The Stigma of Obesity: Examining the Relationship Between BMI and the Treatment of Pain in Surgery Patients

There is a growing epidemic of obesity in the United States and a corresponding increase in the number of morbidly obese patients receiving healthcare. Despite the increasing focus and research on obesity over the years, the prevalence of obesity in the United States has continued to worsen. A stigma against obesity exists in the general public including among healthcare professionals. Attitudes and bias of healthcare professionals against obesity can negatively affect judgment and choices related to the enactment of care, affecting both the quality of healthcare delivered and patient outcomes. Studies have shown that stigmatization against groups of patients such as minorities affects healthcare outcomes, however there is a paucity of research related to outcomes of stigmatization against obese individuals. The purpose of the study was to determine if there is a difference in quality of nursing care as measured by medicating for pain between obese and non-obese post-surgical patients. It was hypothesized that obese individuals will receive less pain medication than non-obese individuals. An underlying assumption based on the literature was that stigmatization of obesity by nurses would be reflected in reduced administration of post-surgical pain medication. The greater the stigma present, the less pain medication will be administered. This study used a retrospective chart review of the electronic medical record of three hospitals within a single healthcare system to compare non-bariatric postsurgical pain treatment among normal weight, over-weight, and obese adult patients as a measure of nurses’ stigma. The final data set contained a total of 1704 cases, with 21.4% (n=365) normal weight, 21.4% (n=365) overweight, 21.4% (n=365) obesity class I, 17.1% (n=291) obesity class II, and 18.7% (n=318) obesity class III individuals. BMI scores ranged from 18.5 to 185.9 (M = 33.1, SD = 11.1). Findings showed differences in total dose of day one post-surgical pain medication among the normal, overweight, obesity class I, II, and III patients. Obesity class III patients received less pain medication than the obesity class I and class II patients and significantly less than the overweight patients. Simple linear regression analyses were used to determine the relationship between BMI and pain medication administration on postoperative day one and day two. Hierarchical linear regression was used to determine the relationship between dose on day one and day two and BMI, while taking into account other variables associated with stigma. The relationship between dose of pain medications and BMI was significant and negatively related. For every 1% increase in BMI there was a .17% decrease in the total morphine equivalent dose of narcotic given on postoperative day one. When controlling for other factors related to stigma, there was a .25% decrease in dose for every 1% increase in BMI. Further research is needed to measure attitudes and biases of nurses along with their administration of pain medication to obese patients. Addressing nurses’ stigma of obesity is essential to improving the quality of care of obese patients.

in the number of bariatric surgeries, such as gastric bypass and banding procedures, performed (Zhao & Encinosa, 2007). The increase in bariatric surgeries is alarming because it suggests that there has been a dramatic increase in the number of surgical patients who qualify (i.e., are morbidly obese) for bariatric procedures.
Attitudes and bias of healthcare professionals against obesity can negatively affect judgment and choices related to the enactment of care, affecting both the quality of healthcare delivered and patient outcomes. Stigmatization of obesity has been described as having a negative impact on health and there are a few studies that have shown that obese individuals receive inferior healthcare when compared to that of normal weight individuals. For example, obese patients were less likely to receive pap smears, mammograms, and colorectal cancer screenings (Fagan, Wender, Meyers & Petrelli, 2011), as well as cervical and breast cancer screenings (Wee, McCarthy, Davis & Phillips, 2000). Physicians spend less time with these patients and were found to only give patients a separate diagnosis of obesity 14.4% of the time (Huang et al., 2004, Bleich, Pickett-Blakely & Cooper, 2011. This lack of a separate diagnosis was an important finding because obese patients were less likely to receive weight loss counseling or education without a diagnosis (Tsai & Wadden, 2009).
It is important to study the effect of obesity on healthcare and health outcomes because of the magnitude of individuals that are affected. In order to promote the best health outcomes in obese individuals, it is important to not only describe the stigmatization of obesity that occurs, but also to identify the consequences of that stigma. Despite this increasing focus and research on obesity over the years, the prevalence of obesity in the United States has worsened. Even with this attention, obese individuals still must access healthcare and stigma still exists among healthcare providers. There are research studies that show that stigmatization against other groups of patients affects healthcare outcomes, but no studies could be found that were related to outcomes of stigmatization against obese individuals related to the treatment of pain. If stigmatization against obese patients and resultant reduction in the quality of healthcare can be documented, then interventions to reduce bias and improve healthcare for obese patients can be developed.

Purpose of Research
The purpose of the study is to determine if there is a difference in quality of nursing care as measured by medicating for pain between normal, overweight, and obesity class I, II, and III post-surgical patients. Treatment of pain is an important quality indicator for hospitals and effective treatment is a requirement of accreditation by the Joint Commission. The amount of pain medication received has been studied in other stigmatized groups, for example racial bias resulting in less pain medication received in the emergency department, but this has never been studied in relation to obesity stigma. It is hypothesized that there is a difference in the amount of pain medication administered between normal, overweight, obesity class I, II, and III patients that results from stigmatization of obesity by nurses The greater the stigma present, the less pain medication is administered.
Research Questions. The research questions for this study are: 1. What is the difference in the total morphine equivalent dose of post-surgical pain medication administered between normal weight, overweight, and obese (Class I, II, and III) adult non-bariatric surgery patients?
2. What is the relationship between pain medication ordered and administered and the BMI of adult non-bariatric surgery patients?
3. What is the relationship between patients' BMI and the receipt of post-surgical pain medication, when accounting for race, gender, age, insurance status, presence of psychiatric diagnosis, and pain score during hospitalization?

Significance of the Study
This problem was selected because of the increase in obesity in the United States and the increase in the number of obese individuals seeking healthcare.
Stigmatization of obesity has no place in healthcare because of the potential to impact the lives of millions of individuals and increase healthcare costs. This study is important because it could contribute to an improvement in quality of care, decrease in healthcare costs, and an increase in quality of life of obese individuals. The study has personal importance due the researcher's twenty-year nursing career in the operating room and being witness to the increased need for consideration of obese patients during care surrounding surgery.

Theory
Symbolic interaction is a useful theory in the development of knowledge related to obesity stigma because it can be used to generate a wide range of researchable hypotheses. Meaning is a central to understanding stigma and its effect on health outcomes. The meaning that nurses attribute to obesity can shape their definitions of situations and perspectives towards the patient. Studying these meanings and the ways of helping individuals to reappraise attitudes and definitions of situations may reduce stigma. The processes of symbolic interaction are useful in describing the range of reference groups utilized by nurses in healthcare, how perspectives and stigma arise within these groups, how these perspectives shape nursing practice, and how they may change. The focus of future nursing research related to obesity stigma should be on understanding how negative perspectives arise and removing them by altering perspectives and definitions of situations, which should in turn remove stigmatizing actions and improve care provided to obese patients.

Premises and Assumptions
Premise 1. A major premise of symbolic interaction, and this study, is that nurses act toward obese patients based on the meaning they attach to being obese.
Negative bias or stigma would be reflected in a reduced pain medication administration. If nurses attach being lazy or lack of self-control to obesity, then biased attitudes may occur.
Premise 2. Another premise is that this meaning arises through interaction with others, for example medical students interacting with biased residents and physicians, or through observing the interactions of others in person or through the media. Negative attitudes toward and interactions with obese individuals can be seen daily through mass media.
Premise 3. The final premise is that meanings are assigned and modified through an interpretive process. The meaning that individuals attach to things is constantly changing. This can be seen in studies that described less negative attitudes towards obesity in individuals who are educated about the uncontrollable causes of obesity.
Assumption. The major assumption of this study is that stigma of obesity is not just an attitude or bias, but translates into behaviors that effect the provision of care given by healthcare providers, meaning that there is a lower quality of care provided to obese as compared to normal-weight individuals. This assumption reflects Premise 1 of Symbolic Interaction, whereby the meaning attached to something is reflected in actions toward it.

Study Design
Treatment of pain was used as an indicator of obesity bias/stigmatization among healthcare providers in this study. A retrospective chart review was performed that examined data from the data warehouse of a large Rhode Island healthcare system. The warehouse consists of multiple data bases that save data from the multiple electronic medical record applications that are used within the system. Data for this study was collected specifically from the medication administration, the admitting, and the computerized physician order entry application. Data from these systems will be utilized to determine any differences or relationships between a patient's weight status, as measured by their body mass index (BMI), and the amount of postoperative pain medication they receive. The five BMI categories used in this study were determined based on classification by the Centers for Disease Control (2015) of normal (18.5 -24.9), overweight (25 -29.9) and the World Health Organization (2014) obesity class I (30 -34.9), II (35 -39.9), and III (≥40). The association of treatment of pain to obesity stigma has not been studied, but has been studied with other stigmatized groups, for example in minorities (Pletcher, Kertesz, Kohn & Gonzoles, 2008;Sabin & Greenwald, 2012).

Summary of Chapters to Follow
In chapter 2, a review of the literature describes the concept of stigma and defines its properties.

Concept of Stigma
To "stigmatize" has been defined as to describe or identify in opprobrious (vulgar, slanderous, abusive) terms (Merriam-Webster.com, 2013). Link & Phelan (2001) described stigmatization as the convergence of distinguishing and labeling human differences, linking labeled persons to negative stereotypes, separating "them" from "us", and status loss and discrimination of the labeled person which co-occur in a power situation that allows the components of stigma to unfold.
Stigma, and the resulting stigmatization, is a complicated, multifactorial concept that was first comprehensively examined in the works of Goffman (1963), and later adapted for various situations in psychology, sociology, and various healthcare professional literatures. Stigma has been much studied, but a unified definition remains elusive. Many studies provide no explicit definition, or they quote Goffman's definition where stigma is an attribute that is deeply discredited (stigmatized). Many studies deconstruct stigma into a list of attributes possessed by the stigmatized, and do not examine it as a language of relationships (Goffman, 1963). The reason for this definitional ambiguity may be the fact that the study of stigma has been multidisciplinary and each discipline has applied a slightly different definition in order to fit a wide variety of professional lenses and situations (Link & Phelan, 2001).
Multiple definitions for this concept may be appropriate given the different research questions being asked, but it does contribute to confusion regarding stigma.
The concept of stigma related to obesity can be defined by describing the stigma, the attitudes and biases related to the stigma, and the effect stigma has related to healthcare. The stigma of obesity sets individuals apart from those who are not obese. The stigma connotes a set of negative attributes for which an obese individual is stigmatized. The obese individual is seen in a negative light due to deficient morals, which causes others to act differently toward the individual. Stigmatization is less likely to occur if there is a perceived cause for the obesity. Research has been conducted that shows that the non-obese react more favorably to obese individuals only if they perceive that the excess weight is beyond the control of the individual, for example the presence of a thyroid condition. This sets up a condition of inequality where those who are perceived to lack control of themselves and do not have any biological excuse for being overweight will be treated less favorably. In a review of literature, Wright & Whitehead (1987) found that fatness was stereotyped and that the more it was perceived that an individual was responsible for their obesity, the more they were disliked. Obese individuals who presented with a condition, such as a thyroid problem, were judged as more likeable, having more self-control, and were judged more attractive than obese individuals who did not have a physical cause to their obesity (DeJong, 1980). Individuals who were educated regarding the controllable causes of obesity are more likely to endorse negative stereotypes than are individuals who were educated about the uncontrollable causes .
The nature of stigma is that it is present among groups that exist outside the societal norm. Crocker & Major (1989) described stigmatization as occurring towards oppressed social categories of people toward which others hold negative attitudes, stereotypes, and beliefs, who are vulnerable to being labeled as deviant, and are who targets of prejudice or victims of discrimination. The recipients of stigmatization receive a disproportionately poor interpersonal or economic outcome relative to members of the society at large. Also, a stigmatized group is an out group relative to the dominant group in a culture or society, whereas an out group is defined by reference to any particular in-group, regardless of which group holds the dominant position in the social hierarchy. Puhl & Brownell (2003) suggested that a stigmatized person possesses an attribute or characteristic that conveys a social identity that is devalued in some particular social context. Puhl & Brownell (2006) also associated stigmatization with weight bias and stereotyping. Balogh-Robinson (2011) described stigmatization as a weight bias, prejudice, discrimination, or stereotype. Stigmatization occurs when any personal attribute is deeply discredited to its possessors; including "tribal stigmata," "abominations of the body," and "blemishes of individual character" (Goffman, 1963). Goffman also described it as the relationship between an attribute and a stereotype. Lewis & Van Puymbroeck (2008) described stigmatization as the discriminatory acts that result from the social disapproval tied to existing negative attitudes toward people perceived as being overweight. Groups affected by stigma. Stigma associated with certain groups requiring healthcare is very prevalent within society. It is described as affecting multiple groups of individuals and occurs in every culture. Stigma has been described within multiple groups, such as stigma related to HIV/AIDS (Herek, Capitanio &Widaman, 2002;Letamo, 2003;Vanable, Carey, Blair & Littlewood, 2006), mental illness (Alonso et al., 2009;West et al., 2011), illicit drug use (Ahern, Stuber & Galea, 2007), epilepsy (Jacoby & Austin, 2007), smokers (Stuber, Galea & Link, 2008), and skin disorders (Chaturvedi, Singh & Gupta, 2005). The common thread that these different types of stigma share with obesity stigma is that the individual possesses and/or displays a mark or behavior that identifies them as belonging outside what is normal or acceptable in society. Stigmatization of a group occurs whether it is perceived that the stigma may or may not be the fault of the individual.
The characteristics of stigma are typically defined by the attributes being stigmatized. Stigma related to HIV is associated with homosexuality, promiscuity, and drug use. Stigma related to mental illness can be associated with inappropriate or bizarre behaviors, instability, and lack of personal hygiene. Stigma related to drug use is associated with lack of morals, criminal behavior, and drug addiction as a disease.
The stigma of epilepsy is characterized by having a perceived mental illness, being possessed, and lacking intelligence. Stigma related to smoking is associated with lacking willpower and putting one's health, or another's health, at risk. Stigma related to skin diseases is associated with being unclean. Characteristics of stigma such as controllability, concealability, and entitativity, greatly affect psychological and behavior reactions to the stigma (Major & O'Brien, 2005) If the stigma is controllable, then individuals are more likely to possess negative attitudes and bias toward the stigma. Stigma that is concealable is less likely to be stigmatized, yet individuals may feel shame and may spend considerable effort trying to hide stigmatized attributes.
Entitativity relates to the cohesiveness of a group. The presence of stigmatized attributes can activate stereotypic beliefs that cause them to be considered not only physically, but psychologically similar to other members of the group. For example, the stereotypic belief that obese individuals are lazy and lack willpower may be applied to all obese individuals.

Effects of stigma. Stigma can greatly affect an individual who is stigmatized
and multiple consequences that can occur. Stigma has been described as decreasing self-esteem, lowering academic achievement, and placing an individual at greater risk for mental and physical health problems (Major & O'Brien, 2005 Stigma of obesity related to healthcare professionals. Negative attitudes and beliefs related to obesity exist in the literature among healthcare students and professionals. Negative or biased attitudes, such as slow, like food, overeat, are insecure, and have low self-esteem, were described as present in physician assistant students (Wolf, 2010) where 13.6% of the physician assistant students studied displayed a high level of fat phobia. Waller, Lampman & Lupfer-Johnson (2012) described weight bias in nursing and psychology students. Both student groups displayed a significant implicit weight bias that was greater towards women than men.
Nursing students were also described as having fat phobia and negative attitudes toward obese patients (Poon & Tarrant, 2009). Registered nurses were compared to nursing students in the study and had a significantly greater fat phobia. Both groups perceived obese people as liking food, over eaters, shapeless, slow, and unattractive.
Dietetic students were more likely to describe a poor health status and diet quality towards obese patients, even with identical health profiles among all individual scenarios (normal and overweight) (Puhl, Wharton & Heuer, 2009). These students displayed a moderate amount of fat phobia and rated obese patients as less likely to comply with treatment. Persky & Ecclesten (2011) described that medical students displayed more negative stereotyping towards obese patients, as well as rating them less likely to comply with treatment recommendations. Students attributed more responsibility to obese patients for potentially weight-related health problems. Few medical students who have fat bias are aware of this bias (Miller et al., 2013). Due to the common lack of explicit bias in healthcare providers, other measures are needed to determine the presence of implicit bias. This could be accomplished with use of a scale that measures implicit bias in combination with a measure of patient outcomes among stigmatized and non-stigmatized groups. The presence of stigmatization in healthcare students is disturbing because of the potential that they will carry these attitudes and biases forward into practice. There may be a great opportunity to change attitudes in healthcare by focusing more study on students. This lack of bias awareness also holds true for other healthcare providers, such as physicians. A review of studies measuring implicit bias in physicians towards stigmatized groups found that obese, black, Hispanic, elderly, and women patients were the target of more bias (Chapman, Hebl & Xu (2001) described physician responses to mock medical records of patients who were average, overweight, and obese presenting with a migraine headache. Physicians viewed heavier patients more negatively and reported that they would spend less time with them than average weight patients. Physicians were more likely to perceive their obese patients as non-adherent to medications (Huizinga et al., 2010), which has been shown in other studies to affect physician prescribing patterns.
Patients perceived as non-adherent may not receive guideline recommended care and may result in a delay in prescribing recommended medications for HIV, acute coronary syndrome, and hemophilia, or intensifying therapy for diabetes. A clinician's own body weight also affects healthcare received. Patients reported less confidence in care provided by overweight physicians (Hash, Munna, Vogel & Bason, 2003) and thin and overweight pediatricians reported more difficulty with weight loss counseling than average weight physicians (Perrin, Flower & Ammerman, 2005). Attitudes and bias of healthcare professionals towards obesity are often not explicitly demonstrated, but implicit (Schwartz et al., 2003).  (Bleich, Pickett-Blakely & Cooper, 2011). While diet and exercise counseling were more likely to occur with a diagnosis of obesity, pediatric care providers were more likely to provide diet and exercise counseling than other specialties, including family physicians and general practice providers (Cook, Weitzman, Auinger & Barlow, 2005).
Postoperative Pain. This study examined the treatment of postoperative pain and its relationship to a patient's BMI. Postoperative pain is an important topic to study since it has significant effects on health outcomes, such as increased lengths of hospital stay and delays in returning to activities of daily living (Morrison et al., 2003). Pain greatly affects the patient's ability to ambulate and immobility has been described as increasing 6-month mortality rates (Siu et al., 2006), while early ambulation has been associated with quicker return of functional capacity and a greater discharge to home after surgery (Oldmeadow et al., 2006). Mental status decline in geriatric patients has also been described in relation to the presence of postoperative pain (Lynch et al., 1998;Duggleby & Lander, 1994). A decrease in postoperative myocardial ischemia has been found in elderly patients receiving effective pain control after surgery (Scheinin et. al, 2000).
Stigma and treatment of pain in other stigmatized conditions. While the study of bias and its effect on the treatment of pain has not been studied in obese patients, it has been studied as it relates to other groups. Stigma related to the provision of healthcare and the treatment of pain has been described in association with age, gender, race, and mental illness.
Studies have described that age is a factor in receiving analgesia, with older adult patients receiving less than younger patients (Jones, Johnson & McNinch, 1996).
Also, several other age-related factors have been described among older adult patients that may affect the administration of pain medication, such as challenges of assessment of pain in older adults, the under reporting of pain, atypical manifestation of pain in older patients, and misconceptions regarding tolerance and addiction (Cavalieri, 2005).
Women experiencing more severe pain than men (Cepeda & Carr, 2003), and a gender bias has been describe in pain management, with women receiving more analgesia than men (Fillingim et al., 2009). In a review of the literature, Hoffmann & Tarzian (2001) found that women were less likely to be taken seriously and receive adequate treatment for pain. Pain was often ascribed to psychiatric causes in women.
In a study examining the effect of race and gender on physician pain management decisions, Weisse, Sorum & Dominguez (2003) found that physicians treat women less aggressively for pain.
In studies related to bias and race, treatment of pain has been demonstrated to be less in minorities than in white patients (Pletcher et al., 2008;Sabin & Greenwald, 2012) and in black patients, unless they exhibited demanding or angry behavior (Burgess et al., 2008). Mills et al. (2011) described that nonwhite patients who presented to the ED for pain were less likely than whites to receive analgesia and waited longer for their opiate medication. Pletcher et al. (2008) described differential prescribing of opioids by race/ethnicity for all types of pain. In a study examining the effect of implicit bias on pediatric physicians' treatment recommendations, Sabin & Greenwald (2012) found an association between implicit bias and patient's race in prescribing a narcotic medication for pain following surgery. It is possible that treatment of pain may also be lacking in other stigmatized groups, such as in obese patients.
Insurance status has been used in multiple studies (Vijayakumar, et al., 1995;Hong, Baumann, & Boudreaux, 2007) as an indicator of socioeconomic status. Bird & Bogart (2000) have described it as a perceived reason for discrimination during healthcare provision.
Having a psychiatric diagnosis is a stigmatized condition and is associated with poorer healthcare outcomes (Zolnierek, 2009). Patients with psychiatric diagnoses often have their physical health problems attributed to their mental illness (Thornicroft, Rose & Kassam, 2007). Primary care providers have been described as having significantly increased negative attitudes towards patients with mental illness, as described by the presence of stereotyping and attributing of negative attitudes (Mittal et al., 2014).
Psychobiological. The Psychobiological approach could provide nurses with insight into the mechanisms that control energy intake and expenditure and has been used by nurses to gain understanding of obesity and to provide education to obese patients. When one understands the causes that lead to weight gain, they can better modify behaviors or the environment to alleviate these causes. This approach incorporates understanding of the biological effects on weight management, such as hunger, craving, hedonic sensations, appetite, meals and their constituents, metabolism, and interactions with brain. Also incorporated is the idea of negative feedback and that if individuals eat too quickly, they may eat a larger portion before satiety signals are stimulated. The environment is also important in that nurses and patients can identify triggers of overeating and reduce or eliminate them.
Understanding metabolism can help identify ways of increasing activity; taking into account domestic, financial, and environmental factors; which will increase metabolism and decrease the effect of starvation metabolism. Understanding energy intake is also important since self-report food intake is often underestimated.
Additional nutrition-focused education could stress the need for more accurate assessment of energy intake. Greater support may potentially reduce environmental stress. Support can be in the form of emotional, informational, instrumental, or appraisal (Peterson & Bredow, 2009).

Socioecological Systems Models.
Obesity stigma is a multifactorial issue and a socioecological model would be useful in determining the individual, family, nurse/clinician, institutional, and societal factors that may increase or decrease stigma (Steele et al., 2011). These systems exist as nested structures, moving from the innermost structure outward. The structures are microsystems, mesosystems, exosystems, macrosystems, and chronosystems (Bronfenbrenner, 1994). In relation to obesity stigma, microsystems consist of activities, roles, and relationships experienced by the obese individual in a given setting with specific physical, social, and symbolic attributes that permit or inhibit engagement or activity in the immediate environment.
Mesosystems are made up of the linkages and processes between two or more settings containing the obese individual, such as the relationship between home and a particular healthcare setting. Exosystems are comprised of linkages and processes that occur between two or more settings where at least one does not contain the obese individual, but where events occur that affect them. Macrosystems are comprised of the overall pattern of micro, meso, and exosystems and could be considered "as a societal blue-print for a particular culture or subculture" (Bronfenbrenner, 1994, p. 40). The chronosystem comprises the change over time of characteristics of an individual and their environment.
Symbolic Interaction. One commonly used theoretical perspective that is helpful in guiding research in the area of obesity stigma is symbolic interaction.
Following the theoretical framework, the actions of human beings toward things are based on the meaning that human beings attribute to them. This meaning is the result of social interaction and can change based on how humans interpret encounters with others (Blumer, 1969). The world exists separately from the individual, but can only be interpreted through the use symbols in the process of interaction. Objectivity can only be approximated because the world is seen through the lens of meaning derived by many individuals.
There are three main premises that underpin symbolic interaction. The first is that humans act toward things based on the meaning that they attach to them. The second is that meaning comes from social interaction. The third premise is that meanings are assigned and modified through an interpretive process. In relation to obesity stigma, nurses act toward obese patients based the meaning they attach to being obese. If nurses attach being lazy or lacking of self-control to obesity, then biased attitudes may occur. This meaning may arise through interaction with others, for example medical students interacting with biased residents and physicians, or through observing the interactions of others in person or through the media. Negative attitudes toward and interactions with obese individuals can be seen daily through mass media. The meaning that individuals attach to things is constantly changing.
This can be seen in studies that described less negative attitudes towards obesity in individuals who are educated about the uncontrollable causes of obesity.
Within situations, actions arise based on an individual's own interpretation of meaning. Stigmatizing actions by nurses towards obese individuals can be explained using this symbolic interaction process (Burbank & Martins, 2010). Following the process, nurses have interactions with reference groups that shape their perspectives regarding obese individuals. These groups can be medical and healthcare organizations, or society at large. When an obese individual enters into the healthcare system and interacts with the nurse, negative perspectives fostered by these reference to its importance, influences other aspects of life, such as political organization, ideology, religion and culture: 'the ideas of the ruling class are in every epoch the ruling ideas: that is, the class which is the 'ruling material force' of society, is at the same time its ruling intellectual force'" (Marx & Engels, 1994, p. 15). Marxist writers authors would analyze obesity stigma as a social problem that is directly linked to the changing mode of production: definitions of obesity stigma and other social problems are influenced by both the economic and social structures and the core values of particular modes of production existing in a historical time period. The goal of critical theory is to create a life free from unnecessary domination (Kim & Holter, 1995).
Habermas described a framework for knowledge specified within three categories; technical, practical, and emancipatory cognitive interest (Kim & Holter, 1995). Technical interest is achieved through the application of empirical-analytic science and predictive knowledge is obtained. Understanding in social life is the orientation of practical interest and is achieved through reflective judgement and interpretive understanding evident in the historical-hermeneutic sciences. Knowledge gained through these two categories is not sufficient for full understanding of social phenomenon. Critical theory goes beyond knowledge gained through empiricalanalytical and historical-hermeneutic sciences by examining power relationships and creating knowledge oriented toward liberating individuals from domination through a process of self-enlightenment.
The medicalization of obesity is evidenced by the increase focus on obesity as a health problem. Much effort and money is spent in the media, weight loss supplements and programs, and bariatric surgeries. Even with all this effort, stigma exists because obesity exists as both a medical and a social problem. The stigma of obesity could be considered a result of ideological hegemony, or how relationships of domination and exploitation are embedded within the dominant ideas of society (Burbank & Martin, 2010). The implicitness of negative attitudes regarding obesity in healthcare relates how society has internalized the idea that obesity is brought upon oneself by sloth and overindulgence. Obese patients accessing healthcare are in a vulnerable position and possess little social power in the nurse/clinician-patient relationship. Patient encounters, such as the administration of pain medication by a nurse to a patient, exist on the micro level. Habermas' communicative action theory, which emerged from critical theory, would allow linkage of macro societal/organizational issues to the micro level of the patient encounter.
Critical Interactionism. Another framework that would be useful in the study of obesity stigma would be critical interactionism. Critical social theory and symbolic interaction are combined, taking into consideration both downstream and upstream factors when developing research related to obesity stigma. Martins & Burbank (2011) compared and contrasted symbolic interaction and critical social theory and described areas of divergence and synergy (Table 1). Obesity stigma is a complex health issue and involves the individual and professional groups, healthcare organizations, and society at large. Both micro and macro approaches need to be incorporated into interventions designed to alleviate the stigma.

Measures of Stigma
Negative attitudes exist within stigmatized groups, such as obese individual.
Scales have been used to assess attitudes of obese individuals towards obesity and obese patients. Scales have also been used to assess attitudes of healthcare providers. The Fat Phobia Scale has been used to assess attitudes toward obese patients (Poon & Tarrent, 2009;Puhl, Wharton & Heuer, 2009;Wolfe, 2010). The Weight Implicit Association Test (IAT) has been used in studies (Miller et al., 2013;Schwartz et al., 2003) to assess implicit weight bias. Poon & Tarrent (2009)  The issue of social desirability is an important consideration for all selfreported measures that assess negative or sensitive issues (Krumpal, 2013). The likelihood of participants telling the truth depends on the perceived risk related to socially undesirable situations. Participants may fear embarrassment, reactions of an interviewer, reactions from family and friends present, or retribution such as loss of job or position. A threat to one's self concept may occur in answering questions negatively that may make them look bad. Nurses want to believe that they treat everyone equally and holistically.
There are several ways to minimize and control for the effect of social desirability bias (Nederhof, 1985). A self-administered survey may increase the likelihood of accurate reporting since there is a higher degree of anonymity than if the survey was administered by an interviewer, although there may be issues with participants not answering sensitive questions if there is a perceived risk of privacy breach. Anecdotally, healthcare staff often expressed concern about the privacy of certain workplace surveys they are asked to take, such as employee satisfaction surveys, even when they are reassured of the anonymity of the survey. During randomizing device, such as a die or coin, to decide which questions they will answer.
The interviewer does not know which question was selected by the participant or their response. Also, the unmatched count technique could be used. Participants are divided into two groups, where one group answers a list of non-sensitive questions and the other answers the same list plus the sensitive questions. Questions that participants answered "Yes" to would be counted and that number reported to the interviewer.
Another method that may be used to decrease social desirability bias would be to include a social desirability bias scale within a questionnaire, for example Latner et al. (2008) described the use of the Marlowe-Crowne Social Desirability Scale when measuring for bias toward obese, homosexual, and Muslim individuals. A higher score on the scale would indicate that a participant may be more likely to under report negative attitudes. Based on the score, researchers would have the opportunity to discard the data, adjust the data to account for the bias, or merely recognize that social desirability bias was a factor within the study and mention it as a limitation.
While the presence of stigma is apparent from studies of obese individuals, the evidence that stigmatization occurs is less conclusive from studies of healthcare providers. It is difficult to assess stigma directly from healthcare providers because of social desirability. Explicit measures of bias are difficult to obtain because subjects may be reluctant to report negative attitudes. It is possible to indirectly measure the presence of bias based on measuring the results of the bias. There is an association between attitude and behavior and the quality of healthcare obese patients receive, for example less screenings for mammography or Pap smear (Ǿstbyte et al., 2005) or the absence of weight loss counseling (Tsai & Waden, 2009). Measurement of these healthcare inequalities would provide a way to identify the presence of stigmatization while decreasing the effect of social desirability. The issue of privacy is not an issue since individual healthcare providers are not directly observed or questioned.
There are several pros and cons to doing studies that link healthcare outcomes to attitude and biases. Biases, such as social desirability, associated with individuals responding to surveys or interviews would be eliminated. Researchers would be studying what was documented and not directly studying healthcare providers. Studies can be done retrospectively potentially giving the researcher access to more data. If more data is required, researchers can look back over a broader period. This would be easier, for example, than trying to recruit more participants to take a survey. There are less ethical issues related to human subjects, such as ensuring confidentiality and the need to obtain informed consent, in retrospective record reviews. Such studies would be measuring actual healthcare outcomes to determine the presence and effect of bias rather than the results of an experiment. These studies can find associations between variables. The strength and direction of the relationships can be determined, opening the way for further study and possibly the determination of causative factors.
The cons of using this type of study are that it would measure an association, which would not demonstrate that obesity bias is a causal factor in obese patients receiving less pain medication and that many be other unknown factors may be affecting pain medication administration besides bias. For example, a patient may have other health issues that are causing increased pain and results in increased pain medication administration, or possibly the patient has a higher tolerance for pain, or other non-medication pain relief are being used. Another con related to doing these studies retrospectively is that the researcher is relying on the accuracy of the data.
Anecdotally, healthcare providers do not consistently paint a vivid clinical picture with their documentation. Patient information is often missing, either in error or because it was never assessed.

Chapter Summary
The concept of stigma and its presence in healthcare was explored. Within healthcare, obesity is a condition of excess adipose tissue that is stigmatized. There were few studies found that examined obesity stigma in healthcare, although it has been described in relation to other healthcare conditions, such as HIV/AIDS, mental illness, illicit drug use, epilepsy, and smoking. Obesity stigma has multiple consequences on healthcare outcomes, both from the patient perspective (e.g., decreased self-esteem, healthcare avoidance, etc.) and from a healthcare provider perspective (fewer referrals for screenings, less time spent with the patient, etc.).
Selection of pain medication administration was based on the findings of the effect the presence of stigma has on patients receiving less medication. Pain medication administration has been studied in other stigmatized groups, but has not been studied in regards to obesity stigma. Pain medication was also selected in order to reduce the effects of social desirability. After review of multiple theoretical frameworks associated with the study of stigma, symbolic interaction was selected for this study.
The selection was based on the frameworks alignment with other studies that explored the meaning individuals held regarding obesity and how it affected attitude.
Following exploration of the concept and theoretical framework, the electronic medical records related to patient demographics, ordering, and medication administration were queried. The measurement of dependent and independent variables is described in Chapter Three. Also described in the next chapter are sample selection, data analysis, and ethical consideration.

METHODOLOGY
Treatment of pain has been used as an indicator of bias/stigmatization among healthcare providers in several studies (Sabin & Greenwald, 2012;Mills et al., 2011;Burgess et al., 2008;Pletcher et al., 2008), was used in this study. Undertreated postoperative pain is associated with negative healthcare outcomes such as longer hospital stays, missed or shortened physical therapy sessions, and less ability to ambulate (Morrison et al., 2003). The relationship of pain treatment to obesity stigma has not been studied, but as described in the previous section has been studied with other stigmatized groups.

Research Questions
This quantitative study was designed to answer the following research questions: 1. What is the difference in the total morphine equivalent dose of post-surgical pain medication administered between normal weight, overweight, and obese (Class I, II, and III) (World Health Organization, 2014) adult non-bariatric surgery patients?
2. What is the relationship between pain medication ordered and administered and the BMI of adult non-bariatric surgery patients?
3. What is the relationship between patients' BMI and the receipt of post-surgical pain medication, when accounting for race, gender, age, insurance status, presence of psychiatric diagnosis, and pain score during hospitalization?

Design
This study used a retrospective chart review to compare pain treatment among normal weight, over-weight, and obese patients as a measure of stigma. This design supported the intent of the study, which is to examine the differences in pain medication administration based on BMI, the relationships of pain medication ordering (as entered by the physician) and administration (as documented by the nurse) between different categories of BMI, and if there is a relationship between BMI and the amount of pain medication administered.

Sample
The sample was obtained through Information System (IS) query of the hospital system data warehouse, which contains data from electronic medical records from three hospitals and has been used by the healthcare system going back to 1993.
The IS analyst employed through the hospital system from where the data was taken was given specific inclusion and exclusion criteria to extract the data.
Inclusion criteria for the study were adult patients, 18 years of age, who must have been specifically admitted for a surgical procedure. Patients who were admitted for bariatric surgery, such as gastric bypass or gastric banding surgery, or were admitted for medical reasons, but ended up needing surgery, were excluded from the study. The reason for excluding bariatric surgery was that prior studies have described that patients who were considered at fault for their obesity were more likely to generate negative attitudes from study participants than patients that were described as not being responsible for their obesity, such as in patients with thyroid conditions (Dejong, 1980). Also, it was possible that nurses working in the specialty of bariatric surgery may display less bias related to an increased exposure to obese patients.
Patients who were admitted to the hospital for medical reasons were excluded because their pain may have been influenced by factors other than surgery. Patients undergoing bariatric surgery could be considered as taking responsibility for their weight and taking action, therefore may experience less stigmatization than non-bariatric patients.
Underweight patients were excluded from the study since being underweight may be associated with other healthcare issues that may impact pain during hospitalization.
Patients who had patient-controlled analgesia ordered in the electronic medical record by a licensed independent practitioner during the post-operative period were excluded since they administer their own pain medications and nurses do not.

Measurement of Dependent Variables
Pain Medication Administration. Data related to pain medication administered during the first (postoperative day one) and second (postoperative day two) 24-hour period, for a total of 48 hours after surgery, was collected. This included the number of times that intravenous narcotics, oral narcotics, and non-narcotics were administered, as well as the average dose during each 24-hour period. The electronic medication administration records for the sample were examined to determine which pain medications were administered and are described in Table 2. Pain medications were identified from a list extracted from the electronic medical record data based and if any of these medications that were administered during postoperative day one or day two the patients were included in the study. Intravenous narcotics of interest during review of the medical record included morphine, hydromorphone, fentanyl and meperidine. Non-intravenous narcotics included hydrocodone and oxycodone. Nonnarcotics included acetaminophen, ibuprofen, ketorolac, and diclofenac. Since it is difficult to do an overall comparison of pain medication due to differing amounts, strengths, and doses used between the IV and PO pain medications, narcotic equivalency used in this study. This method has been used in other studies to facilitate analysis between different opioid medications (Olson, Hanson, & Michaud, 2003, Fillingim, Doleys, Edwards, & Lowery, 2003, Allen, et al., 2003. Conversion tables should be taken as approximations and not as absolute doses. Tables often describe different conversion and dose calculations and caution is prescribed when using these in actual clinical practice (Shaheen et al., 2009). The conversion table used for this study was based on information from GlobalRPh (2015). Narcotic equivalency is an approximation and compares a given oral or intravenous pain medication dose to the equivalent dose of oral morphine and approximations used in this study are listed in Table 3. For example, oxycodone is 1.5 times stronger than oral morphine; therefore 10 mg of oxycodone would be equivalent to approximately 15 mg of oral morphine.
Conversions were done for each medication and added together to determine the total equivalent dose of oral morphine that each patient received during the first and second 24-hour period after surgery.

Measurement of Independent Variables
Age. Age was measured in years and was selected as a variable because of its potential effect on the treatment of pain. Older adult patients have been shown to receive less pain medications than younger patients (Jones, Johnson & McNinch, 1996), therefore it was important to account for age as a potential bias contributing to the amount of medication received by the patient.
Gender. Gender was measured as male or female. This variable was chosen because the described differences in the experience of pain between men and women.
It was necessary to measure gender to account for bias since women were less likely to be taken seriously, receive adequate treatment for pain, have pain ascribed to psychiatric causes, and be treated less aggressively.
Race. Race, as valued within the electronic medical record, was described as  (Vijayakumar, et al., 1995, Hong, Baumann, & Boudreaux, 2007 and has been perceived as a reason for discrimination during healthcare provision. Use of insurance status is not an exact measure of socioeconomic status, but status may be inferred. For example, Medicare is offered to low income individuals.
Psychiatric Diagnosis. Presence of psychiatric diagnosis was positive if any psychiatric diagnosis listed in the Diagnostic and Statistical Manual of Mental Disorders exists for the patient during the hospitalization. Mental illness has been described as a stigmatized condition and therefore it was important to account for it during this study.
Pain Score. Patients with higher pain scores receive more pain medication, so it was important to account for pain score to understand if less pain medication is administered to obese patients. Pain was assessed on a scale of 0-10, with 0 being no pain and 10 being the worst possible pain ever. This scale is a subjective measure of pain that is reported to the nurse by the patient. Pain score was entered by the nurse into the medication administration as ordinal values and were described as mild (1-2), moderate (3-4), moderate/severe (5-7), and severe (8-10).

Data Analysis
IBM SPSS Statistics 23 was used to analyze data for this research study. Data for the first research question was analyzed using one-way ANOVA. The dependent variable was pain medication administration and the independent variable was BMI category. Pain medication administration was divided into total narcotic equivalency doses, and the total dose of intravenous and oral narcotic and non-narcotic medications given during a first and the second day postoperative period. The groups used were the BMI classifications of normal weight, overweight, and obese class I, obese class II, and obese class III.
In addressing the second research question, the number of intravenous and oral narcotic, and non-narcotic medications ordered and administered were compared between the five BMI categories. A chi-square test was used to determine if there was a significant difference between expected (ordered by a licensed independent practitioner) and observed (administered by the nurse) medication, dose, and frequencies.
Regression analyses were used to address the third research question. Simple linear regression was used to determine the relationship between the total morphine equivalent dose and BMI. Separate analyses were performed for the dose on postoperative day one and day two to determine if there were any differences in the relationship between days. Hierarchical linear regression was used to determine the relationship between dose on postoperative day one, and then the dose on day two, and BMI after controlling for other factors related to stigma and to the amount of pain each patient reported. The independent variables for each regression analysis will be BMI (continuous), race (categorical), gender (categorical), age (continuous), insurance status (categorical), presence of psychiatric diagnosis (categorical), and average pain score during hospitalization (categorical). Dummy variables were created for each of the categorical variables within the regression analysis.

Ethical Considerations
Institutional Review Board approval was obtained from both the hospital system (Appendix A) and the University of Rhode Island (Appendix B) prior to conducting this study. Consent was not obtained because this was a retrospective chart review using de-identified data, posing no more than minimal risk, not affecting the rights and welfare of the subjects, and the consent would have been the only documentation linking the study to the patient. A waiver would not adversely affect the rights and welfare of the subjects since the study was a retrospective chart review and all identifiers was removed. The sample size required was large and including only those samples/records/data for which consent can be obtained would prohibit conclusions to be drawn or bias the sample such that conclusions would be skewed.
Also, since the potential time period being looked at was ten years, the proportion of individuals likely to not be able to be contacted due to having relocated or died would be a significant percentage of the subject population and the research results may not be meaningful and lose statistical power. All data points were supplied by IS query and did not contain one of the 18 HIPAA identifiers (U.S. Department of Health and Human Services, 2012). All data points were supplied by IS query and were deidentified by the Information System (IS) staff prior to delivery for data analysis and each patient was be assigned an identification (ID) number. Chart review was not performed remotely. The study did not include the use of investigational drugs, devices, or psychological interventions. Breach of confidentiality would have been the only possibility, but was prevented by the use of de-identifying the data. The potential benefits to research subjects as a result of the study would be the identification of obesity stigma as a problem as related to post-operative pain medication administration, creating an impetus to develop interventions aimed toward healthcare providers, thereby improving outcomes and the quality of care.  Table   4. The table lists ICD 9 procedure code descriptions that were performed on more than one patient. The descriptive statistics for each variable based on BMI are listed in Table 5.

Dependent Variable
Pain medications documented as given during the first 48 hours after a procedure were hydromorphone, morphine, fentanyl, meperidine, oxycodone, ketorolac, acetaminophen, and ibuprofen. A breakdown of frequency can be seen in Table 6, which displays the pain medication that were listed as being administered, as well as the number of patients that the medication was given to. Hydromorphone was a common pain medication during the first 48 hours after surgery and was given to 86.7% (n=1478) of patients during the first 24 hours and 18.2% (n=310) during the second 24 hours. Descriptive statistics in Table 7

Independent Variables
Age. The age of the study population ranged from 18 to 96 (M = 57.1, SD = 15.9). Descriptive statistics for age within each BMI category are listed in Table 9.
There was a significant difference in the mean age of each group, F (4, 1699) = 32.9, p < .0001). Post hoc comparisons (Table 10) using the Tukey HSD test indicated that the mean age for normal BMI (M = 60.1, SD = 17.9) was significantly greater than the mean age of the obesity class II (M = 55.2, SD = 15.6) and obesity class III (M = 49.1, SD = 13.9) groups. The mean age of the overweight group (M = 61, SD = 15.1) was also significantly greater from the mean age of the obesity class II and III groups. The mean age of the obesity class III group (M = 49.1, SD = 13.9) was significantly less than all other groups.
Gender. There was a significant relationship between gender and the five BMI groups, X 2 (4, N= 1704) = 57.3, p <.0001. There were a higher proportion of females than males in the obesity class III group, which consisted of 30% male (n= 94) and 70% female (n = 224) ( Table 11). The obesity class III group had a larger proportion of females than all the other BMI groups and the overweight group had a larger proportion of males than all the other BMI groups (Table 12).
Race. There was no relationship between race and the five BMI groups although overall, a vast majority of cases, 86.9% (n=1481) listed a race of "White".
The other race categories made up far less of the total population, with 7.2% "Black", .3% "Asian", and 5.5% "All Other".

Insurance Status.
There was a significant relationship between insurance status and the five BMI groups, X 2 (16, N = 1704) = 33, p = .007. There were a higher proportion of overweight (.22), obesity class I (.23), and obesity class III patients (.2) with private insurance than the normal BMI group (.19) (Table 13).
Pain Scores. Pain scores were only recorded in 48% (n=824) of cases and overall frequencies of mild, moderate, moderate/severe, and severe pain are listed in Table 14. There was a significant relationship between BMI category and moderate pain, X 2 (4, N = 1704) = 12.6, p = .01. The overweight group had a higher proportion of reported moderate pain (.26) than the obesity class II group (.13), t (654) = 3.12, p = .002 (Table 15). There were no other significant differences in moderated pain between the BMI groups. There was no significant relationship between BMI and mild pain. No groups had a pain score of moderate/severe or severe recorded.

Research Question
Research Question One. What is the difference in the total morphine equivalent dose of post-surgical pain medication administered between normal weight, overweight, and obese (Class I, II, and III) adult non-bariatric surgery patients?
Data for the total postoperative day one and postoperative day two oral morphine equivalent dose of pain medication contained several extreme outliers which were removed prior to performing an analysis. These outliers showed dose values approaching 1015 mg. High doses such as these could harm patients and were most likely documented in error. Doses greater than 120 mg were removed because many dosing guidelines recommend a maximum morphine equivalent dose of 120 mg per day (Franklin et al., 2012, Braden et al., 2010. These high doses may have been entered in error, or represented patients with high tolerance to narcotics. A significant difference was not found among the BMI groups related to total doses greater than 120 mg. Descriptive statistics of the study sample after removal of outliers are listed in Table 16. Histograms and Q-Q plots for postoperative day one and day two values demonstrated a positive skew (Figures 1 through 4) and were transformed to approximate a normal distribution. A Box-Cox transformation (Osborne, 2010) was used to obtain normally distributed values for postoperative day one and day two total morphine equivalent dose. The value of lambda that yielded the smallest value for mean square residual was 0 for the postoperative day one total morphine equivalent dose and for the postoperative day two total morphine equivalent dose. Therefore, a natural logarithmic transformation was performed on the two variables. After transformation, both variable histograms and Q-Q plots (Figures 5 through 8) better approximated a normal distribution, although the assumptions of normality and homogeneity of variance were not met. The Shapiro-Wilks and Levine statistics demonstrated a significant difference from normality and from homogeneity of variances and therefore the Brown-Forsythe robust test of means was used determine the difference in mean between the BMI categories (Brown & Forsythe, 1974).
Games-Howell's procedure was used for post hoc multiple comparisons due to the heterogeneity of variances (Keselman & Rogan, 1978). A one-way ANOVA was performed with the log-transformed total postoperative day one oral morphine equivalent dose of pain medication dependent variable and the BMI classification independent variable and there was a significant difference in the mean dose of pain medication given for each of the BMI categories, F (4, 1468) = 2.72, p = 0.03. Games-Howell's post hoc test (Table 17) revealed that the geometric mean of the total postoperative day one oral morphine equivalent dose for overweight individuals was significantly different (p = .004) and 1.22 times as much as for obesity class III individuals (95% CI: 1.05 to 1.42 times as much). While there was not a significant difference with the other groups, there was a downward trend in dose between the overweight group and the obesity class I and II groups (Figure 9). The postoperative dose for the normal BMI group was 1.1 times as much as the obesity class I group, and 1.01 times as much as the obesity class II group. There was not a significant difference in dose between the normal BMI and all other groups.
A one-way ANOVA was performed with the log-transformed total postoperative day two oral morphine equivalent dose of pain medication dependent variable and the BMI classification independent variable. The analysis resulted in no significant difference in mean dose of pain medication given for each of the five BMI categories.

Research Question Two. What is the relationship between pain medication ordered and administered and the BMI of adult non-bariatric surgery patients?
Two-way contingency table analyses were conducted to evaluate if pain medication ordering and administration were related to BMI. There were two separate electronic medical record systems used to for ordering and medication administration and the dose and frequency were not documented in equivalent units between each system. The medication ordering system used "Units" while the medication administration system used "mg". Also, documentation of medication administration tended to be grouped by the nurse. Multiple doses were grouped into one dose that would span an eight-hour shift. Because of this the comparison between dose/frequency ordered and administered to determine if they were related was not possible.
The two variables were pain medication ordered and administered (Not Ordered -Not Administered, Not Ordered -Administered, Ordered -Not Administered, Ordered -Administered) and BMI (Normal, Overweight, Obesity Class I, Obesity Class II, Obesity Class III). Pain medication ordered/administered and BMI were found to be significantly related for hydromorphone, X 2 (8, N=1680) = 23.03, p = .003, morphine, X 2 (12, N=1682) = 46.77, p < .0001, meperidine, X 2 (8, N=1682) = 29.93, p < .0001, oxycodone, X 2 (12, N=1682) = 21.1, p = .05, and acetaminophen, X 2 (12, N=1682) = 49.2, p < .0001. There was a borderline significant relationship with ketorolac, X 2 (8, N = 1682) = 15.22, p = .055. A significant relation was not found between fentanyl ordering/administration and BMI. Follow-up pairwise comparisons were conducted to evaluate the difference among these proportions. Tables 18 and 19 show the number and proportion results of the crosstabulation. Independent sample ttests were performed for crosstabulation table columns. All t-tests had a significant result for Levine's test for equality of variances; therefore the value when equal variances are not assumed was used. The results of the t-tests are listed in Table 20.
While there were significant differences in proportion found, there were none found that suggest that the higher BMI groups consistently had smaller proportions for ordering and administration when compared to the lower BMI groups.
Hydromorphone. In the overweight BMI category, the proportion of hydromorphone ordered and administered (.96) was greater than the proportion not

Research Question Three. What is the relationship between patients' BMI
and the receipt of post-surgical pain medication, when accounting for race, gender, age, insurance status, presence of psychiatric diagnosis, and pain score during hospitalization?
Separate analyses were performed for the total oral morphine equivalent dose during the first 24-hours (postoperative day 1) and for the second 24 hours (postoperative day 2) after surgery. Simple linear regression was used to determine the relationship between the independent variable total postoperative oral morphine equivalent dose and the dependent variable BMI. Hierarchical multiple regression analysis was then performed to determine the relationship after controlling for other factors related to stigma such as age, gender, race, insurance status, psychiatric diagnosis, and also factors related to the amount of pain each patient reported (pain score).

Total postoperative day one morphine equivalent dose. Extreme outliers were
found that had cutoff values for Cook's Distance greater than .0024 (Fox, 1991) and Leverage values greater than .0023 (Montgomery, Peck & Vining, 2012). In examining the outliers, BMI was found to be greater than 151 in four of the cases.
While the weight of these patients appeared to be realistic adult weights, 55 kg -87 kg (121 -192 lbs.), the heights were all measured as less than .724 m (2.37 ft.). Most likely the height values were entered in error and therefore these four values were removed prior to all regression analyses.

Simple linear regression assumptions. A simple linear regression was
performed and the assumption of normality of residuals was not met, as shown in positively skewed histogram ( Figure 10) and non-linear P-P plot ( Figure 11). Also, the assumption of homogeneity of variances (homoscedasticity) was not met as seen by a non-random distribution of points on a scatterplot of studentized residuals against the predicted value ( Figure 12). The variances appeared to increase as a function of the predicted value.
After performing a natural log transformation of the total postoperative day one oral morphine equivalent dose, heteroscedasticity was still apparent after transformation (Figure 13), therefore the independent variable BMI was also The Durbin-Watson statistic was used to determine independence of errors, d = 1.97, which was between 1.5 and 2.5 (Hutcheson & Sofroniou, 1999) and therefore the null hypothesis was not rejected and it could be concluded that the errors were not autocorrelated (Montgomery, Peck & Vining, 2012)  Hierarchical multiple regression. Hierarchical multiple regression was performed to investigate the relationship between the total postoperative day one oral morphine equivalent dose and BMI, after controlling for other factors related to stigma, such as age, gender, race, insurance status, and presence of psychiatric diagnoses, and also controlling for level of pain. accounted for within the sample. The moderate/severe and severe pain independent variables were not used in any of the regression analyses since there were no cases in the sample where the presence of moderate/severe or severe pain was reported.

Hierarchical regression assumptions described with outliers intact.
Preliminary analyses were conducted to ensure no violations of the assumptions of normality of variances, linearity, and homoscedasticity. The assumption of normality of residuals was reasonable based on a normally distributed histogram ( Figure 18) and linear P-P plot ( Figure 19). Evidence of linearity was provided by scatterplots of standardized residuals versus predicted values that demonstrated a random distribution of points that were distributed with a roughly constant variance for the total model ( Figure 20) and each of the continuous independent variables, age and BMI, in partial regression plots (Figures 21 and 22).
Examination of the correlation matrix (Table 22) and the variable inflation factor (VIF) values for the independent variables suggested that multicollinearity was not an issue (Montgomery, Peck & Vining, 2012). There was little to low correlation between the independent variables. VIF values were all less than 10 (1.013 -1.310).
Most Eigenvalues were well above zero; however the values for moderate pain and BMI were close to zero (.042 and .002). The Condition Index for moderate pain was 10.66 suggesting a weak to moderate degree a multicollinearity and the value for the log transformed BMI was above 30 (44.682) suggesting a high degree (Callaghan & Chen, 2008). No evidence of multicollinearity was assumed since only one independent variable had large variance proportions corresponding to each large condition indices.
The bivariate and partial correlations showed small but significant relations to total oral morphine equivalent dose and are shown in Table 23. As can be seen, BMI and age were negatively and significantly correlated, indicating that as BMI and age increase the amount of pain medication on postoperative day one decreased.

Hierarchical regression assumptions described with outliers removed.
Preliminary analyses were conducted to ensure no violations of the assumptions of normality of variances, linearity, and homoscedasticity. The assumption of normality of residuals was reasonable based on a normally distributed histogram ( Figure 23) and linear P-P plot ( Figure 24). Evidence of linearity was provided by scatterplots of standardized residuals versus predicted values that demonstrated a random distribution of points displaying a roughly constant variance around the horizontal line for the total model ( Figure 25) and each of the continuous independent variables, age and BMI, in partial regression plots (Figures 26 and 27).
Examination of the correlation matrix (Table 25) and the variable inflation factor (VIF) values for the independent variables suggested that multicollinearity was not an issue (Montgomery, Peck & Vining, 2012). There was little to low correlation between the independent variables. VIF values were all less than 10 (1.013 -1.291).
Most Eigenvalues were well above zero; however the values for moderate pain and BMI were close to zero (.041 and .002). The Condition Index for moderate pain was 10.68 suggesting a weak to moderate degree a multicollinearity and the value for the log transformed BMI was above 30 (44.581) suggesting a high degree (Callaghan & Chen, 2008). No evidence of multicollinearity was assumed since only one independent variable had large variance proportions corresponding to each large condition indices.
The bivariate and partial correlations showed small but significant relations to total oral morphine equivalent dose and are shown in Table 26. As can be seen, BMI and age were negatively and significantly correlated, indicating that as BMI and age increased the amount of pain medication on postoperative day one decreased. Total postoperative day two morphine equivalent dose. The four extreme outliers found with BMI greater than 151 were also removed in examining the postoperative day two morphine equivalent dose. These were most entered in error and therefore these four values were removed prior to the analysis. The mean BMI after removal of the outliers was 32.8, SD = 9.01, with a minimum BMI of 18.5 and a maximum of 97.3. Fifty-five cases were highly influential and had high leverage with Cook's Distances greater than .0024 and leverage values greater than .018. One extreme outlier was found that had large cut off values for Mahalanobis Distance, p < .001. No clinical significance of the outliers was readily apparent, therefore analyses were run first without removal of outliers and then after outliers had been removed to examine any differences or patterns. The moderate/severe and severe pain independent variables were not used in the regression analysis since there were no cases in the sample where the presence of moderate/severe or severe pain was reported.

Simple linear regression assumptions. A simple linear regression was
performed and the assumption of normality of residuals was not met, as shown in positively skewed histogram ( Figure 28) and non-linear P-P plot ( Figure 29). Also, the assumption of homogeneity of variances was not met as seen by a non-random distribution of points on a scatterplot of studentized residuals against the predicted value ( Figure 30). The variances appeared to decrease as a function of the predicted value.
After performing a natural log transformation of the total postoperative day two oral morphine equivalent dose, heterogeneity of variances was still apparent after transformation (Figure 31), therefore the independent variable BMI was also transformed using a natural logarithmic transformation. After transformation of BMI, the scatterplot of studentized residuals against predicted values (Figured 32) demonstrated a relatively random display of points that were spread fairly constant over the range of values of the total postoperative day two oral morphine equivalent dose provided evidence of homogeneity of variances. Also, the assumption of normality of residuals was reasonable based on a normally distributed histogram ( Figure 33) and linear P-P plot ( Figure 34). The scatterplot of total oral morphine equivalent dose and BMI indicated that the assumption of linearity was reasonable since the points were roughly symmetrical in distribution around the diagonal line ( Figure 35).
The Durbin-Watson statistic was used to determine independence of errors, d = 1.86, which was between 1.5 and 2.5 (Hutcheson & Sofroniou, 1999) and therefore the null hypothesis was not rejected and it could be concluded that the errors were not autocorrelated (Montgomery, Peck & Vining, 2012)  Hierarchical multiple regression. Hierarchical multiple regression was performed to investigate the relationship total postoperative day two oral morphine equivalent dose and BMI, after controlling for other factors related to stigma, such as age, gender, race, insurance status, and presence of psychiatric diagnoses, and also controlling for level of pain. Table 28 contains descriptive statistics for variables used.

Hierarchical regression assumptions described with outliers intact.
Preliminary analyses were conducted to ensure no violations of the assumptions of normality of variances, linearity, and homoscedasticity. The assumption of normality of residuals was reasonable based on a normally distributed histogram ( Figure 36) and linear P-P plot (Figure 37). Evidence of linearity was provided by scatterplots of standardized residuals versus predicted values that demonstrated a random distribution of points that displayed a roughly constant variance around the horizontal line for the total model ( Figure 38) and each of the continuous independent variables, age and BMI, in partial regression plots (Figures 39 and 40).
Examination of the correlation matrix (Table 29) and the variable inflation factor (VIF) values for the independent variables suggested that multicollinearity was not an issue (Montgomery, Peck & Vining, 2012). There was little to low correlation between all of the independent variables. VIF values were all less than 10 (1.019 -1.401) Most Eigenvalues were well above zero; however the values for moderate pain and BMI were close to zero (.039 and .002). The Condition Index for moderate pain was 10.66 suggesting a weak to moderate degree a multicollinearity and the value for the log transformed BMI was above 30 (44.194) suggesting a high degree (Callaghan & Chen, 2008), but multicollinearity was not assumed since only one independent variable had large variance proportions corresponding to the large moderate pain and BMI condition indices.
The bivariate and partial correlations showed no significant relations among the continuous independent variables to total oral morphine equivalent dose and are shown in Table 30.

Hierarchical regression assumptions described with outliers removed.
Preliminary analyses were conducted to ensure no violations of the assumptions of normality of variances, linearity, and homoscedasticity. The assumption of normality of residuals was reasonable based on a normally distributed histogram ( Figure 41) and linear P-P plot (Figure 42). Evidence of linearity was provided by scatterplots of standardized residuals versus predicted values that demonstrated a random distribution of points and a roughly constant variance around the horizontal line for the total model ( Figure 43) and each of the continuous independent variables, age and BMI, in partial regression plots (Figures 44 and 45).
Examination of the correlation matrix (Table 32) and the VIF values for the independent variables suggested that multicollinearity was not an issue (Montgomery, Peck & Vining, 2012). There was little to low correlation between the independent variables and VIF values were all less than 10 (1.012 -1.404) Most Eigenvalues were well above zero; however the values for moderate pain and BMI were close to zero (.038 and .002). The Condition Index for moderate pain was 11.2 suggesting a weak to moderate degree a multicollinearity and the value for the log transformed BMI was above 30 (50.22) suggesting a high degree (Callaghan & Chen, 2008). No evidence of multicollinearity was assumed since only one independent variable had large variance proportions corresponding to each large condition indices.
The bivariate and partial correlations showed small but significant relations to total oral morphine equivalent dose and are shown in Table 33. As can be seen, gender was negatively and significantly correlated, indicating that when the gender was female, the amount of pain medication on postoperative day one decreased. There was a positive and significant correlation between the dose of pain medication and

DISCUSSION AND CONCLUSIONS
Obesity is a stigmatized condition and since other stigmatized groups have been shown to receive less pain medication it was hypothesized that obese individuals would receive less pain medication postoperatively than lower BMI groups. This study set out to explore ordering and administration practices of pain medication between normal, overweight, obesity class I, obesity class II, and obesity class II groups and has identified differences and relationships between groups. The literature on patient outcomes related to pain and obesity stigma is non-existent. The intention of this study was to begin exploration into obesity as a stigmatized condition that affects pain management and answer three questions: 1. What is the difference in the total morphine equivalent dose of postsurgical pain medication administered between normal weight, overweight, and obese (Class I, II, and III) adult non-bariatric surgery patients?
2. What is the relationship between pain medication ordered and administered and the BMI of adult non-bariatric surgery patients?
3. What is the relationship between patients' BMI the receipt of post-surgical pain medication, when accounting for race, gender, age, insurance status, presence of psychiatric diagnosis, and pain score during hospitalization.

Independent Variables
There were differences in age (Tables 9 and 10), gender (Tables 11 and 12), insurance status (Table 13), and pain score (Tables 14 and 15). The differences in age, with the mean age decreasing as BMI increased, was most likely due to the heaviest patients dying earlier than the lighter patients since there numerous diseases associated with an increased weight (e.g., heart disease, cancer, etc) (Peeters et al., 2003).
Another explanation for the decreasing age could have been due to the heavier patients needing to be hospitalized and needing surgery at an earlier age due to the presence of obesity-related diseases. Women are more susceptible to becoming obese and there are approximate three obese women for every two obese men (Wells, Marphatia, Cole & McCoy, 2012) and the higher proportion of obese females supports this. A greater proportion of private insurance in the overweight and obese groups compared to the normal BMI group may be explained by the sampling and types of surgeries performed. The sample included patients who were admitted on the same day as their surgeries and a vast majority of these surgeries are planned and scheduled ahead of time through the surgeons' offices. Patients who schedule elective surgeries may be more likely to have private insurance. Also, another explanation could have been that the heavier patients were also younger and therefore may still have been employed and receiving private insurance through their employer. The findings related to pain scores were important since it showed that the obese III group was not complaining of moderate pain more or less than the normal, obesity class I, and obesity class II groups and there were no differences in mild pain between all groups. This variable may not have added much understanding of stigma since there were no moderate/severe or severe pain scores recorded. In 2013, the healthcare system had multiple locations for which pain score may have been documented, including a paper record. One explanation for the low percentage of pain scores recorded in the medication administration system may have been that they were documented on paper. Also, it is possible that the pain scores recorded in the medication administration system could have been pain scores recorded after the administration of medication. They may have only been reflecting the improvement of pain, but it was not possible to determine this from the data set.

Research Question One
While there was a general downward trend of the mean dose of pain  2015). Studies have described a tendency to perceive higher weights as normal (Tschamler, Conn, Cook & Halterman, 2009, De La O et al., 2009, Johnson, Cooke, Croker & Wardle, 2008. In a study examining perceived discrimination, Carr & Friedmnan (2005) found no difference between normal and overweight groups. In addition to using a normal BMI population as a comparison group, comparing overweight patients to the heaviest patients may also be appropriate. Examining the results from this perspective, there was a significant difference in the dose of pain medication given on postoperative day one between the overweight and obesity class III groups. The overweight group received more pain medication than the obesity class I and II groups and significantly more than the obesity class III group. Since the heaviest patients received less pain medication, stigmatization associated with patient weight may be present. This receipt of less pain medication by the heaviest groups may be as a result to negative attitudes from the healthcare providers. This would fit with Lewis & Van Puymbroeck's (2008) description of discriminatory acts tied to negative attitudes.

Research Question Two
Comparing overall pain medication ordering and administration practices among the five BMI categories was inconclusive in demonstrating clear patterns and differences in proportion between groups. A greater proportion of acetaminophen was ordered and not administered to obesity class III patients than normal and overweight BMI patients, which does support the literature related to less pain medication administration for stigmatized groups, although this may not be a significant finding since acetaminophen is usually not the primary choice for postoperative pain medication. Another interesting finding was that there were a proportion of patients that were administered hydromorphone, morphine, fentanyl, meperidine, oxycodone, and acetaminophen without the medication having been ordered by a licensed independent practitioner. It is possible that these drugs may have been administered to patients as a result of a verbal order that never was entered into the electronic medical record. Finding an overall relationship among all medications that supported bias towards higher weight patients was difficult. This may have been due to other factors, such as most pain medications are ordered on an "as needed" basis and clinicians have different interpretations of the intent of these orders (Gordon et al., 2008). Also, the "habitus" may have differed between postoperative units (Lauzon Clabo, 2008 (Huizinga, 2012). Further study is needed to determine if ordering of pain medication is affected by weight.

Research Question Three
There was a significant relationship found between the total postoperative oral morphine equivalent dose administered on postoperative day one and BMI, while controlling for other variables related to other stigmatizing conditions. BMI was the highest weighted predictor. It has been described that obese patients do not need any more pain medication that normal BMI patients (Patanwala, Holmes & Erstad, 2014), but this study found that there was a decrease in pain medication dose as BMI increased. Age was not as weighted a predictor of receiving less pain medication as being black or being female, although it could be considered comparable to these when accounting for the scale. For example, the gender and racial category could only increase by a unit of one, since the variables were measured as either being black or not and being male or being female. There was a larger range of possible increases with age. Worker's Compensation patients and patients with a psychiatric diagnosis received a much greater proportion of pain medication dose, 90.8% and 23.2% more respectively. This finding may have been due to the characteristics of each group, such as being more demanding for pain medication or having surgical procedures that may be inherently more painful, but it may have also been due to the low amount of cases within the groups. There were only 19 Worker's Compensation patients and 149 patients with a psychiatric diagnosis in the sample. The findings of this study support the hypothesis that BMI is related to the treatment of pain and that increased weight, i.e. obesity, contributed to a lesser dose of pain medication administered. Past studies have linked stigma of race, gender, and age to pain medication administration practices that resulted in the receipt of less pain medication. While there was a low percent of the variance in total morphine equivalent dose explained by regression model (.6%) and a very low correlation between independent and dependent variables, this was understandable when taking into account the many factors in healthcare and human behavior which may impact the dose of medication received by the patient.
BMI and the other independent variables were only a small portion of what may account for the dose of pain medication. The purpose of this study was to understand the relationship between the variables and small but significant relationships were found. These findings are supported by past research and there was persuasive initial evidence that obesity, also a stigmatized attribute, may have impacted the administration of medications for pain when taking into account other factors related to stigmatization.

Limitations
There were several limitations that affected this study. After removal of duplicate cases and cases with missing information, the obesity class II and III groups contained fewer cases than the other groups, 291 and 318 respectively, and therefore those groups did not achieve the needed cases to yield a power of .95 and the possibility of incorrectly rejecting the null hypothesis is higher. Also, the number of patients receiving pain medication on postoperative day two was less than on day one.
There were only 525 patients for which pain medication was documented on postoperative day two. If interested in determining if any differences total dose of pain medication continued past postoperative day one then ensuring that there are more patients who received pain medication on day two are included in the sample would increase the probability of finding a significant result.
Another limitation was that data was not collected through a random sample, but through IS query based on criteria. Sampling from three different hospitals did strengthen the generalizability of this study; however non-random sampling decreases the ability of the results to be generalized to a larger or different population. While differences in independent variables could be explained from the literature, some of these differences may have introduced sampling bias since the sample was not equally balanced. Also there was an unequal sample size among the five categories. The obesity class II and III groups had fewer cases than the other groups due to exclusion and removal.
The use of BMI was another limitation. BMI is a measure of weight status that depends upon a patient's height and weight to determine a value that places them within a category that can be used to flag the patient as being under weight, normal weight, overweight, or obese. A patient may have a very low percentage of body fat, but be considered obese by measurement if they have a large muscle mass combined with a shorter stature. Some patients falling into this category may be present in the sample, but it was impossible to determine this based on the data acquired.
Another limitation was that pain score was recorded in the electronic medical record for only 48% of the cases. This may have been due to the multiple locations that nurses document pain, including both paper and electronic records. Also, the pain scores that were recorded in the medication administration system were categorical, grouping pain scores into categories, and not continuous. This would not precisely capture pain scores reported by each patient. There was not an adequate representation of each patient's pain as there would be if an integrated electronic medical record existed.
Finally, the categorization of race within the electronic medical record was suspect. There were a large proportion of patients classified as "White" within the study. Many of these patients had languages other than English listed as a primary language. "White" therefore was not an accurate representation of race since it appeared that Hispanic, as well as other race designations, were most likely included in the variable category.

Symbolic Interaction
The actions that healthcare providers take toward their patients are based on the meaning that they attribute to them or their conditions. If there is a negative meaning attributed, then care of the patient may be adversely affected. This was described in past studies examined the impact of racial bias on pain medication administration. This study supports the premise that meaning is reflected in action since there was a difference in the amount of pain medication a patient received based on their BMI. As BMI increased, the amount of pain medication administered decreased. Stigma related to obesity may have been a factor. Since a difference was found, further research that incorporates measures of attitude or bias is important in determining the presence and extent of obesity stigma, and the meanings that are held by healthcare providers. The third premise of symbolic interaction, that meanings are assigned and modified through an interpretive process, will be important in developing interventional studies designed to modify meanings that are attached to obesity. The decision to use symbolic interaction, and not critical theory or critical interactionism, as a framework to guide meaning and assumptions was based on the micro-level view of the research. Pain medication administration was viewed at the individual perspective. The assumption that nurses derive meaning from obesity which creates stigma and negatively affects healthcare outcomes, i.e. pain medication administration, was central in this study. Pain medication administration had never been examined in relation to obesity stigma; therefore it was important to first understand if there were any differences in administration based on BMI. Future research could incorporate critical social theory, using a critical interactionist perspective to address the individual/micro level as well as the organizational/societal/macro level.

Implications for Research
The findings of this study did show that there was a difference in the receipt of pain medication depending on BMI, and that as BMI increased the dose of medication decreased. Only an implied association of these findings to stigmatization can be made and more research is needed to strengthen the evidence. Mixed-methods research that combines the measurement of outcomes with a tool to measure attitudes and biases would be valuable in strengthening the theory that obesity stigma negatively affects the treatment of pain. Also, expanding the study to include different patient or healthcare facility types would increase the generalizability of any findings.
Incorporation of qualitative research examining the meaning of obesity and its effect on pain control would contribute to the usefulness of symbolic interaction as an explanatory theory regarding the under treatment of pain in obese patients. Further research using the theory would be useful in developing interventional studies. Studies that use "priming" as an intervention may be effective in reducing stigma/stereotyping that impacts patient care (Burgess, van Ryn, Crowley-Matoka & Malat, 2006).
"Priming" can be described as providing the subject(s) of interest with information that generates a specific attitude desirable by the researcher.

Implications for Education
Prior research involving students has described the presence of weight bias.
Biases may continue on into clinical practice, maintaining an environment where stigma is perpetuated. Educational interventions aimed at students would improve attitudes, decrease bias, and ultimately may improve pain management in obese patients. It would be important to incorporate education designed to decrease obesity stigma into nursing curriculum. Continuing education for clinicians may also decrease obesity stigmatization. Bariatric sensitivity training has been used successfully in decreasing weight bias in nurses (Falker & Sledge, 2001) and may have a positive impact on the treatment of pain.

Implications for Practice
Negative attitudes and bias toward obese patients have been described as often being implicit among healthcare providers.  Oral (Adults and Children >12 yr): 325-650 mg q 4-6 hr or 1 g 3-4 times daily or 1300 mg q 8 hr (not to exceed 4 g or 2.5 g/24 hr in patients with hepatic/renal impairment). Intravenous (Adults and Children ≥13 yr and ≥50 kg): 1000 mg q 6 hr or 650 mg q 4 hr (not to exceed 4 g/day or less than 4 hr dosing interval).
Crosstabulation between BMI and Gender.  Bivariate and partial correlations between log transformed total morphine equivalent dose on postop day 1 and predictor variableshierarchical regression outliers intact.