Race, Social Class, and Child Abuse: Content and Strength of Race, Social Class, and Child Abuse: Content and Strength of Medical Professionals’ Stereotypes Medical Professionals’ Stereotypes

Black and poor children are overrepresented at every stage of the child welfare system, from suspicion of abuse to substantiation. Focusing on stereotypes as a source of bias that leads to these disparities, the current study examines the content and strength of stereotypes relating race and social class to child abuse as viewed by medical professionals. Doctors, nurses, and other medical professionals (Study 1: N = 53; Study 2: N = 58) were recruited in local hospitals and online through snowball sampling. Study 1 identified stereotype content by asking participants to list words associated with the stereotype that either (a) Black or (b) poor children are more likely to be abused by their parents, and responses were organized into construct groups. Study 2 determined stereotype strength by asking participants to rate how strongly the constructs generated in Study 1 related to either the race-abuse or social class-abuse stereotype. The content of stereotypes linking child abuse to Black or poor children are confounded, with approximately half the constructs shared by both stereotypes. Of the 10 shared constructs, only “Stressed” and “Neglect” differed in strength, with both significantly more strongly related to the social class-abuse than race-abuse stereotype, all t s(36–37) < -2.23, p s < .03, d s > .71. This research documents the existence, content, and strength of stereotypes that link race and social class to child abuse. These stereotypes have the potential to lead to medical misdiagnosis of abuse for Black and poor children.


Content and Strength of Medical Professionals' Stereotypes
Research has indicated the existence of racial and class disparities in child welfare contacts, finding that a child's injuries are more likely to be reported as resulting from abuse when the child is either (a) Black rather than of any other race (Flaherty et al., 2008;Hampton & Newberger, 1985; but see Laskey et al., 2012) or (b) from a lower versus higher social class (Flaherty et al., 2008;Hampton & Newberger, 1985;Laskey et al., 2012). Black and Hispanic children have been found to be as much as nine times more likely to be suspected of being abused as compared to White children (Lane, Rubin, Monteith, & Christian, 2002). Yegidis and Morton's (1999) analysis of data from the National Child Abuse and Neglect Data System revealed that, in one state, abuse and neglect were substantiated in cases involving Black children at a rate seven times greater than should be expected based on the demographics of the population in that state. Similar patterns of overrepresentation emerge when examining social class. In Laskey et al.'s (2012) empirical study, pediatricians diagnosed abuse more often in child patients who were from low versus high socioeconomic backgrounds.
It is possible that abuse is suspected and substantiated at disproportionately high rates in cases involving Black or poor children because these children are either actually disproportionately more likely to experience abuse or to use public services that bring them into contact with mandatory reporters. It is also possible, however, that stereotyping and bias contribute to the disparities. To begin, there may be an increased tendency to look for or suspect maltreatment among Black or poor children. Observers also might overlook maltreatment that could be affecting White or middle-and upper-class children. And even after maltreatment is suspected, bias can affect the likelihood that it will be reported to child welfare services or substantiated. Ards, Myers, Chung, Malkis, and Hagerty (2003) accounted for the potential for racial differences in exposure to reporters to produce differences in maltreatment rates by examining non-welfare cases from the National Incidence Study of Child Abuse and Neglect (NIS). Their analysis showed that maltreatment allegations were made more frequently for Black versus White children. Further, this difference accounted for half of the racial disparity in non-welfare cases that were subsequently substantiated and reported to the federal government. In a more recent study by Hymel et al. (2018), racial and ethnic minorities were disproportionately represented in samples of young children evaluated and reported for abusive head trauma at pediatric intensive care units. Moreover, these disparities manifested only among children who were not classified as having experienced abuse. That is, children who were evaluated and reported but ultimately not determined to have experienced abusive head trauma were more likely to be from minority versus White families, whereas no similar racial/ethnic difference emerged among children who were classified as abuse victims. As well, Jenny, Hymel, Ritzen, Reinart, and Hay (1999) conducted a retrospective chart review to identify children who had experienced abusive head trauma. Their results revealed that abuse was significantly more likely to have been missed by a physician in a previous evaluation when children were White rather than racial minorities (37% versus 19%). And with regard to social class, Drake and Zuravin's (1998) analysis of NIS data revealed that lower family income significantly increased the likelihood that suspected maltreatment would be reported to authorities.
Experimental research also points to bias as a factor contributing to race and social class disparities in decision making related to maltreatment. Using data from a national survey of mandated reporters, Zellman (1992) examined responses to several vignettes describing potential physical or sexual abuse in which family social class and race were experimentally manipulated.
In line with Drake and Zuravin's (1998) findings from the NIS, the reporters were more likely to label a child's injury as abuse and perceived the incident as more serious and deserving of a report to authorities when the child's family was described as Black rather than White or as having lower versus higher socioeconomic status.
These findings indicate that nonevidentiary factors are influencing decision-making surrounding child abuse in ways that disadvantage certain children. We suggest that at least some of the disparities in suspicion, reporting, and substantiation of child abuse can be accounted for by stereotypes that link race and social class to child abuse. It is well established that stereotypes can have a subtle yet biasing influence on the way people perceive others, process information, and form judgments, regardless of the extent to which people consciously endorse the stereotypes (e.g., Devine, 1989). It also has long been recognized that child abuse is a socially constructed phenomenon, subject to being identified on the basis of stereotypes about the types of people who are likely abusers (e.g., Gelles, 1975). Indeed, research has documented stereotypes that link crime, violence, and child abuse to Blacks (Devine, 1989) and child abuse to low social status individuals (Gelles, 1979). There also is evidence that nurses believe child abusers are stereotypically Black and from lower socioeconomic backgrounds, and that child abuse is caused by social class status (Krowchuk, 1989).
Although social psychologists have done substantial work identifying the content of different stereotypes (e.g., Devine, 1989;Cuddy, Fiske, & Glick, 2008;Madon, 1997), systematic analyses have not yet identified either the constructs that make up the stereotypes that link race and social class to child abuse, or how central the various constructs are to those stereotypes. Therefore, we performed two studies to identify the content and strength of the constructs that comprise the race-abuse and social class-abuse stereotypes in samples of medical professionals, as they are mandated to report suspected abuse and are common sources of system contact (Sedlak et al., 2010). This work takes the necessary first step toward understanding how stereotypical beliefs related to race and social class might affect diagnostic decision-making and practices of medical professionals in potential child abuse cases.

Study 1
Method Participants. The sample included 53 doctors, nurses, and other medical professionals, 74% of whom were women. Participants were 37 years old on average (SD = 14, range = 21 to 72) and predominantly White (77%, with 8% Asian, 8% Hispanic, 2% Black, 2% Native American, and 4% of other backgrounds). Fourteen other participants were excluded because they worked in a nonmedical setting, had no contact with patients, or had missing data.
Stereotype content measure. The content of the race and social class stereotypes was assessed with a measure modeled from Devine (1989) and Madon (1997). Specifically, participants read that, "A common stereotype in our culture is that [Black or poor] children are more likely than other children to be abused by their parents. Please list all the words you can think of that are associated with this stereotype. Note, these characteristics may or may not reflect your personal beliefs. So, please list words that you know to be part of the stereotype whether or not you believe the stereotype to be true." Following Madon (1997), participants' individual word responses were grouped based on relatedness. Then, one word that best exemplified each group was chosen to represent the stereotype construct. For example, for the race-abuse stereotype, "Uneducated" (n = 8), "Education" (n = 2), "No Education" (n = 1), "Lack of Education" (n = 1), "Uninformed" (n = Demographic measures. Participants were screened for eligibility based on occupation, age, and status as U.S. citizens. They also reported their gender, educational attainment, race/ethnicity, and socioeconomic status. Procedure. Participants were recruited in person at hospitals and online via snowball sampling. Specifically, we solicited local and known medical professionals to participate via in person communication, email, and Facebook. Individuals who received invitations online were also asked to forward the survey to other eligible potential participants they knew. Participants provided informed consent and were randomly assigned to consider either the stereotype that (a) Black children or (b) poor children are more likely than others to be abused. They then completed the stereotype content measure appropriate to the experimental condition to which they had been assigned and reported their demographics. Participants were debriefed and thanked. All procedures were approved by the university Institutional Review Board.

Results
Participants provided an average of 5.71 words for the race-abuse stereotype (SD = 3.67, range = 2 to 16), yielding a total of 131 words that were organized into 25 construct groups. For the social class-abuse stereotype, participants provided an average of 5.31 words (SD = 2.66, range = 1 to 11), providing a collection of 162 words that were organized into 24 construct groups. The constructs are presented in Figure 1. As can be seen, approximately half the constructs are shared by the race-abuse and social class-abuse stereotypes, revealing a considerable degree of confoundedness.

Study 2
To better understand potential effects on medical decision-making, we next explored how strongly related the various constructs identified in Study 1 are to the race-abuse and social classabuse stereotypes.
Eighteen other participants who had missing data were excluded.
Stereotype strength measure. Following Devine (1989) and Madon (1997), participants rated the strength of the relations between the race-abuse or social class-abuse stereotype and those constructs identified in Study 1 that were derived from at least two participants' responses (n = 17 constructs per stereotype). Specifically, participants read, "A common stereotype in our culture is that [Black or poor] children are more likely than other children to be abused by their parents. Please rate each word below for how strongly it relates to this stereotype. Note, these characteristics may or may not reflect your personal beliefs. So, please rate each word for how strongly it relates to the stereotype, whether or not you believe the stereotype to be true." Responses were given on a scale that ranged from 0 (Not at all related) to 4 (Extremely related) and mean scores were calculated for each of the 17 stereotype-related constructs.
Demographic measures. Participants completed the same measures used in Study 1.

Procedure.
Researchers posted fliers at different local hospitals and emailed unknown medical professionals (e.g., heads of pre-medical and nursing departments at colleges and universities, hospital chiefs of staffs) to invite participation in the study. Emails included a request that the individual forward the survey to their colleagues in the medical field. 1 As in Study 1, participants provided informed consent and were randomly assigned to either the race or social class condition. They completed the stereotype strength measure appropriate to their experimental condition, reported their demographics, and were debriefed and thanked. These procedures were approved by the university Institutional Review Board. Figure 2 depicts the degree to which participants perceived each construct to represent the race-abuse or social class-abuse stereotype. These results shed light on the relations both within and between stereotypes. They still appear to be confounded, as the four most strongly rated constructs (i.e., "Drugs," "Neglect," "Poor," and "Stressed") are the same for both the raceabuse and social class-abuse stereotypes. Moreover, of the 10 constructs that overlapped across stereotypes, eight ("Culture," "Cycle," "Drugs," "Misbehaving," "Poor," "Uneducated," "Unmarried," and "Unskilled") were rated as statistically similar in strength for the race-abuse and social class-abuse stereotypes (all ts(36-38) < |1.78|, all ps > .08, all 95% CIs included 0, all Cohen's ds < .57). Only "Stressed," t(36) = -2.40, p = .02, 95% CI [-1.36, -0.11], d = .78, and "Neglect," t(37) = -2.23, p = .03, 95% CI [-1.80, -0.09], d = .71, were rated differently, with both rated as significantly more strongly related to the social class-abuse than race-abuse stereotype.

Discussion
To our knowledge, these studies are the first to investigate the content of stereotypes that link race and social class to child abuse as well as the relative strengths of different attributes of those stereotypes. The results provide a basis upon which we can refine our understanding of the negative impact of race and social class stereotypes on child health and welfare outcomes. One important concern is whether these stereotypes bias medical decision-making in ways that lead to disparities in misdiagnosis of abuse. Differences across the two stereotypes could lead medical professionals to be more predisposed to label children's injuries as products of abuse or to perceive their abuse as more serious or deserving of being reported if they are Black rather than from a low social class background (e.g., Zellman, 1992). Moreover, the considerable overlap between the two stereotypes could lead medical decision makers to assume Black children are also of low social class and exacerbate bias that leads Black child patients and families to be overrepresented in the child welfare system (see, e.g., Verhaeghen, Aikman, & Van Gulick, 2011). Preliminary evidence of this comes from Keenan et al.'s analysis of pediatricians' consultation notes in cases involving suspected child physical abuse. Their results revealed that pediatricians perceived underrepresented minority children as having lower socioeconomic status than White or Asian children, and pediatricians perceived that children who had either low or middle perceived socioeconomic status were at higher social risk than others for abuse and, in turn, were more certain in believing that they had probably been abused (Keenan et al., 2017). This is especially concerning in light of experimental research showing that physicians and nurse practitioners who were exposed to simulated cases of potential child abuse were nearly nine times more likely to misdiagnose an accident as abuse when the child was described as from a high versus low risk social setting (Anderst, Nielsen-Parker, Moffatt, Frazier, & Kennedy, 2016).
Further empirical studies are needed to better understand how the race and social class stereotypes act separately and in concert to affect medical professionals' decision making related to child abuse.
Another issue highlighted by this research is that the ways in which child abuse stereotypes related to race and social class converge and diverge might affect children's and families' outcomes. For instance, constructs such as "Depression" may lead to mental health interventions for low social class parents, whereas constructs such as "Batterer" or "Scary" may lead to criminal justice interventions for Black parents. The differences we found in strength ratings for the constructs of "Stressed" and "Neglect" support the idea that the stereotypes could lead to disparate outcomes in cases of suspected maltreatment, as they may paint a picture of low social class individuals being more in need of aid versus Black individuals being more deserving of punitive action. Further, among the constructs that differed across the two stereotypes, the one most strongly related to the social class-abuse stereotype was "Depression" whereas the one most strongly related to the race-abuse stereotype was "Ghetto." Future research should examine whether the content and strength of the stereotypes generalize to service providers who work with children and families after they are referred to the child welfare system. Such work will yield important insights about whether stereotypes about race, social class, and child abuse contribute to disparities in the likelihood of criminal justice intervention.
These exploratory studies are not without limitations. Using online samples allowed us to access medical professionals, who are a hard-to-reach population, from beyond our immediate geographic area. Even so, our samples are composed heavily of participants who are women, White, and nurses. Although underrepresentative of men and other medical professionals, our samples are similar to the racial and ethnic composition of medical school graduates from 1974 to 2015 (Association of American Medical Colleges, 2016). Due to the nature of snowball sampling, we cannot report response rates or demographic differences between those who elected to participate and those who did not. Future work should use nationally representative samples including a broader range of medical professionals to determine whether our findings generalize to the field at large.
Yet, a measure of confidence stems from the fact that, following previous research (e.g., Devine, 1989), participants reported not their personal beliefs but rather their perceptions of the race-abuse and social-class abuse stereotypes as they exist in our culture. As explained by Fiske, Cuddy, Glick, and Xu (2002), this approach reduces participants' motivation to provide socially desirable responses and, thus, yields a more accurate picture of stereotypes as they are perceived to be in the field. One might suggest that this requires an assumption on the behalf of participants that stereotypes about race or social class (a) include a presumed predisposition for abuse and (b) are generally prevalent in the culture. Yet prior research has suggested the existence of such stereotypes (Devine, 1989;Gelles, 1979;Krowchuk, 1989). Further evidence that the stereotypes we identified are not merely methodological artifacts is garnered by reviewing the considerable overlap across participants' responses: most construct groups were based on multiple words (68% of race-abuse constructs were based on multiple words, with M = 5.24 words per construct, and 75% of social class-abuse constructs were based on multiple, words with M = 6.75 words per construct). Although this suggests the stereotypes are in fact real, further research is still needed to observe whether medical professionals actually endorse them. Of importance, however, even when people do not specifically endorse a stereotype, it may affect their decision-making (Arkes & Tetlock, 2004).
In conclusion, our results provide the foundation for future empirical research regarding the role stereotypes play in contributing to disparities in the child welfare and criminal justice systems. Our findings demonstrate a degree of consensus among medical professionals regarding the existence and content of stereotypes that link race and social class to child abuse. A wealth of social psychological research has shown that stereotypes can affect subjective judgments and decision-making, particularly in ambiguous situations (e.g., Devine, 1989; for review, see Nickerson, 1998). Still the question remains as to whether medical decision makers are affected by stereotypes about child abuse in actual cases. Do the stereotypes cause medical professionals to overdiagnose child abuse in Black and poor patients and/or underdiagnose it in White and middle-/upper-class patients? Can such effects occur even if medical professionals do not consciously endorse the stereotypes? The answers to such questions have implications for understanding the origins of misdiagnoses that subject Black and low social class children to unnecessary intervention and lead other causes of their symptoms to be overlooked, while simultaneously failing to identify actual abuse affecting White and higher social class children. Now that we know the content and strength of the stereotypes linking race and social class to child abuse, future researchers can measure their influence on medical professionals' diagnostic decision-making.

Strength of Relations between Constructs and Child Abuse Stereotypes
Note. The response scale ranged from 0 (not at all related) to 4 (extremely related). Constructs related to the social class-abuse stereotype appear above the horizontal line whereas constructs related to the race-abuse stereotype appear below it.