported more pride in their ethnic identity than those with lighter skin.

Academic Outcomes
Although there has been considerable research showing how academic achievement is associated with skin color among adolescents (Hannon et al., 2013;Hunter, 2002;Perreira et al., 2019;Ryabov, 2016;Thompson & McDonald, 2016), less attention has been given to the associations between experiences with discrimination based on skin color and academic achievement.One study with Add Health data that included African American, Asian American, Latinx, Multiracial, and Native American adolescents demonstrated that those with darker skin had lower grade point averages than those with lighter skin (Thompson & McDonald, 2016).Similarly, a retrospective analysis of the Add Health data with African American, Asian, and Latinx young adults indicated that participants with darker skin were less likely to graduate high school than those with lighter skin (Perreira et al., 2019).Another study reported that Asian Americans and Latinx individuals with darker skin obtained less educational attainment than those with lighter skin (Ryabov, 2016).A national study of African American students aged 12-16 indicated that students with darker skin had a greater chance of school suspension than those with lighter skin (Hannon et al., 2013).A study with Mexican American women in the United States indicated that those with lighter skin attained more schooling than those with darker skin (Hunter, 2002).Findings from Latin America (Bolivia, Brazil, Colombia, Dominican Republic, Ecuador, Guatemala, Mexico, and Peru) showed that adults with darker skin completed fewer years of schooling than those with lighter skin (Telles et al., 2015).Lastly, a study with Mexican children indicated that those with darker skin had lower math scores than those with lighter skin (Hailu, 2018).
Research has also illustrated how skin color is associated with schooling experiences.Crutchfield et al. (2022) conducted a scoping review of quantitative and qualitative studies on colorism experienced by students of color (African Americans, Native Americans, Asians, and Latinx) in primary and secondary schools in the United States.The authors concluded that youth with lighter skin had a higher social standing (e.g., "White-looking" privilege) than their counterparts.In one study, researchers used in-depth interviews with six Mexican adolescents to determine how skin color was related to their social identities in a predominantly White institution (Fergus, 2017).Similarly, findings indicated that Mexican students with lighter skin reported that their teachers and peers perceived them as White or Italian and treated them better than their counterparts who had a darker skin color.These researchers also discovered that the lighter-skinned Mexican students were aware that they could pass for White and thus avoid some of the discrimination that other people in their ethnic group faced.Lastly, a study of Latinx adolescents indicated that satisfaction with skin color was positively associated with the importance of doing well in school (Kiang et al., 2020).

Mental Health
Prior studies have shown how differences in skin color have been linked to mental health outcomes (Almanzar, 2020;Calzada et al., 2019;De Casanova, 2004;Kiang et al., 2020).However, there is limited research on the relationship between adolescents' direct experiences with colorism and mental health.For example, a study of Latinx adolescents in the United States indicated that skin color satisfaction was positively associated with self-esteem and inversely associated with depressive symptoms (Kiang et al., 2020).A mixed-methods study examined how skin color was related to one's self-esteem among adolescent females in Ecuador (De Casanova, 2004).The findings demonstrated that lighter skin color was reported to be more beautiful than darker skin.In addition, it was found that those with darker skin had lower self-esteem than those with lighter skin.
Skin color has also been associated with mental health among Latinx children.A longitudinal study examined how skin color was related to Dominican-origin and Mexicanorigin children's mental health (Calzada et al., 2019).Results indicated that mothers of children with darker skin reported more mental health issues that their children experienced compared to mothers of children with lighter skin.However, contrary to deficit beliefs about people with dark skin, a recent study with adults found that Latinx participants with darker skin color endorsed higher rates of skin color satisfaction than those with lighter skin color (Almanzar, 2020).Prior research has also shown that holding a positive view of oneself was associated with darker skin among African American, Asian American, Latinx American, and White/ European American adolescents (Fegley et al., 2008).
Skin color differences in mental health outcomes have been well documented among African Americans.A longitudinal study of African American adolescent girls investigated how skin color was associated with self-esteem from middle school to high school (Adams et al., 2020).Results indicated that adolescent girls with lighter skin reported higher self-esteem than those with darker skin.Among African American adults, skin color was negatively associated with self-reported health (Hargrove, 2019) and positively associated with psychiatric disorders (Oh et al., 2021) and psychological distress (Hamler et al., 2022).A study found that skin color was negatively related to African American women's mental health (Keith et al., 2010).An analysis of Add Health data with African American, Asian, and Latinx young adults indicated that those with darker skin reported more perceived stress and depressive symptoms than those with lighter skin (Perreira et al., 2019).

Substance Use
There is some limited evidence that skin color is associated with substance use among Latinx adolescents.Researchers examined skin color and substance use (alcohol, cigarettes, and marijuana) among Mexican adolescents (Ayers et al., 2012).Findings indicated that the odds of using alcohol, cigarettes, and marijuana were higher for adolescents who were more indigenous in appearance and identified as third-generation, whereas the opposite pattern was shown for adolescents who were first-and second-generation.Indigenous appearance was measured by asking adolescents to indicate their skin tone from very light to very dark and to report how much their physical features, including eye color, hair color and texture, and the shape of their nose, were like the majority White/European racial/ethnic group.The prevalence of substance use among Latinx groups (Villalobos & Bridges, 2018) warrants investigation about colorism and substance use.

Present Study
To address the current knowledge gaps in the field, we addressed the following research question and hypotheses: Among Latinx adolescents, how is colorism associated with developmental outcomes, including academic outcomes (letter grades), mental health (depressive symptoms and self-esteem), and substance use (alcohol, tobacco, cannabis, and illicit substances)?We hypothesized that colorism would be inversely associated with academic outcomes, given past studies (Kiang et al., 2020;Thompson & McDonald, 2016).Second, we hypothesized that colorism would be positively associated with depressive symptoms and negatively associated with self-esteem because of prior research (Hamler et al., 2022;Kiang et al., 2020;Oh et al., 2021).Third, we hypothesized that colorism would be positively associated with substance use, given research in this area (Ayers et al., 2012).
Socioeconomic status (SES) was assessed via adolescent-reported maternal education.Response options ranged from 1 (no high school diploma) to 6 (doctorate).The sample average was 1.94 (SD = 1.21), indicating an average maternal education between no high school diploma and a high school diploma/GED.The High School 1 sample average was 1.76 (SD = 1.06), indicating between no high school diploma and a high school diploma/GED degree.The High School 2 sample average was 3.19 (SD = 1.49), indicating between an Associate's degree and a Bachelor's degree.Relevantly, adolescents self-reported their skin color, "How would you describe your skin color?"Response options ranged from 1 (very light) to 5 (very dark).The sample average skin color was 2.53 (SD = 0.69).

Colorism
The Colorism Scale (Centeno et al., 2023) comprised two subscales: (1) Colorism-School (five items; α = .84)and (2) Colorism-Community (eight items; α = .90).The prompt included the question, "How often have you experienced each of the following because of your skin color?"Sample items from the Colorism-School subscale were, "You were discouraged from joining an advanced level class" and "You were wrongly disciplined or given after-school detention."Sample items from the Colorism-Community subscale were, "You were hassled by police" and "You received poor service at a restaurant or store."Response options ranged from 1 (rarely) to 5 (often).Scores were generated by calculating non-missing averages for each subscale.Higher scores indicated more colorism.See (Centeno et al., 2023) for scale development information, including information about the Exploratory Factor Analysis and Confirmatory Factor Analysis.

Academic Achievement
Academic achievement was measured with one item: "Generally, what grades do you get in your classes?"Response options ranged from 1 (A's) to 5 (F's).Responses were reverse coded so higher values indicated higher grades.Adolescents have been shown to reliably report academic achievement (Crockett et al., 1987).

Mental Health
Depressive Symptoms.Depressive symptoms were measured with the Center for Epidemiological Studies Depression Scale (CESD; Radloff, 1977).The measure included 20 items.A four-point Likert scale ranging from 1 (rarely or none of the time/less than 1 day) to 4 (most or all of the time/5-7 days) was used.Four items were reverse-coded.Scores were generated by calculating a non-missing average.Higher scores indicated more depressive symptoms.The measure yielded high internal consistency in the current study (α = .92).The CESD has shown high internal consistency with a Cronbach's alpha of .86 in a recent study with Mexican-descent adolescents (Piña-Watson et al., 2019).
Self-Esteem.Self-esteem was measured using the Rosenberg Self-Esteem Scale (Rosenberg, 1965).Participants were asked to indicate how much they agreed with each statement on a four-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree).The measure included 10 items.Five items were reverse-coded.Scores were generated by calculating a non-missing average.Higher scores indicated higher selfesteem.The measure has yielded good internal consistency in the current study (α = .84).The Rosenberg Self-Esteem Scale has shown strong internal consistency in a prior study including Mexican American adolescents with a Cronbach's alpha of .85(Arbona & Power, 2003).

Substance Use
We assessed the use of alcohol, tobacco, cannabis, and other illicit substances.The response options were yes (1) and no (0).Composite scores for each substance were created by summing yes responses.For alcohol, a single item asked: "Have you ever had a whole drink (more than a sip or a taste) of an alcoholic beverage?"For tobacco, 22 items assessed the use of tobacco products, such as cigarettes, vape devices, and non/water pipes.For cannabis, a single item asked: "Have you ever used marijuana, weed, hash, blunt, or other cannabis?"For other substances, two items assessed the use of opioids, stimulants, and hallucinogens.

Racism
Subtle.We measured subtle racism using the Ethnic Microaggressions Scale (Huynh, 2012) which contained 13 items and across three subscales.Sample items were "Someone tells you that racism does not exist anymore," and "Someone tells you that no one discriminates against your ethnic/racial group."Response options ranged from 0 (0 times/never) to 5 (almost every day/all the time; α = .93).
Overt.We measured overt racism using the Adult and Peer Discrimination Measure (Way, 1997) that compromises of 24 items and 4 subscales.A sample item for peers was, "How often have you been insulted by teenagers because of your race or ethnicity?"A sample item for adults was, "How often are you treated unfairly by teachers/staff because of your race or ethnicity?"Response options were 1 (never) to 5 (all of the time; α = .97).

Covariates
To address alternative hypotheses, we included age, gender, SES, immigration status, and skin color as covariates (see Participant section for details and Table 1).

PROCEDURE
Data were collected from two high schools in the western region of the United States.Recruitment speeches were given in classrooms by trained researchers.Study materials were distributed to interested students that included an assent form, parental consent form, and the study survey.Surveys were completed on the students' own time and returned to the research team at the high school.Anonymity was preserved by providing separate envelopes for the surveys and consent forms.Compensation was $20.The affiliated institutional review board of the university approved the procedure (H20-010).Data were analyzed using Stata (Version 17).

Colorism Prevalence and Association with Demographic Characteristics
Table 1 shows the associations between colorism and demographic characteristics.On average, adolescents reported experiencing colorism at school and in the community between rarely and sometimes.Analyses were conducted to determine if Colorism-School and Colorism-Community subscales were associated with demographic characteristics, including gender (man/woman), immigration status, age, SES, skin color, as well as racial/ethnic discrimination.T-tests indicated that Colorism-School did not differ by gender (p = .94)or immigration status (p = .31).Similarly, t-tests indicated that Colorism-Community scores did not differ by gender (p = .53)or immigration status (p = .46).Pearson correlations indicated that Colorism-School was not associated with age (p = .11),SES (p = .77),or skin color (p = .22).Similarly, Colorism-Community was not associated with age (p = .09),SES (p = .30),or skin color (p = .53).Further, t-tests indicated that Colorism-School (p = .42)and Colorism-Community (p = .15)did not differ across the two high schools.Colorism-School and Colorism-Community were not associated with demographic characteristics (values not shown).However, Pearson correlations indicated Colorism-School and Colorism-Community were positively associated with both overt and subtle racism.
Correlations shown in Table 1 indicated that Colorism-School and Colorism-Community subscales were strongly associated with mental health and substance use.Specifically, Colorism-School and Colorism-Community were positively associated with depressive symptoms, tobacco use, and illicit substance use, and negatively associated with self-esteem.However, Colorism-School and Colorism-Community were not associated with academic achievement.T-tests indicated that adolescents who reported more Colorism-School (t(667) = 2.80, p = .005,d = .25)and  Colorism-Community (t(667) = 4.01, p < .001,d = .36)were also more likely to use alcohol.Whereas adolescents with higher scores on Colorism-Community (t(667) = 3.31, p = .001,d = .31)were more likely to use cannabis, although Colorism-School was not associated (p = .17).

Developmental Outcomes
Analyses were conducted to determine the associations between colorism and developmental outcomes (see Tables 2 and 3).Multiple linear regression analyses indicated that Colorism-School and Colorism-Community subscales were negatively associated with academic achievement and self-esteem, and positively associated with depressive symptoms after controlling for covariates (see Tables 2 and 3 for Colorism-School and Colorism-Community analyses, respectively).For substance use, multiple linear regression analyses indicated that Colorism-School and Colorism-Community subscales were positively associated with tobacco and illicit drug use after controlling for covariates.Logistic regression showed that Colorism-School and Colorism-Community subscales were positively associated with alcohol use.Although Colorism-Community was positively associated with cannabis use, Colorism-School was not associated.Semipartial correlations for multiple linear regression models indicated small to medium effect sizes and odds ratios for logistic regression models indicated small effect sizes (Cohen, 1988).

DISCUSSION
We investigated how colorism was associated with developmental outcomes among Latinx adolescents.Colorism was defined as the interpersonal experience of discrimination based on skin color.We sought to extend research focused on skin color differences by investigating adolescents' direct experiences with colorism.Examining colorism among Latinx adolescents is vital in order to provide insight into the processes underlying academic and health disparities among this racial/ethnic group.Therefore, the primary aim of our study was to conduct research on the experiences of colorism among Latinx adolescents and to examine the association between colorism and key indicators of development, including academic achievement, mental health, and substance use.Findings extend research that has focused on skin color differences by showing how adolescents' reports of interpersonal experiences with colorism are also associated with developmental outcomes.First, colorism was negatively associated with academic achievement.These results extend prior research on skin color differences in academic outcomes (Crutchfield et al., 2022;Fergus, 2017;Hailu, 2018;Hannon et al., 2013;Hunter, 2002;Kiang et al., 2020;Perreira et al., 2019;Telles et al., 2015;Thompson & McDonald, 2016) by illustrating how adolescents' direct experiences are connected to their academic achievement.Second, colorism was positively associated with depressive symptoms and negatively associated with self-esteem.These results build on prior studies about skin color differences in mental health outcomes (Adams et al., 2020;De Casanova, 2004;Fegley et al., 2008;Hamler et al., 2022;Hargrove, 2019;Kiang et al., 2020;Oh et al., 2021;Perreira et al., 2019) and show how colorism is related to multiple indicators of psychological well-being.Lastly, we demonstrated that colorism was positively associated with using several substances, including alcohol, tobacco, cannabis, and illicit drug use.These results add to the limited extant literature indicating that skin color differences are associated with substance use (Ayers et al., 2012) by demonstrating how experiences with colorism are associated with several substances.Overall, findings demonstrated that colorism was associated with key developmental outcomes among Latinx adolescents.
Results indicated that colorism among Latinx adolescents included multiple individuals (peers, school personnel, and adults in the community) and settings (schools and communities).Thus, Latinx adolescents reported experiencing colorism in the classroom when trying to sign up for a club or at the local store when trying to make a purchase.Findings indicated that the pattern of associations between colorism and developmental outcomes was similar across schools and communities.Our study lends credence to the Phenomenological Variant of Ecological Systems Theory (Spencer, 1995), which contends that it is crucial to understand adolescents' perspectives across social contexts, such as in schools and communities.Results also support ecological systems theory (Charles, 2021) which highlights how colorism may vary across contexts such that individuals with darker skin have a different lived experience than those with lighter skin.
Research on colorism is essential because there is a need to eliminate biases that perpetuate a pigmentocracy, a hierarchy based on skin color (Chavez-Dueñas et al., 2013;Dixon & Telles, 2017;Hannon et al., 2013;Hunter, 2002Hunter, , 2007;;Telles et al., 2015).Colorism has its origins in colonialism, in which light skin was associated with White Europeans and was perceived to be more desirable and superior to the darker skin of Black and Indigenous people (Chavez-Dueñas et al., 2013;Hunter, 2002;Keith & Monroe, 2016).Our results with Latinx adolescents illustrate these theoretical perspectives and demonstrated that skin color hierarchies and preferences for lighter skin are still occurring today.Although our study focused on the interpersonal experience of colorism, the existence of this form of discrimination has roots in systems of oppression (Chavez-Dueñas et al., 2013;Hunter, 2002;Keith & Monroe, 2016).
Our study has implications for how researchers conceptualize colorism and its impact on the well-being of minoritized adolescents.Unfortunately, findings from our study show how colorism continues to insidiously infiltrate communities and the lives of adolescents, underscoring the importance of identifying ways to reduce colorism in order to promote the health and well-being of Latinx adolescents.Results may help researchers better understand the harmful impact of colorism on the education, mental well-being, and risk-taking behavior of Latinx adolescents.Implications from the current study include the creation of programs that teach adolescents, school personnel, and community members about the existence of colorism and its association with key developmental outcomes.Hence, anti-colorism educational programs are needed in order to eliminate skin color bias.

Limitations and Future Directions
Our study had several limitations.First, a longitudinal design was not employed.This would be ideal for determining the mechanisms that underlie the relationships between colorism and developmental outcomes.Second, alcohol, cannabis, and illicit substances were measured with just a few items compared to tobacco.Additional research is needed with other substances measured in greater detail.Third, the current study employed a scale that was adapted from Fisher et al. (2000) scale on racial discrimination.It is possible that the scale may not have captured how colorism is experienced among Latinx groups.Future studies should create a scale to assess colorism specifically.Qualitative methods might be especially useful for this endeavor.
A fruitful area of future research includes drawing from intersectional perspectives and considering how colorism intersects with other identities.Such a line of inquiry could include adolescents with a greater range of skin colors given that the sample average skin color was relatively low in this study.Collecting information about generational status would also be useful.Future studies should also investigate colorism among mixed-race adolescents.Mixedrace identification for Latinx in the United States is connected to Latin America's racist and colonial history of whitening the race.For example, four in ten of Latinx respondents identifying as mestizo, mulatto, or mixed race also reported that their race is White (Gonzalez-Barrera, 2015).Indeed, prior research has found that many Latinx individuals identify as mestizo, or mixed race, and tend to have a preference for lighter skin color, a phenomenon known as blanqueamiento, meaning "whitening the race" (Charles, 2021).Examining how mixed-racial and ethnic groups experience colorism would be important for our understanding of this topic.

CONCLUSION
The current study examined the relationship between Latinx adolescents' experiences with colorism and their developmental outcomes.This research provided a unique examination of the interpersonal experiences of colorism across school and community contexts.We found evidence linking colorism to a broad array of key indicators of health and wellbeing, including academic achievement, psychological well-being, and substance use.The study broadened the perspective by emphasizing the importance of addressing colorism in both community and school environments among adolescents.The profound negative outcomes spotlight colorism's insidious ties to the legacy of colonial-era pigmentocracy, reinforcing the pressing need to challenge and dismantle such biases.Our findings serve as a call to action for the development of tailored educational programs designed to eradicate skin color bias.These efforts are particularly significant for Latinx adolescents, who represent a growing demographic in the United States.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the author(s).

FUNDING
Our work was supported by funding awarded to Zena R. Mello, PhD, from The Regents of the University of California, Tobacco-Related Disease Research Program [Grant T31IP1855].The data that support the findings of this study are available on request from the author, ZRM.The data are not publicly available due to the privacy of research participants.
). Immigration status was assessed with one open-

TABLE 1
Correlations and Descriptive Statistics for Colorism-School, Colorism-Community, Academic Achievement, Mental Health, and Substance Use Among Analyses for each developmental outcome were conducted separately.Values for linear regression are shown in the upper portion and values for logistic regression are in the bottom portion.All regression models controlled for age, gender, maternal education, immigration status (values not shown), and skin color.CI = confidence interval.
a The 95% CI is calculated around the OR.*p < .05. **p < .01. ***p < .001.RESEARCH IN HUMAN DEVELOPMENT Analyses for each developmental outcome were conducted separately.Values for linear regression are shown in the upper portion and values for logistic regression are in the bottom portion.All regression models controlled for age, gender, maternal education, immigration status (values not shown), and skin color.CI = confidence interval.aThe 95% CI is calculated around the OR.*p < .05. **p < .01. ***p < .001.