Journal of Public Health

, Volume 20, Issue 2, pp 111–124

Familial and societal causes of juvenile obesity—a qualitative model on obesity development and prevention in socially disadvantaged children and adolescents

Authors

  • Wolfgang Weimer-Jehle
    • ZIRN, University of Stuttgart
  • Jürgen Deuschle
    • ZIRN, University of Stuttgart
    • Institute of SociologyLeibniz University Hanover
    • KATALYSE Institute for applied environmental Research
Original Article

DOI: 10.1007/s10389-011-0473-8

Cite this article as:
Weimer-Jehle, W., Deuschle, J. & Rehaag, R. J Public Health (2012) 20: 111. doi:10.1007/s10389-011-0473-8
  • 507 Views

Abstract

Aim

The issue of excess weight and obesity among our young people is currently under discussion as one of the most serious problems in public health. Extensive work has been done to analyse the problem, to indicate the drivers, and to create prevention programmes. Much research remains to be done in the field of modelling the complex impact network of familial and societal influences on juvenile obesity. To achieve this, the forecasts and results issued by the various disciplines must be integrated. The aim of our work has been to create a causal-loop model of obesity in socially disadvantaged children and adolescents that allows qualitative simulation, group-specific risk assessment, as well as the identification of key factors for prevention.

Subjects and Methods

The model was created in cooperation with 18 experts from the field of obesity research. The participants were drawn from eight different disciplines including medicine, sociology, and prevention. Four expert workshops pinpointed 43 main obesity drivers at the individual, familial, and societal level; these were rated according to their causal interdependence and impact. The computer-based method of cross-impact balance analysis was used to evaluate the model and to produce risk profiles for different societal and individual context situations.

Results

The model analysis reveals that there is no one single key factor that can be expected to act as an effective prevention factor for every scenario. Instead, both the risks and the effectiveness of prevention measures depend strongly on the specific characteristics of an individual’s own environment.

Conclusion

Consequently, it would appear sensible to approach the design of prevention programmes from a group-specific, multi-factor and multi-level perspective.

Keywords

ComplexityQualitative modellingHealth promotionJuvenile overweight/obesityPreventionSocial disadvantageSystemic implicationsSystems analysis

Introduction

In recent years juvenile obesity has been the subject of more public attention than practically any other aspect of public health. Frequently afforded primetime mass media coverage, obesity among children and adolescents is regarded as a social problem in all industrialised nations (cf. OECD 2010; WHO 2009). The efforts being made to reverse this trend are enormous. Among scientists, the topic is a favourite focus of disciplinary and interdisciplinary research projects, not to mention a whole series of journals. Bearing the title 'IN FORM—German national initiative to promote healthy diets and physical activity', a national action plan has been launched by protagonists of the Nutrition and Health Field of Action (cf. BMELV—German Federal Ministry of Nutrition, Agriculture and Consumer Protection, and BMG—German Ministry of Health 2008) with the aim of preventing poor nutrition, a lack of exercise, obesity, and related diseases. On the one hand the problem of obesity has become a lucrative business for those marketing prevention. On the other there are signs that the obese are mobilising and beginning to challenge their exclusion from society. To date however such action remains isolated, lacking the power to influence the public discourse and the social stereotypes about fat people propagated therein. There is scarcely a single social subsystem that does not—subject to its own specific characteristics—categorise obesity as a problem.

Diseases directly linked to nutrition and exercise such as obesity play an essential role in the ‘prevention research’ funding programme sponsored by Germany's Federal Ministry for Education and Research—BMBF. In addition to conventional modes of approach (investigation into disease causes and occurrence, diagnosis, treatment and health care research), this research funding also covers social-ecological and psychological research (Hausdorf 2009); in 2008 the programme established the medical Competence Network of Obesity Research. Moreover, obesity prevention forms the focus not only of the German Ministry of Health’s funding programme ‘Alliances for healthy lifestyles and living conditions’, but also of a whole range of health promotion campaigns and projects (e.g. the Kinderleicht regions) run by the German Federal Ministry of Nutrition, Agriculture, and Consumer Protection.

The manifold efforts of the scientific community and the health care system have nevertheless so far failed to have the desired effect. Surveys confirm there is a strong link between being overweight and/or obese and social status: boys and girls from socially disadvantaged families are three times as likely to be obese as children from families with a high social status (BzgA 2010). The link between socio-economic factors and obesity has formed the focus of numerous investigations involving various different countries; in the process education has been pinpointed as the inequality indicator (e.g. income-related factors, education, professional status, employment, residential district) having the greatest influence (Knesebeck 2009).

Despite extensive research, however, it still has not been possible to recognise the main determinants, understand the increasing spread of the disease, or develop sustainable and successful prevention strategies. To date meta-analyses and reviews have not produced any criteria for success supported by hard evidence (cf. BzgA 2010; Kremers et al. 2009; Müller 2009; Hebebrand et al. 2005; Roth et al. 2002, Campbell et al. 2001; Pudel and Westenhöfer 1991). This, despite the fact that research into the subject is no longer confined to the medical sector—other disciplines have since discovered it as well. Research into influencing factors now addresses more than simply an individual’s predisposition towards the disease and his or her eating and exercise habits. Since the 1980s there has been an increasing focus on familial and societal conditions (cf. Neumark-Sztainer 1999). The huge significance of these environmental factors is an aspect that figures in almost every publication on obesity.

One result of the intensification of interdisciplinary research has been an increasing appreciation of the complexity of obesity. Typical discussion points are
  • the increasing number of factors understood to influence the occurrence of obesity, as well as the moderator variables relevant to the context and

  • the indirect relationship between these factors and the energy balance.

The “never-ending stories about complexity” bemoaned by Manfred Müller (2009) have sparked calls for a systematic integration of the insights acquired by the various disciplines (cf. Schneider and Hoffmann 2011; Hebebrand et al. 2004; Willich et al. 2001). What is needed is a meta model that contributes to a better understanding of the complex interactions, thus effecting an improvement in prevention measures. This is the opinion of Maziak et al. (2008): “Understanding these complex interactions is important for the formulation of rational research and intervention strategies that will address the needs and circumstances of different groups of the society” (Maziak et al. 2008). Manfred Müller is even more direct: “We presently do not have a unifying theory to explain population obesity and a so-called meta model has to be developed as a next step. (…) Since prevention strategies cannot work without an adequate model I consider this as a sine qua non for future prevention programmes” (Müller 2009).

In recent years impact models on obesity have been produced by a number of different research groups (cf. Schneider and Hoffmann 2011; Butland et al. 2007; Davison and Birch 2001). Despite all the differences in the methods used, the models’ various scope, the reasoning given for the selection of factors, and the assumed causal relationships, these projects have had one thing in common: the attempt to collate current research in its sheer complexity. These models may be appreciated as the systematic visualisation of the current state of research. They do not however provide a systematic simulation of the causal mechanisms. They are capable—with the help of thought experiments, for instance—of providing hints on system suggestibility and effective prevention starting points. Nevertheless it is not possible in this way to illustrate the manifold system states generated by preventive intervention within such a complex network of interactions. What is needed is a suitable formal method of analysis. After all, bearing in mind the virtually incalculable number of possible system states associated with the multiple factors, their range of possible states and varying degrees of causal relationships, cognitive constraints are bound to limit the success of any thought experiment.

During the empirical study conducted in consultation with 18 experts, we addressed the identification of relevant influencing factors, the visualisation of the interplay between obesity drivers and inhibitors, the systematic simulation of their relationships, and finally the development of approaches to identify suitable levers for prevention. We used cross-impact balance (CIB) analysis to assess the interaction network. This is a formal method for analysing qualitatively formulated networks, implemented for the first time within the context of our research into juvenile obesity. The use of a formal, computer-assisted method of analysis made it possible to systematically record and analyse the multitude of potential interactions, even in extensive networks, and to develop a model that would go beyond the current state of research in the modelling of juvenile obesity, with the potential to provide inspiration for the design of effective prevention measures.

Research questions

Our research addressed the following questions
  • What are the principal factors of a juvenile obesity causal model?

  • How are these factors related?

  • How can an understanding of such interaction be used to identify group-specific approaches for preventive measures?

Background

Our study is part of the project ‘Participative development of obesity prevention concepts for socially disadvantaged children and adolescents—target group-specific strategies for strengthening the health-related resources of nutritional and exercise responsibility in consideration of systemic implications’, which is funded by the Federal Ministry for Education and Research as part of its prevention research programme (‘Health Research: Scientific Research for the People’) and coordinated by Regine Rehaag (for more information see http://adipositas.katalyse.de).

The aim of the project is to reconstruct both the experience of being fat as seen by overweight socially disadvantaged children and adolescents, and possible explanations about why people become fat as presented by experts, with the idea of channelling these findings into prevention programmes. The method consists of two modes of approach: the target group perspective and the expert perspective.

Target group perspective—group discussion as a qualitative approach

Group discussions—Dicke Freunde (German pun: ‘dick’ meaning ‘best’ and ‘fat’)—were held with socially disadvantaged: in summer 2009 with overweight children and adolescents, and in spring 2010 with the parents of this target group. The aim was to understand how overweight socially disadvantaged individuals experience and shape their everyday lives, what health strategies they pursue, and what obstacles, (exclusion) experiences, coping strategies, and desires emerge from the narratives shared. The group discussions were organised according to gender, age (11–13 and 14–16 year olds), and cultural background (German or Turkish origin). A total of 60 juveniles took part in the discussions. Additional information was gained from group discussions involving the parents of overweight children (two groups of mothers separated according to cultural origin—German or Turkish, and a mixed group of German and Turkish fathers). The juveniles classified themselves as overweight, while recruiting took place via schools and paediatrician practices in the urban districts selected—the socially disadvantaged origins of those taking part were specifically monitored. The preferred survey method was group discussion, since the aim was to investigate orientations typical of the social background and the individuals’ collective experience; for these to be mobilised an atmosphere of “mutual reference and challenge within a (group)-discourse” context was necessary (Bohnsack 2010).

The relatively high tendency of children from socially disadvantaged backgrounds to be overweight has been established by numerous studies. Furthermore this group also demonstrates a syndrome of factors known to promote obesity, such as poor eating habits, lack of exercise, high levels of media consumption, and—especially among children of Turkish extraction—specific cultural factors (cf. Kleiser et al. 2009; Schenk et al. 2008; Toschke et al. 2005; Power et al. 2003; Langnäse et al. 2002; Garn et al. 1981).

Expert perspective—the qualitative procedure of discursive risk and cross-impact balance analysis

At the same time, ZIRN—Stuttgart University’s ‘Interdisciplinary Research Unit on Risk Governance and Sustainable Technology Development’—pursued a parallel line of investigation, systematically recording how experts regard the complex causal structure of juvenile obesity. This produced a model that allowed expert assessments on the interplay of obesity inhibitors and drivers to be illustrated in the form of influence networks and different system statuses to be simulated using computer-assisted cross-impact balance analysis (CIB, Weimer-Jehle 2006). By entering the group discussion results it was possible to subject the model to a plausibility check (cf. Chapter Model plausibility check).

Two different perspectives on the issue of obesity were therefore established during the process of empirical analysis using parallel lines of investigation: the expert perspective thanks to the technique of discursive risk analysis and the everyday perspective by means of group discussions.

Subsequent stages in the project will include
  • combining the results of the two lines of study with the aim of deriving quality criteria and structural features suited to target group specific-prevention initiatives

  • developing an evaluation tool—guide/manual—for the purpose of discursive (self-) evaluation

  • holding expert workshops to validate results and testing applications to optimise findings

The following chapters of this article concentrate on the expert perspective survey and analysis.

Methods

Cross-impact balance (CIB) analysis, which is used in this project to provide a methodological approach to model creation, conceptualises systems as interaction networks. The creation of a cross-impact model on the development of obesity among socially disadvantaged children and adolescents therefore requires
  • the identification of major direct and indirect influencing factors on the weight trends of individuals belonging to this target group and

  • the assessment of interactions between these factors and weight trends and between the factors themselves.

How the interaction network thus defined then behaves and how intervention measures would influence the network can be analysed using a simple yet systematic balance procedure. CIB belongs to a group of cross-impact techniques originally proposed for the field of technology foresight (Gordon and Hayward 1968) and provides a method that is specifically designed to analyse qualitatively defined interaction networks. In the past CIB has been used for analyses concerned with energy supply, energy consumption, innovation, sustainability, water supply and environmental scenarios (Weimer-Jehle and Kosow 2011; Renn et al. 2009; Fuchs et al. 2008; Förster and Weimer-Jehle 2004; Aretz and Weimer-Jehle 2004; Förster 2002).

In view of the complexity associated with obesity risk revealed by obesity research, an expert panel capable of reflecting the unique range of the topic by virtue of its interdisciplinary nature was convened to assist in formulating the network (cf. Chap. Acknowledgements). Two criteria were taken into account when selecting the 18 experts from the fields of nutrition, exercise, medicine, sociology, psychology, treatment and prevention practice, and culture (youth food culture and Turkish culture): proven expertise on the topic of juvenile obesity and a clear interdisciplinary research orientation, with the related experience of communicating with other disciplines.

The information required for the model was collated during four 1-day workshops and one 2-day workshop. Between the workshops, surveys were used to record preparatory information. The first workshop served to select the main factors that would need to be included in the model. The three workshops that followed aimed at assessing the factors with regard to their interrelatedness and their impact on the model’s core factor, the energy balance of an overweight (but not obese) person. This basis enabled the project team to create a qualitative impact model, which was then analysed using the CIB algorithm. A final workshop was held for the experts to discuss the results and derive proposals for possible prevention measures.

The result of the expert discussion described is a qualitative model for the project target group's energy balance. The model consists of
  • 42 factors, which describe the personal behaviour, as well as the familial and social environmental conditions associated with the individuals, and the energy balance—as the 43rd factor—on which the model factors have a direct or indirect impact,

  • One hundred ninety-eight impacts between these factors characterised either as positive or negative, and with a weak, medium, or strong weighting.

Using this information the model describes an interdependent system, which on the one hand enables consideration of the obesity risks and their social context, and on the other can also be referred to for formal analyses using the CIB method.

Results

List of factors

The first result of the expert discourse was to limit the number of major direct and indirect energy balance-influencing factors being entered in the model. The selection of factors for the model is synonymous with a conceptualisation of the obesity problem by the panel of experts and in this sense a major stage in the project. It necessarily includes the critical decision as to which factors, considered less relevant, are to be excluded from the analysis. Against the background of the resources available and the projected scope of the model, the experts were given the option of selecting some 50 or so factors. During the selection process the experts’ assessment of the significance of various topics became clear; at the same time the suitability of the model for certain issues was determined, to the exclusion of others. The expert discussions identified the factors listed in Table 1.
Table 1

Factors considered by experts as especially significant in the occurrence of obesity among children and teenagers from socially disadvantaged families—and thus included in the model

Personal context factors

Dependent factors

 a. Education (formal and informal)

A. Professional status of parents/income

 b. Enculturation

B. Peer pressure

 c. Positive impact of family structure on upbringing

C. Positive energy balance

 d. Family cohesion

D. Enjoyment, positive experience in physical activity

 e. Fatalistic attitude to life

E. Feeling socially isolated/stigmatised

 f. Genetic/epigenetic predisposition to obesity

F. Exercise opportunities

 g. Sex/gender: girls subject to unique influences

G. Conflict/stress/boredom

 h. Overweight mother/parents

H. Specific and/or repeated experience of stigmatisation

 i. Physical inactivity

I. Body related self-esteem

 j. Immigrant background

J. Physical activity

 k. Affinity toward physical activity

K. High media consumption

 l. Smoking

L. High media availability in the home

 m. Healthy sleep pattern

M. Physical activity, nutrition, and health literacy

 

Societal context factors

N. Positive parental role model—exercise habits

 n. Deinstitutionalisation/erosion of the family

O. Positive parental role model—eating habits

 o. Loss of diurnal structure

P. Positive parental role model—media consumption

 p. Health-promoting range of foods

Q. Positive parental rearing practices

 q. High levels of media offerings

R. Quality of baby food/toddler nutrition/breast-feeding/weaning

 r. Quality of physical activity on offer

S. Rituals to promote healthy eating and exercise habits

 s. Being slim—society's established beauty ideal

T. Self-efficacy

 t. Social acceptance of being overweight

U. Eating habits that promote weight gain

 u. Social support via secondary socialisation agents

V. Effectiveness of hunger/satiety mechanism

The factors can be divided into two groups: ‘dependent’ factors, the existence or otherwise of which is to be explained by the model and ‘context’ factors, which produce relevant influences, yet which cannot themselves be explained using the model. The group of dependent factors also includes the model’s central core factor, the positive energy balance, which is shown in a different colour in the ‘dependent factors’ column in Table 1. The dependent factors relate to the individual characteristics and the familial environment.

‘Context’ factors are not intended to be explained by the model, partly because some are actually not subject to influence (such as genetic/epigenetic disposition towards obesity or sex) and partly because to explain them would go beyond the scope of the model. In this way addressing social and social environmental backgrounds—e.g. related to level of education or smoking behaviour—would involve similarly comprehensive models as the one we are using to weigh up obesity risks. For the group of context factors, we therefore assume that for the subject of the model, it is their impact on the other model factors that is most relevant; for pragmatic reasons, further analysis of these factors at this point must and can be dismissed. The context factors are divided into ‘personal context’, which incorporates the context conditions that vary from individual to individual, and ‘societal context’, which may be subject to change as society develops, yet is generally the same for all affected at a certain point in time and within a specific environment.

For the purpose of the model analysis each factor was allocated two possible statuses: either the ‘characteristic exists’ or the ‘characteristic does not exist’. The model does not take gradations or other means of differentiation into account. For the factor ‘J. Physical Activity’, for instance, this means that either significant physical activity exists or it doesn't. This simplification is not a result of the method selected, but is intended to limit the effort required to survey the interrelatedness of the various factors.

Factor interrelation

Once the factor interrelations are processed and formulated for operability, the list of factors becomes a model. Only then can system-related observations be made concerning the significance of factors for the occurrence of obesity and/or its prevention. These factor interrelations were specified during three workshops involving a panel of experts (cf. Chap. List of factors) and characterised in accordance with the CIB analysis method as ‘positive’ or ‘negative’ (cf. Table 2), and as ‘weak’, ‘medium’, or ‘strong’. The consolidated outcome of the expert discussions resulted in 198 interactions for the model on juvenile obesity presented in this paper. Of these 75 were judged to be weak, 77 medium, and 46 as strong. During analysis of the network those with a weak influence were given a weighting of 1, those of medium intensity a weighting of 2, and the strongest interactions a weighting of 3.
Table 2

Interpreting positive and negative influences

Positive influence of X on Y:

Negative influence of X on Y:

If…

…then the following is promoted:

If…

…then the following is promoted:

Factor X exists

Factor Y exists

Factor X exists

Factor Y does not exist

Factor X does not exist

Factor Y does not exist

Factor X does not exist

Factor Y exists

For example, Fig. 1 shows the interrelations of the ‘Physical Activity’ factor. Green arrows denote positive influences; red arrows stand for negative influences (cf. Table 2). Line thickness determines the strength of influence. Grey boxes represent dependent factors, while context factors are illustrated using pale blue boxes. Eight dependent factors and five context factors are shown to have sometimes positive, sometimes negative influences on physical activity. The factor also has an impact on itself—producing a self-reinforcing effect. By contrast physical activity has an influence on three factors, including the model’s core factor the positive energy balance.
https://static-content.springer.com/image/art%3A10.1007%2Fs10389-011-0473-8/MediaObjects/10389_2011_473_Fig1_HTML.gif
Fig. 1

The interconnectedness of the factor ‘Physical activity’

If the cross-impact images of all the factors are combined to create an overall picture, it produces an image such as the one found in the Appendix (Fig. 4). As the mental models presented by the experts are aggregated to produce the different aspects, the complexity of the obesity problem becomes all too clear.

A few figures suffice to underline the potential of such large and intricate networks. Our obesity model features 43 binary factors. This basically means that, if the potential configurations of all the binary factors were to be combined with each other, the model would produce 243 = 8,796,093,022,208 potential states (configurations). With reference to the division of the 43 into 21 context and 22 dependent factors, we can also claim that the model offers 221 = 2,097,152 different potential contexts, and for every single context situation 4,194,304 different configurations of the dependent factors.

This multitude of combinations is thus limited by the 198 network interactions. By adding this information, the network ‘learns’ the expert views of the panel—in the process acquiring content. Taking these interactions into account means we are no longer looking at approximately 4 million configurations per context situation. This is because randomly generated configurations of dependent factors almost always feature internal contradictions with regard to the interactions (e.g. the configuration for a particular factor produces the statement ‘Factor exists’, although the majority of influencing factors would seem to indicate that the ‘Factor does not exist’).

Cross-impact balance analysis (CIB, Weimer-Jehle 2006; Weimer-Jehle 2008) may be used to extract plausible (‘consistent’) configurations from the set of potential configuration combinations. CIB helps determine plausible configurations by using software to construct all the configuration combinations, checking in the process for each configuration whether every dependent factor reflects a status (prescribed by the configuration) that conforms to the sum of influences seen to have an impact. Any configurations deviating from this criterion, even in the case of a single factor, are rejected as inconsistent—the highly selective process therefore serves to identify a small number of plausible configurations from the multitude of potential combinations.1

Model plausibility check

During work on the project, we had the opportunity to check model plausibility by comparing case-related model results with the real-life circumstances of individuals. On the basis of 13 group discussions conducted with mainly overweight children and adolescents in 2009 and the parents of overweight children in 2010 (cf. Chapt. Background), a total of 128 evaluations was made for a sample of 13 individuals using the personal model factors. These case-related characteristics were fed into the model to estimate the obesity risks of individuals demonstrating such factors. These were compared with the current weight (normal or overweight) of the 13 sample individuals in order to assess model plausibility.

In each case, when comparing model results with empirical data, we checked—once all the empirical evaluations concerning the relevant individual (not including the weight evaluation) had been fed in—whether the model produced the correct conclusions for the ‘positive energy balance’ factor. On this basis the model was evaluated for each sample individual, checking for each scenario created whether the energy balance result produced by the model tallied with the empirical findings. If this proved to be the case, the model result was recorded as a hit.

In terms of the model’s core factor ‘C—Positive energy balance’, the hit rate for all 13 sample individuals and all the scenarios relating to them was approximately 90%. Furthermore the same approach was applied to all other factors for which empirical data were available and for which the model was designed to produce results. The model’s average hit rate for all factors was approximately 76%.

When evaluating the hit rates achieved, it is vital to remember that, even if the model were perfectly valid in a generic sense, the unique nature of individual lives would necessarily have a limiting effect. Bearing this in mind, the model plausibility check may be regarded as successful—within the empirical limits set. A truly accurate measure of the model’s performance would require a sample group featuring an approximate balance of obese and normal weight individuals and a much higher number of cases.

Model analysis

Results can either be extracted from the model simply by reasoning—e.g. by tracing the direct and indirect impacts on a factor (cf. Fig. 4), or via formal model analysis using the CIB method. Formal analysis has the advantage of systematically recording the above-mentioned extremely high number of potential model states and numerous effective interactions, allowing quantities of analyses and analysis variants to be conducted without investing too much time in the process. Despite the fact that the model also undoubtedly offers potential for discursive analysis, the following analyses therefore focus on formal analyses using the CIB method.

Definition of risk categories and risk profiles

Firstly, an initial analysis of the model without any factor-related restrictions (‘baseline case’) is made to introduce the concept of risk categories and profiles. This means that the 21 factors create a scope of 221 = 2,097,152 different ‘context cases’, since the two results—‘Factor exists’ and ‘Factor does not exist’—are as yet open to each context factor without restriction. For the statistical analysis of this potential scope N = 3,000 random context cases were created, and for each of these samples all 222 = 4,194,304 possible configurations (‘scenarios’) for the 22 dependent factors checked according to the CIB consistency criterion (cf. Chapter Factor interrelation). This revealed that each context case generated a typically single-figure number of scenarios for the dependent factors, representing the various conceivable types of individual development for persons living in the circumstances described by the context. Every context case thus analysed was allocated to one of three risk categories:
  • The ‘red’ risk category comprised context cases that only featured scenarios tagged with the factor ‘Positive energy balance exists’. This corresponds to a prognosis of weight gain.

  • The ‘green’ risk category stands for context cases which only included scenarios tagged with the factor ‘Positive energy balance does not exist’.

  • Finally, the ‘amber’ risk category comprised context cases that featured both scenarios with and without the factor ‘Positive energy balance exists’. In such instances the model diagnosed contingency, i.e. the model's logic predicted mixed weight development for individuals within a group under the influence of the circumstances described by the context.

When the model is not subject to any context or other factor-related restrictions (baseline run), approximately 19% of context cases are allocated to the ‘red' risk category, some 62% of context cases are classified as ‘amber’, and the last 19% of context cases are assigned to the ‘green’ risk category. Whereas if some of the context factors are restricted—for instance the social context factors are selected to reflect the current societal realities in Germany (as assessed by our panel of experts)—the risk profile shifts dramatically in a negative direction (cf. Fig. 2).2 On the other hand, if a few important personal factors are rated positively (e.g. existence of high education, positive impact of family structure on upbringing, family cohesion), these can more or less balance out the impact of the societal environment. This is explained by the resilience-promoting nature of these factors. When looking to achieve a lasting improvement in the risk profile in the model, it is also necessary to change the societal context. As a rule any dramatic improvement in the risk profile within the model resulting from improvements in the societal context can only be achieved with a combination of measures; isolated change tends to remain ineffective.
https://static-content.springer.com/image/art%3A10.1007%2Fs10389-011-0473-8/MediaObjects/10389_2011_473_Fig2_HTML.gif
Fig. 2

A few examples of context cases and their risk analysis using the obesity model. Expert assessment of the current social context in Germany: deinstitutionalisation/erosion of the family: yes; loss of diurnal structure: yes; health-promoting range of foods: no; high media availability: yes; quality of physical activity: no; being slim as beauty ideal: yes; social acceptance of obesity: no; social support via secondary socialisation agents: no

Analysis of factor influences

Using the concept of a risk profile introduced in Chap. Definition of risk categories and risk profiles results concerning the systemic significance of a single factor to the energy balance analysis can also be derived. To achieve this, the factor under examination is set to give a result that will increase the risk of obesity, while all the other factors remain without restriction. An analysis of the model then reveals to what extent the risk profile has worsened in comparison with the baseline case. Figure 3 shows the 20 factors found under analysis to produce the most dramatic shift in the direction of ‘red’. If these factors are made to produce the opposite (risk-reducing) results, we see a similar movement into the ‘green’ risk category.
https://static-content.springer.com/image/art%3A10.1007%2Fs10389-011-0473-8/MediaObjects/10389_2011_473_Fig3_HTML.gif
Fig. 3

The 20 factors with the strongest individual impact on energy balance. The table shows risk profiles produced by model analysis when one single weight gain-promoting characteristic of the respective factor is prescribed. The average margin of error for the calculations is approx. 1 %

One central result of the analysis is that no single factor is so significant that it can be made solely responsible for an unfavourable weight trend (or conversely a favourable one). Although primary influence factors such as ‘Physical activity’ and ‘Weight gain-promoting eating habits’—both characterised in the model by a strong direct impact on the energy balance (cf. Fig. 4)—have a pronounced influence on the risk profile, there is still scope for movement in both trend directions. Most factors, if regarded in isolation, are limited in their impact. Only in combination with other factors are they capable of effectively shifting the risk profile in one direction or another, as can be seen from the table: the impact network of obesity remains beyond the realm of simple explanation, even in reduced ‘model reality’ terms.

The table shows risk profiles produced by model analysis when one single weight gain-promoting characteristic of the respective factor is prescribed. The average margin of error for the calculations is approximately 1%.

The unique significance of the ‘Physical activity’ and ‘Eating habits’ factors apparent in Fig. 3 naturally comes as no surprise; the difficulty lies in determining how such factors can best be influenced. Although discussion on prevention now acknowledges the significance of education as a factor, the promotion of the latter with reference to obesity prevention is not actively considered.

What may possibly be new to prevention practice is the conclusion that arises from Fig. 3: i.e. that parental rearing practices, conflicts/stress, nutrition and health literacy, family cohesion, as well as secondary socialisation agents, are among the most important aspects needing to be addressed by general (not case-related) obesity prevention initiatives.

Ways of using the model to assess individual cases

The model has been designed to analyse the obesity risk of groups, since the blanket analysis of interactions on which the model is based only achieves validity in a group context. The model becomes less meaningful the closer we get to analysis at a personal level, since the complexity and contingency of individual lives make it impossible to forecast interactions in an individual context. This does not however mean that the model is wholly unsuitable for analysing individual cases—the encouraging results from applying the model to empirical data during another line of project investigation (cf. Chapter Model plausibility check) clearly indicate the model’s potential in this area. When dealing with the results of analysis conducted at a personal level, it nevertheless remains important to consider that the unique nature of individual lives necessarily has the potential to produce different cross-impacts to those suggested by the model.

There now follows a sample analysis of an individual case, Janina3 (a participant in the group discussion involving girls aged 14–16, whose grandmother attended the group discussion for mothers of overweight children), linked to the development of personalised prevention approaches. The group discussion allowed an assessment of Janina’s environment with relation to 13 model factors, 6 of them from the personal context group (Table 3); in addition eight generic assumptions were made about social context (cf. Chap. Definition of risk categories and risk profiles), thus enabling the model to be fed with information on 14 of the 21 context factors and 7 of the 22 dependent factors.
Table 3

Group discussion-based factor assessments for ‘Janina’

A. Professional status of parents/income

High*

B. Peer pressure

No

D. Enjoyment, positive experience in physical activity

Yes

G. Conflict/stress

Yes

H. Specific and/or repeated experience of stigmatisation

No

Q. Parental rearing practices

Yes

U. Eating habits that promote weight gain

Yes

 a. Education

High*

 e. Positive impact of family structure on upbringing

No

 f. Family cohesion

Yes

 i. Sex/gender:

Female

 j. Overweight mother/parents

Slightly

 n. Immigrant background

Polish

*The strength of factor is interpreted in relation to the socially disadvantaged target group

The model produced a total of 160 scenarios for Janina's environmental circumstances. The high number of scenarios resulted from residual context uncertainty (no information was available for seven context factors). Of the 160 scenarios related to this case, exactly half showed a favourable trend and half showed an unfavourable trend in weight development. As long as a considerable share of the context remains unknown, it therefore remains impossible to accurately assess Janina’s weight trend. Interestingly this ambivalence corresponded with the impression that Janina made during the group discussion. At the time of the group discussion, she was seriously obese, in correspondence with half of her model scenarios. Owing to her personality, her attitudes and her environment, however, there appeared to be a real chance of favourable development. From the contributions made by Janina and her guardian to the relevant group discussions, it may be assumed that the early loss of both parents—critical event in her life—has played a major role in her weight development thus far.

Counselling should aim to identify ways of promoting a trend reversal. One approach relevant to the case selected is the fact that Janina’s analysis produced a scenario set of mixed scenarios. The scenario set was sorted according to weight trend and the two subsets (favourable vs. unfavourable weight trend) compared with each other. In the process it became clear where and in association with which factors the two subsets typically deviated from one another. This means the model can be used to help identify those factors hidden in the anamnesis that might play a significant role in the weight trend of the individual in question. In Janina’s case the following approaches for support counselling arose from the differentiation of the factors4:
  1. 1.

    Attempt to combat any existing physical inactivity on the part of the individual

     
  2. 2.

    Attempt to motivate the individual to get sufficient sleep

     
  3. 3.

    Attempt to establish a positive influence on the hunger-satiety mechanism

     
Bearing in mind that there is a chance Janina could be genetically/epigenetically pred isposed to obesity, the model attached particular importance to the third measure. In addition to these measures, checks concerning the extent to which the empirically determined unfavourable factors are subject to direct interaction would also need to be made. Such checks specifically apply to the factors ‘Conflict/stress’ and ‘Weight gain-promoting eating habits’. Since it directly targets recognised problem areas, this section of the recommendations reflects the conventional prevention strategy. Nevertheless the model also offers insights as to how best to approach the problem areas identified. With respect to Janina’s empirically proven weight gain-promoting eating habits, the model specifies the following, among others, as those factors that have a major influence on the encouragement or inhibition of eating habits (cf. Appendix—Fig. 4):

Parental rearing practices

(In Janina’s case empirically proven to exist)

Conflict/stress

(In Janina’s case empirically proven to exist)

Physical activity, nutrition, and health literacy

(No empirical evidence in Janina's case)

Positive parental role model—eating habits

(No empirical evidence in Janina's case)

Rituals

(No empirical evidence in Janina's case)

Health-promoting range of foods

(Assumed to be non-existent)

These factors need to be checked for potential influence, despite the fact that in this case the manner of rearing practices has already been positively assessed, and the range of foods available cannot be influenced at a personal level, relying at best on a political initiative. In addition to the strong factors already mentioned, the model specifies other factors of a medium and weak weighting that are not discussed here for reasons of brevity.

In general these model recommendations refer to Janina as a person. For other individuals with a different profile the model would generate different recommendations. It was not possible to assess the appropriateness and effectiveness of these recommendations within the scope of the project. The analysis nevertheless shows that model-based scenario analyses may be suited to generating specific recommendations for use in prevention practice.

The model analyses described in this paper serve as an example of the type of analysis potential found in the model. Group-specific living conditions, for example, may also be prescribed for the model, thus determining the most effective prevention approaches for the group in question; in addition the model may also serve as the basis for a general discussion into the effects of social change on the risk profiles of certain groups. When interpreting model results, it is nevertheless crucial to remember that the model is not based on empirical data, but rather on a discursive survey of experts. In terms of the model this means that it represents the reconstruction of the issue by a well-informed panel of experts, and not reality itself. For this reason and since the description of interactions in model form necessarily involves a considerable amount of simplification and generalisation, the model does not claim to furnish any evidence. Instead it should be principally regarded as a heuristic tool, capable of deducing the major facets of the complex origins of obesity and pinpointing potential contexts and effects in a systematic and transparent way, while drawing on a qualitative database formed from expert assessments.

In the future we intend testing the model in additional practical contexts associated with obesity prevention, thereby gaining further insights into the model’s validity and practical suitability.

Prevention: applying the model

The next stage foresees the application of the cross-impact model in the Lale project—iss bewusst & sei aktiv! (Be aware of what you eat and keep fit.) The project aims to prevent obesity and promote healthy living among families of Turkish extraction, taking the tulip—the Turkish ‘lale’ gives its name to the initiative—as a symbol of well-being. The project is being run by North Rhine Westphalia’s Ministry for Climate Protection, Environment, Agriculture, Nature Conservation and Consumer Protection (MKULNV) and its Department of Nutritional Policy and Sustainable Consumption. In a multipliers training course designed in consultation with nine additional project partners and covering nutrition, fitness, and relaxation, bilingual nutrition and fitness experts are currently undergoing instruction in the prevention of obesity among families of Turkish extraction.

At the beginning of 2012 we will present the cross-impact model to the steering group, with the intention of subsequently testing its practical suitability together with the course participants. On the basis of the experience gained during this pilot phase, the Lale project steering group will then assess whether a relevant module should be included in the training curriculum. Should the approach be permanently adopted, the tool will be made available to other interested parties involved in obesity prevention.

Discussion

Although there have been years of intensive research into obesity and the topic these days involves a broader disciplinary spectrum than ever before, relatively few clear results exist on its origins. Even the influence of nutrition remains unclear. While a handful of studies has revealed what can at most be termed a weak connection between eating habits and weight status, others have failed to prove even the weakest of links (cf. Müller et al. 2011 for instance). When it comes to prevention, this ambiguity presents a two-fold problem. Not only because there is uncertainty concerning the suitability of the approach, but also because it is difficult as a result to clearly address the issue of responsibility. Maziak et al. expound upon the problem as follows: “In the absence of consensus about the causal pathways leading to the obesity epidemic, it is hard to devise a public health response that can affect its course’ (Maziak et al. 2008). Marion Nestle’s observations are in the same vein: “Everyone knows that American children are becoming fatter, but not everyone agrees on the cause’ (Nestle 2006). In an expert Delphi on preventing obesity in children conducted by our research group, empirical confirmation of these observations was provided. Among those stakeholders in attendance—representing the food industry, health insurance companies, and public authorities—“there was agreement on the need to act, but a failure to reach a consensus as to whom should take responsibility for implementing the measures (Zwick and Schröter 2011).

Our model is not able to solve either of these two problems. We can however help to make the problems more transparent. This is prerequisite to finding a solution. As far as the problem of uncertainty is concerned, our panel of experts was also forced to work within the currently ambiguous state of research. This should be remembered when interpreting the results of the model. The information fed into the model about the factors and their interactions cannot be any clearer than the research results from which our experts derived the information in the first place. But what the model does reveal is as follows: in terms of juvenile obesity there are no key factors that bear the main burden of responsibility. And because no such factors exist to dominate the remaining influences, the explanation also remains unclear. No universally applicable ‘law of nature’ on obesity exists. Instead there are many factors involved in an equally intertwined and fragile network of interaction. Seen in isolation the impact of each one of these factors is pretty weak. Yet it only takes a handful of the factors in this system to assume a different result, and the energy balance result can be seen to shift from its prior state. Which factors are involved varies from case to case. This is why our model does not feature any principal factors. Those conducting cause-effect studies should therefore not expect clear results—as the entire topic under investigation is anything but unequivocal.

As far as responsibility is concerned, however, it should not be assumed that this finding provides scientific legitimacy for inactivity or an excuse to delegate responsibility elsewhere. The opposite is the case. No one—e.g. the media, the food industry and the schools—is being asked to take sole responsibility for the situation; instead, there are many stakeholders in our society, all of whom need to share in the responsibility. The model may be regarded as a plea in favour of a multilateral approach to prevention.

In our opinion, measures should not simply focus on cancelling out risk factors (e.g. forbidding snack foods and soft drinks in schools). Instead, in accordance with the results of the model analysis, greater emphasis should be placed on promoting resilience factors (e.g. reinforcing individual self-efficacy). Since there is a limit to the restrictions that can be imposed upon a free society for the sake of preventing obesity (e.g. market limitations), it is these resilience factors that are called upon to play the greater role. The effectiveness of the latter becomes clear when we consider that the majority of children and adolescents alive today—despite the obesogenic environment—still manage to have a healthy energy balance. Although this project did not address those children and adolescents who, despite the existing context factors, maintain a normal weight—no doubt a worthwhile subject—our model still provides information, based on the assessments of the expert panel, as to which factors provide effective protection against the risk factors under the given circumstances. We may assume that these resilience factors produce a positive result among children with a normal weight.

Footnotes
1

Network analysis using CIB is described in detail by Weimer-Jehle 2006 and at http://www.cross-impact.de

 
2

The share of risk categories included in the risk profile should not be directly interpreted as probabilities, since the probabilities of the different context cases are not rated during counting. As long as no systematic link exists between the probability and the risk category of the various context cases, the risk profile provides a qualitative model result on the risk character of a context case that is suitable for comparison.

 
3

Name changed.

 
4

The factor ‘Smoking’ also emerged. Since this factor cannot, however, be considered as a measure owing to the associated health risks it is not listed among the support counselling recommendations.

 

Acknowledgements

The authors would like to thank the members of the expert panel, without whose untiring commitment (which far exceeded our expectations) the model would never have been a success.

The expert panel consisted of

Prof. Dr. Silke Bartsch, Department Everyday Culture and Health, Karlsruhe University of Education

Aytekin Celik, Stadtjugendring Stuttgart e.V.

Prof. Dr. Ingrid Hoffmann, Department of Nutritional Behaviour, Max Rubner-Institut, Federal Research Institute of Nutrition and Food, Karlsruhe, Germany

Eva Hummel, Department of Nutritional Behaviour, Max Rubner-Institut, Federal Research Institute of Nutrition and Food, Karlsruhe, Germany

Daniela Kahlert, Department of Sports and Health Care, Potsdam University, Germany

Dr. Robert Jaeschke, Rehabilitation Clinic for Children, Fachkliniken Wangen, Germany

Dr. Lars Libuda, Research Institute of Child Nutrition, Dortmund, Germany

Dr. Claudia Müller, Department Life Sciences and Facility Management, Zurich University of Applied Sciences, Switzerland

Prof. Dr. Manfred Müller, Institute for Human Nutrition and Food Sciences, Christian-Albrechts-University, Kiel, Germany

Dr. Andreas Oberle, Social Paediatric Centre, Centre for Paediatric Medicine, Olgahospital, Medical Centre Stuttgart, Germany

Dr. Claudia Peter, Institute of Social Research, University of Frankfurt, Germany

Regine Rehaag, Institute for sociology, Leibniz University of Hannover, KATALYSE Institute for Applied Environmental Research, Cologne, Germany

Dr. Ulla Simshäuser, Department Health Promotion, Heidelberg University of Education, Germany

Prof. Dr. Petra Wagner, Institute of Exercise and Public Health, Leipzig University, Germany

Corinna Willhöft, Department of Nutritional Behaviour, Max Rubner-Institut, Federal Research Institute of Nutrition and Food, Karlsruhe, Germany

Prof. Dr. Eva Wunderer, Social Work, University of Applied Sciences Landshut, Germany

Claudia Ziegler, Children's Hospital Auf der Bult/Medical School Hannover, Germany

Dr. Michael Zwick, Institute for Social Sciences, University of Stuttgart, Germany

Thanks are due also to Ms. Sophia Alcantara for painstakingly creating the system graphs.

The project was funded by the German Federal Ministry for Education and Research prevention research programme.

Conflict of interest

The authors declare that they have no conflict of interest.

Copyright information

© Springer-Verlag 2012