Influence of body mass index on health complains and life satisfaction

Beyond health concerns, maintaining an optimal weight according to one’s height is crucial for both cognitive and non-cognitive development, as well as for various aspects of life, including social relationships [1, 2]. This is important through life, and it becomes particularly prominent during adolescence, a period marked by strong stigmatization of heavier body sizes and its impact on social acceptance and friendships [3]. This study aims to explore the influence of body mass index (BMI) on health complaints and psychosocial relationships of Spanish teenagers following a causal methodological approach.

Children with excess weight experience poorer physical health. In this regard, childhood obesity is associated with the development of a number of medical complications, such as type 2 diabetes, hypertension, sleep apnea or cholesterol disorders [4]. All these health issues manifest as musculoskeletal, neurological and gastrointestinal pains [5] and lead to higher healthcare expenses compared to children with a healthy weight [6].

In addition to adverse effects on health status, being overweight is associated with lower life satisfaction [7]. Firstly, somatic complaints reduce the quality of life of children who are overweight [8]. Secondly, overweight children are also more likely to suffer from depression and anxiety disorders [9], which are also negatively associated with life satisfaction [10]. A third channel mediating the relationship between life satisfaction and overweight is social bias. Research indicates that overweight children are more frequently targets of bullying due to factors like social marginalization, lower self-esteem and body dissatisfaction, among others [11‐14]. In fact, adolescents themselves perceive weight status as one of the main reasons to be bullied by their peers [15], which underscores the prevalence of weight-based stigma among teenagers [16]. This discomfort in the learning environment may explain why overweight children are more likely to be absent from school and tend to have a lower academic performance compared to their normal-weight peers [17‐19]. Due to the associated risks of excess weight and its escalating prevalence, childhood overweight and obesity are considered as a worldwide epidemic in modern society [20].

The percentage of children with excess weight is particularly alarming in Spain, where 40.6% of children¹ fall into the overweight or obese category. This figure surpassed that of neighbouring countries, with rates of 16.5% in France, 29.6% in Portugal and 29.8% in Italy. Specifically, Spain is ranked third among the European countries with more overweight children [21]. This is surprising, given that Spain encompasses the Mediterranean diet, argued as one of the healthiest [22]. In this context, it is especially interesting to deepen the understanding of the consequences of childhood weight on perceived health (psychosomatic complaints) and school life in Spain. Therefore, our first aim is to provide new evidence on this issue. Specifically, we use the data from the “Health Behaviour in School-aged Children” for Spain—2014 (the last survey publicly available); using this dataset, we investigate how body mass index conditions students’ perceived health and school life, including violence, bullying issues and peer support.

Yet, despite the widespread interest in the consequences of weight status, relatively few works have tried to mitigate the endogenous part of weight status [23]. In this regard, children’s weight status can be endogenous since it is associated with socioeconomic conditions and individual’s surroundings [24] and these variables are determinants of some outcomes explored in this study, such as life satisfaction. Moreover, the dependencies explored herein could suffer from reverse causality problems. Let us give an example: Do children feel low because they are overweight, or could it be that feeling low predisposes them to gain weight? We found a limited number of studies which have tried to control the biases produced by such confounding problems of weight status on diverse outcomes. Previous research focusing on children's development have used the weight of a biological relative [25‐28], genetic markers [28], child’s height [29] and past weight status of child [28, 30] as instrument for child's weight.

In light of these methodological considerations, the evidence provided herein contributes to the existing literature in two significant ways. First, our research implements an instrumental variable procedure, using as instrument the frequency of alcohol consumption. While previous literature has frequently addressed the association between body mass index or overweight/obesity status, few studies have gone a step further and try to get close to a causal effect. The instrumental variable approach allows us to obtain consistent estimates of the influence of body mass index. Second, we provide new empirical evidence about the influence of body mass index on health complaints and school-related factors in Spain. Although some research studies have explored the consequences of weight status in the Spanish case (see [31, 32], to the best of our knowledge, there are no previous large-scale studies which have systematically investigated the link between psychosomatic complaints and the body mass index in this country.²

This research is based on the Spanish records from the 2013–14 Health Behaviour in School-Aged Children Survey (HBSC). The HBSC is a collaborative study coordinated and sponsored by the World Health Organization (WHO), carried out each 4 years. The goal is to collect information about adolescents’ health. According to the WHO, “health is a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”.³ Following this definition, they collect data related to physical health, well-being, and social environment at home and at school, food and diet, substance use and physical activity. The survey consisted of an online questionnaire completed by adolescences aged 11, 13 and 15 years old in the classroom setting. To obtain a representative sample of the Spanish population of these ages, participants were recruited using a multi-stage random sampling stratified by conglomerates, taking into account the age, the Autonomous Community and the ownership of the school (public or private) [34]. The Spanish data were collected between March and December 2014.

The number of students participating in Spain was 11,136. To be eligible for the present research, students must have answered the questions related to both height and weight. In particular, around 86% of the sample answer both questions; this left a total of 9,565 students for consideration in the current analysis. Table A1 (Appendix) shows the descriptive statistics of the variables employed in the research. We have analysed if there are any differences between the sample used and the sample excluded by performing a test of mean differences and we have identified significant differences in the higher frequency of feeling psychosomatic complaints, high levels of life satisfaction, high participation in fights and bullying outcomes, and in the set of variables which measure the socioeconomic status (Table A1, Appendix). This should be taken into account when interpreting the results. Besides that, missing answers in the dependent variables may reduce the sample.

Three sets of outcomes are explored in this study: subjective health complaints and life satisfaction, fighting and bullying, and peers’ support. Table A2 (Appendix) summarises the outcome variables—the exact wording of the questions can be seen at Table A1 (Appendix).

The main variable of interest in this research is the body mass index (BMI henceforth). This variable is a measure used to classify the population into severe thinness, thinness, normal weight, overweight and obesity, and it is derived from the relationship between height and weight:

Both weight and height are self-reported. We intend to identify empirically the effect of body mass index on health and school outcomes. To begin, we estimate the following ordered probit model:where \({O}_{i}^{*}\) is a latent outcome variable, further \({O}_{i}^{*}=k\) if \({d}_{k-1}<{O}_{i}^{*}<{d}_{k}\). The different outcomes are:

The explanatory variables can be classified as:

A summary of the set of control variables can be found in Table A3 (Appendix). Since there are observations with missing socioeconomic and cultural characteristics, we replace those missing values with “0” and we add an indicator variable (i.e. missing flag variable). By doing so, we can maximize the number of observations in the sample; \({u}_{i}\) is the idiosyncratic error term.

The estimated \(\beta\) coefficient captures the influence of one point increase of student’s BMI on the outcome of interest, controlling for student’s characteristics. However, this estimation may suffer from bias due to the unobserved factors that may be correlated with BMI and the outcomes under studied. For instance, greater fast food intakes, which lead to gain weight, have been found to be associated with lower life satisfaction [35]. Besides that, around 50% of body mass index variation is due to individual choices and environment, which suggests that BMI is not exogenous [36]. To solve this, we employ the instrumental variable (IV) methodology to try to obtain the causal effect of BMI on the set of health and school outcomes. The IV approach has been used by many authors to solve the endogeneity problem of BMI (for instance [37‐39]).

We attempt to address the potential endogeneity of BMI by using the frequency of alcohol consumption as instrumental variable (\({Z}_{i}\)). In particular, the question that students’ answer in the HBSC survey is: “On how many days (if any) have you drink alcohol in the last 30 days?”, with seven possible answers (never, 1–2 days, 3–5 days, 6–9 days, 10–19 days, 20–29 days and 30 days or more). This categorical variable has been recoded as a set of binary variables.

To ensure the proper application of this methodology, the instrument must satisfy a number of conditions. In Appendix B, we detail the relevant properties as well as their application to this context.

Due to the categorical nature of the outcomes, we do not directly apply the two-stage least squares (2SLS) estimation to fit the data. Instead, the model is estimated using a conditional (recursive) mixed process estimator. The base model is composed of two equations:where \({Z}_{i}\) is the frequency of alcohol consumption, \({\varepsilon }_{i}\) and \({\omega }_{i}\) are idiosyncratic random error terms for each equation. Equation (3) would be some kind of first stage in 2SLS, while Eq. (2) would be reduced form. The dependent variable in Eq. (2) is an ordered outcome and it is estimated using ordered probit model. The dependent variable in Eq. (3) is continuous and the model is estimated using Ordinary Least Squares (OLS). The inference using this structural equation model exploits the presence of an instrumental variable through its inclusion only in Eq. (3), which is correlated with the outcome only through its effect on the BMI. This is implemented using the Stata command cmp [40], which uses a canned routine to assure that standard errors are correctly estimated.

First, we estimate the influence of BMI on health outcomes, bullying and peer support assuming that BMI is exogenous—Eq. (1) in the methodology section. Second, we implement an instrumental variables (IV) estimator that addresses the concern about the endogeneity of BMI—Eqs. (2 and 3) in the methodology section. The full set of estimates is presented in the Appendix (Tables A4, A5 and A6). Given that probit coefficients have not a meaningful interpretation far from the sign, we have computed average marginal effects to know how large and important differences are. Tables 1, 2, 3, 4, 5 and 6 show average marginal effects for the key variable (BMI). Each category of the dependent variables requires a separate estimation, which will be denoted by k. For instance, “headache” will present \(k=0, 1, 2, 3, 4\) marginal effects estimations. Tables also report the standard errors, which can be used to calculate the confidence intervals at a specific significance level:where \(\widehat{\beta }\) is the estimated average coefficient;\(\alpha\) is the significance level; \(SE\) is the standard error; and \({z}_{1-\frac{\alpha }{2}}\) is the percentile of the normal distribution.

The empirical evidence confirms that BMI conditions the daily life of Spanish students. Firstly, we found that BMI is associated with more psychosomatic complaints, such as backaches, headaches and sleeping disorders. This positive association between BMI and psychosomatic complaints had been described in other countries like Finland [41], Germany [42] or Sweden [43]. In addition to this, an increase of BMI is associated with lower life satisfaction; this effect is within the trend observed in western countries, in which excess of weight is regarded negatively, and hence reduces subjective well-being [44].

Secondly, BMI increases the frequency of participating in physical fights and bullying other students at school. Conversely to the previous studies for Spain [31, 32], we have found that BMI is not significantly associated with the risk of bullying victimization at school (face-to-face bullying). However, our results note that BMI is significantly associated with the bullying developed in the online realm (cyberbullying). This entails that overweight students suffer more verbal than physical abuse [45]. According to Kowalski et al. [46], the perceived anonymity of the aggressor and the higher accessibility are adverse effects of online bullying compared to face-to-face bullying, which justify its expansion.

The specific organization of a school day in Spain further illuminates the notable association of bullying victimization online and its absence during regular school hours. In Spain, the majority of schools follow a condensed schedule, with only morning sessions [47]. Consequently, the opportunities for in-person interactions out of the classroom are limited to the school break which is short, while they persist or even increase out-of-school, given the time that children spend connected to the Internet. In fact, Spanish adolescents spend most of their free time using the mobile phone, tablet or computers [48]. On average, children between 9 and 16 years old connect to the Internet more than 3 h a day in Spain, mainly for activities related to communication and entertainment [49].

Lastly, results show that the increase of BMI makes more difficult to feel peer support. Specifically, a weight gain reduces the chances of feeling friend’s help and counting on friends. The social isolation experienced by overweight students might be based on the weight stigma, which acts as a foundation for social disapproval: “overweight or obesity are negatively stereotyped as being lazy, lacking willpower and self-discipline, unmotivated to improve their health” (p. 402, [50]. An underlying driver of the lower social support might be the constant exposure to unrealistic beauty standards prevalent on the Internet. In particular, Calado et al. [51] highlight a significant association between exposure kinds of media topics related to body image—such as dieting, fashion, and fitness—and body dissatisfaction, which affects more to female than male.

Policy interventions resulting from these findings may be driven in different ways. On the one hand, they should be focused on helping children to achieve a healthy BMI. As it happens in other countries, such as the United States, health assessment at school can be a useful tool to inform parents and children about healthy weight ranges. In Spain, given that medical check-ups are available through the National Health System, there is no health assessment at school. However, we find it relevant to develop this action at school, since it ensures to reach the target group.

On the other hand, the high prevalence of overweight among Spanish youth may be attributed to their lifestyle. Research by Grosso and Galvano [52] demonstrated a low adherence to the Mediterranean diet among children and adolescents in Spain and a decreasing trend in this adherence. This relationship was mediated by various social and demographic factors; notably, individuals from socioeconomically advantaged backgrounds tended to exhibit higher adherence to the diet. In this context, a second recommendation is to implement nutrition education programs at school to promote healthy eating habits from a young age, paying special attention to low-income schools.

In addition, this study alerts to higher discrimination as a result of the increase of BMI, particularly in online environments. Consequently, policy actions should be targeted at dealing with the social stigma of overweight. Cultivating a culture of non-discrimination within school is a challenge that involves the entire educational community. In the classroom, there are various approaches, like discussing and identifying weight stereotypes, designing mechanisms to monitor incidents of discrimination or involving students in peer mediation activities. Regarding the role of parents, they should pay particular attention to the time that children spend on the Internet using social networks which, together with an inadequate use, may foster cyberbullying. In this sense, enhancing parents’ awareness of children’s use of social networks can help prevent cyberbullying. A practical approach to monitoring online activity is to use parental control apps.

Some limitations apply to the present study. First of all, the use of IV tries to deal with the endogeneity problem of BMI, but there might be other unobservable variables that we cannot control for. For instance, prior research indicates a significant correlation between adolescent life satisfaction and family structure [53], a relationship that we currently lack information on. Secondly, it is worth noting that all measures were self-reported. While this approach is advisable for some variables, for others the results should be interpreted with caution. For example, when it comes to assessing bullying comprehensively, anonymous self-reports tend to be the most reliable, as school personnel and classmates may not always be aware of every instance [54]. However, when it comes to self-reporting weight and height, there can be a risk of misclassification. Previous investigations indicate a pattern of underreporting for weight, along with over-reporting for height, which leads to a lower BMI and tends to underestimate the prevalence of overweight and obesity [55].

Further research that incorporates objective measurements of height and weight is needed to investigate the impact of BMI on experiences of bullying and peer support in a more accurate way. Additionally, it is important to note that the majority of studies examining the relationship between BMI and the school context are correlational. While this research takes a step towards establishing causality, more causal statements are needed. To better address these questions, the implementation of longitudinal assessments throughout the school period would be particularly valuable.