Identifying the ‘red flags’ for unhealthy weight control among adolescents: Findings from an item response theory analysis of a national survey

Weight control behaviors are common among adolescents. Data from the 2009 Youth Risk Behavior Survey suggest that more than 60% of adolescents exercised for weight loss in the 30 days before the survey and 4% vomited, 5% took diet pills, and 10% went without eating [1]. In a nationally representative sample of adolescents in New Zealand, approximately two-thirds of females and one-third of males had attempted weight loss in the previous year with concerning proportions of young people fasting, vomiting, or smoking more cigarettes for weight loss [2]. Given the increasing prevalence of overweight and obese children and young people, weight control behaviors may be indicated for some obese/overweight adolescents as long as they are healthy weight control strategies (e.g. eating less junk food).

Unfortunately unhealthy weight control behaviors are common, especially among overweight young people, and there is abundant evidence that unhealthy weight loss behaviors lead to poor outcomes for adolescents. For example, adolescents who used diet pills, vomited, took laxatives, took diuretics, fasted, used food substitutes, skipped meals or smoked cigarettes for weight loss had high levels of depression and were more likely to develop suicidal behaviors into young adulthood [3]. Likewise, other studies have found that adolescents who fasted, used food substitutes, skipped meals, smoked cigarettes, took diet pills, vomited, took laxatives or used diuretics had poorer nutrient intakes [4] and adolescents who used more extreme weight control behaviors (defined as vomiting and/or taking diet pills) ate fewer fruits and vegetables and ate more high fat foods [5] than those who were not dieting or using only moderate weight control behaviors (defined as anything other than vomiting or diet pills). Furthermore, it has been reported that approximately 20% of severe dieters (defined as the top third of students reporting calorie counting, reducing food, and skipping meal strategies) go on to develop a new eating disorder [6]. In general, adolescents who diet (change how they eat in order to lose weight) are more likely to develop binge eating behaviors and gain more weight [7] over time, compared to those who do not.

Given the pervasiveness of weight control behaviors among young people and the potential for adverse outcomes, it is concerning that clinicians rarely ask young people about these behaviors or conduct routine screening of adolescents about weight control behaviors [8]. This may be due to uncertainty about which questions to ask, specifically around what weight loss behaviors are more likely to lead to adverse outcomes and whether certain weight loss strategies are healthy or unhealthy (e.g. skipping meals). It also may reflect a lack of awareness that overweight and obese young people are more likely to use unhealthy weight control strategies than their normal weight peers [9, 10].

One of the challenges that health care professionals face is limited time to ask the most important questions on all aspects of a young person’s life. The standard protocol for interviewing adolescents in a healthcare setting includes discussion on numerous aspects of the lives of adolescents, including home, school, eating, friends, drugs, sexuality, depression, and safety [11]. While comprehensively assessing body satisfaction, perceived weight and eating and exercise behaviors systematically may give an indication of a young person’s risk, clinicians may not always have the time to fully enquire about every possible of weight loss strategy. Therefore it would be useful for clinicians to better understand which weight loss behaviors are healthy or unhealthy, which unhealthy weight loss behaviors are more severe and require urgent attention, and which weight loss behaviors are less severe but signify ‘red flags’ as unhealthy weight control strategies.

Item response theory (IRT) analyses of weight control behaviors can provide clinicians with information on how various weight loss behaviors relate to an underlying continuum of risk, from less severe to more severe behaviors and therefore a better understanding of which behaviors place young people at higher risk of harm. IRT can be used to identify the validity of underlying latent constructs of behavior (e.g. unhealthy weight control, healthier weight control) by relating these continuums to other outcomes such as mental health concerns. IRT analyses produce two main parameters of interest to clinicians: item thresholds and factor loadings. The item thresholds parameters correspond to item difficulty and reflect the endorse-ability of a behavior (e.g. vomiting) along the continuum of the latent construct (e.g. unhealthy weight control). Item factor loadings identify those behaviors which are more accurately able to discriminate students along the continuums of the latent constructs.

Item response theory analyses also provide information of possible systematic biases or measurement equivalence between demographic groups [12]. These arise when individuals with equivalent levels of underlying risk from unhealthy weight loss behaviors respond differently to specific weight loss behaviors depending on their demographic characteristics. For example, males and females may have differing levels of risk from endorsing the same weight loss behavior. IRT analyses that assess measurement equivalence between demographic groups are important for clinicians to better understand how weight control behaviors may differ in their meaning between demographic groups and to lessen the chance of unbiased diagnoses.

The specific objectives of the current study are: 1. to determine which weight loss behaviors are on the unhealthy and healthy continuums of weight control, 2. to identify weight loss behaviors that are ‘severe’ (or difficult) and may require urgent attention and behaviors that are less difficult, but are the ‘red flags’ signifying unhealthy weight loss strategies, 3. to explore the association of unhealthy and healthy weight control with symptoms of depression and wellbeing among students, and 4. to examine how weight control may vary among demographic groups and which demographic groups are engaging with healthy and unhealthy weight control behaviors.

Data for the current study were collected as part of Youth’07, a national survey of the health and well-being of New Zealand secondary school students (approximate ages 13-18). Full details of the methodology and survey design of the Youth’07 survey are described elsewhere [13] with a brief description outlined below. Data for the Youth’07 survey were collected in 2007.

Participating students were randomly selected through a two-stage clustered sampling design. First, 115 schools were randomly selected for participation then students were randomly selected for participation from the school roll. In total, 9,107 students agreed to participate in the survey. The final response rate for schools was 84% and for students 74%. Among the most common reasons for students not participating were being absent from school or being unavailable on the day of the survey [13].

School principals consented to participation in the survey on behalf of the Boards of Trustees. Selected students and their parents were provided with information sheets about the survey. Students consented themselves to participate in the study on the day of the survey. The University of Auckland Human Subject Ethics Committee granted ethical approval for the study.

All data were collected during the school day. On the day of the survey, students arrived at a designated room where they were given an anonymous login code to access the survey. The Youth’07 survey included a 622 item multimedia questionnaire administered on an internet tablet, anthropometric measurements, and identification of their census mesh block number (based on their residential address) to determine small-area neighborhood deprivation.

Drawing on a large, nationally representative population of adolescents, the current study has been able to confirm that there are two underlying constructs of weight control behaviors which are consistent with what previous literature has referred to as healthy and unhealthy weight control [4, 5, 9, 10, 22, 23]. The unhealthy construct in the current analyses included behaviors such as vomiting, using diet pills, smoking cigarettes, fasting and skipping meals. The current study also demonstrated that unhealthy weight control is positively associated with depression. This is consistent with previous research which has suggested that using unhealthy weight control behaviors is associated with poor outcomes in terms of their nutritional status and emotional wellbeing [3, 4, 24].

We found that skipping meals and fasting were among the most common unhealthy weight loss behaviors and had the lowest item difficulty on the unhealthy weight control continuum. This indicates that these behaviors may warrant discussion in routine clinical assessments and may be appropriate to use as screening questions for unhealthy weight control behaviors. In the current sample, skipping meals was a commonly used weight loss strategy with 15% of all students and 30% of students who have attempted weight loss in the past year reporting the behavior [2]. Skipping meals is a commonly reported weight loss strategy among American adolescents as well, with previous studies estimating that 20-30% of adolescents in the general population report this behavior [25, 26]. Furthermore, skipping meals is not an effective weight loss strategy as available research on successful weight loss suggests that regular meal consumption, particularly breakfast, is important for weight loss maintenance. [27, 28] Thus, regular screening for fasting and skipping meals specifically, may be as quick and useful marker for other risk behaviors, including more unhealthy weight control.

In the current study, among the more difficult indicators of the unhealthy weight control construct were use of diet pills, smoking for weight loss and vomiting. Some previous research has referred to these types of weight control behaviors as extreme weight control behaviors and demonstrated that young people who use these weight control behaviors are at risk for multiple poor outcomes. For example, Story et al. [24] found that young people who had vomited or used diet pills for weight loss in the past week were less likely to eat fruits and vegetables and more likely to consume high fat foods. Similarly, Neumark-Sztainer et al. [29] demonstrated that young people who had vomited or used diet pills for weight loss were more likely to engage in a number of risk-taking behaviors, such as substance use and unprotected sexual activity. More recently, Crow et al. [3] observed that young female adolescents who used diet pills, laxatives, diuretics or vomited for weight loss were significantly more likely to report suicidal attempts and suicidal ideation into young adulthood than those who did not.

It was of interest that two behaviors, counting calories and eating fewer carbohydrates, loaded positively onto both the healthy and unhealthy weight control constructs and that these were more difficult behaviors on both constructs. Thus, when young people report the use of these behaviors for weight control in clinical settings, clinicians need to discuss the range of other weight control behaviors students are engaging in to be able to determine how healthy their weight loss strategies are.

The healthy weight control construct identified in the current study is consistent with recognized weight control recommendations encouraging dietary improvements and exercise, [30] such as eating more vegetables and fewer sweets. Yet there is mixed evidence to suggest that adolescents who engage in weight control behaviors commonly described as healthy weight control achieve better health and wellbeing [7, 23–25, 27]. It is of interest, then, that we observed only a slight relationship between healthy weight control behaviors and positive emotional wellbeing. As discussed above, the bulk of the previous research has looked at the relationship between emotional wellbeing and unhealthy weight control behaviors rather than healthy weight control behaviors.

In the current study, we found evidence of differential item functioning suggesting measurement non-equivalence or item response bias across demographic groups. It is known that the prevalence of weight loss behaviors varies by age, gender and ethnicity among New Zealand young people [2], and internationally as well [31]. The varying prevalence of these behaviors by demographic groups and differential item functioning found in the current study suggests that there may be differing underlying cultural and gendered meanings of these behaviors and weight control more generally. One of the limitations of the current study is that our analysis of differential item functioning only concerns non-invariance of threshold parameters and not factor loadings across groups. While we were able to model some of the more significant direct effects to account for non-invariance of item threshold across groups, it is likely that non-invariance of factor loadings between groups also exists. However after the addition of direct effects the relationship between covariates and the healthy and unhealthy latent factors did not change substantially.

While the comparison of latent factor mean values between groups relies on the assumption of measurement invariance, our results indicated that young people who were female or of Maori or Pacific ethnicities were more likely to endorse more difficult weight control behaviors on the unhealthy weight control construct. Thus, these young people who report weight control behaviors may warrant extra clinical attention and monitoring to prevent poor outcomes associated with unhealthy weight control.

The strengths of the current study lie in its large, nationally representative sample of young people, well validated measures of both depression and positive mood, and use of analytic techniques to confirm what previous researchers have hypothesized. That said, there are a few limitations to the current study that warrant consideration in its interpretation. First, while our study is generalizable to secondary school students in New Zealand, this may not be true for young people in other parts of the world. This is particularly true for the weight control behaviors included in the current study as they may not be inclusive of the most commonly employed weight control behaviors used by young people in other regions of the world. For example, our study did not include items on diuretics or laxatives, yet these behaviors are reported by approximately 2% of young people in the US. [32] The current study is also limited in its depth of questioning around the weight control behaviors. For example, more information on frequency of weight control behaviors and the effects of the unhealthy weight control strategies would add depth to our understanding of the issue, but was not easily addressed in this large, broad health survey.

Weight control behaviors are common among young people and unhealthy weight control behaviors pose unique risks to their wellbeing. Routine assessment of weight control strategies by clinicians are warranted, particularly for screening for meal skipping and fasting for weight loss as these behaviors appear to ‘flag’ strategies that are associated with poor mental wellbeing.

JU conceived the research questions, interpreted the data, and drafted the manuscript. SD and ER acquired the data, conducted and oversaw data analyses, and revised the manuscript for intellectual content. SA and SC acquired the data and revised the manuscript for intellectual content. All authors read and approved the final version of the manuscript.