A feasibility study of using a diet optimization approach in a web-based computer-tailoring intervention for adolescents

Abstract

Objective:

Adolescents are an interesting but neglected target group in obesity prevention. We assessed the feasibility of using a diet optimization approach for computer-tailored nutrition interventions for adolescents.

Method:

Development of an optimization model based on the public health approach to diet optimization. On the basis of food frequency questionnaires (FFQ) of 48 adolescents (14–17 years) optimized diets were calculated.

Results:

The optimization calculations for all cases resulted in individual advice. On a total of 137 items included in the FFQ, the individualized advice included changes in a minimum of 36 and a maximum of 88 items (mean: 61 items), recommendations for changes in the food items ranged from less than 1 g day−1 up to 1660 g day−1. In almost all cases a higher intake of fruit and vegetables was recommended; some unexpected advice was also generated (for example, to decrease the consumption of brown bread and to increase the consumption of pizza). The strengths and weaknesses of the optimized diets are discussed.

Conclusion:

Using the optimization approach is a step forward in nutrition tailoring interventions but the model used in the present feasibility study still needs to be refined.

Introduction

Adolescence is a period when the physiological need for the consumption of a diet with good nutritional quality is particularly important.1 During this period, adolescents gain 15–20% of adult height and 50% of adult weight while they accumulate up to 45% of their skeletal mass.2 With regard to overweight and obesity prevention, targeting this population with messages concerning their eating habits is important. The school is a setting in which almost every adolescent can be reached; however, until now school interventions aiming to improve eating habits have not shown important changes in the dietary behaviour of adolescents, positive results were found only for subpopulations and for specific food items.3, 4, 5

Computer tailoring, a theory-driven health education strategy, has become more popular in the past decade and quite strong, short- and medium-term, positive results in adult populations were reported, especially for fat intake.6 Computer tailoring can be defined as the adaptation of health education to a specific person through a computerized process.7 If used several times it can be regarded as a method for self-monitoring, a successful strategy to control or prevent weight gain.8 Until now only three computer-tailored interventions on eating behaviours for adolescents have been evaluated.9, 10, 11 In the first study,9 the effectiveness of a tailored intervention to reduce fat intake was investigated in families with two adolescents. After 4 weeks, there was a reduction in the percentage of energy from fat, but only in those adolescents with high baseline fat intake. In a second study, fat intake and physical activity were targeted through a tailored intervention offered during eight 45-min lessons in low-income, culturally diverse students. The intervention was effective for reducing the percentage of energy from fat.10 In the third intervention,11 all students in the classroom went through a single computer-tailoring intervention on fat intake on individual computers. The intervention showed a decrease in fat intake in two sub-samples, namely girls in technical--vocational education and in boys and girls following general education who had read the whole advice provided.

Until now all computer-tailored interventions on nutrition are targeting one or just some aspects of the eating habits, for example, fat intake, fruit and vegetable consumption, soft drinks, and so on.6, 12 However, in the context of obesity prevention, research and interventions focusing on multiple dietary behaviours simultaneously are frequently advocated.13, 14

On the basis of promising results of previous studies with computer-tailored interventions in adults and in adolescents on specific aspects of their eating habits, the Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) Study Group aimed at developing a computer-tailoring intervention for enhancing the total diet of adolescents. The HELENA study is described in detail by Moreno et al.15

The aim of this study is to describe the concept and development of the HELENA intervention and to discuss the use of a diet optimization approach for computer-tailored interventions on nutrition for adolescents.

Materials and methods

Development of the tailoring instrument

The development of the HELENA tailoring intervention was based on the procedure of developing computer-based interventions described by Brug et al.16 For the HELENA intervention the following components are (1) a validated food frequency questionnaire for measuring the dietary intake of adolescents; (2) a corresponding food composition database; (3) an optimization model adapted from the public health approach for diet optimization, described by Gedrich et al.17

The intervention starts with an introductory page giving information on the aim of the intervention, some demographic questions including sex, height and weight, and screening questions for eating disorders. Adolescents identified as at risk for eating disorders could not continue to fill in the questionnaires and get standard advice on healthy eating.

To measure the energy intake and target nutrients (fat, fiber and calcium) of the adolescents a food frequency questionnaire (FFQ) was chosen. An FFQ was developed based on the validated computerized FFQ for fat intake developed in Belgium.18 This FFQ was extended and adapted to measure the intake of the target nutrients: energy, fat, fiber and calcium. In total, 137 items (food groups, for example, fresh fruit; and individual food items, for example, white bread) were identified as contributing substantially to the overall intake of these components. A validation and reliability study was conducted in 48 Flemish adolescents (results will be published elsewhere).

For calculating the nutrient intake a food composition database was built with a total of 137 items mainly from the German so-called Federal Food Code (BLS),19 and supplemented with the Dutch20 and the Belgian21 Food Composition Tables. For these items, all nutrients (energy plus 30 nutrients) included in the optimization approach were available from the food composition databases used.

For optimizing the usual diets of the respondents the ‘public health’ approach, described by Gedrich et al.17 was chosen. It aims at determining optimized dietary patterns of individual persons or person groups based on information on their dietary habits and their corresponding nutrient intake recommendations. This approach assumes that a person's diet meets their subjective nutritional requirements (that is, personal preferences) quite well,22 but usually does not completely meet the objective nutritional requirements (that is, the standards for a healthy diet). Nutrition optimization, therefore, aims at determining a diet that totally meets the standards for a healthy diet, whereas deviating from the actual diet as little as possible (for reasons of acceptability). This deviation is considered as the sum of squared differences between the actual and the optimized amounts of food (to be) consumed for a given period of time. By means of quadratic programming this sum is minimized subject to a set of constraints given by the standards for a healthy diet.17

A diet was defined to be healthy, when all the related nutrient intake recommendations are met. For the computer-tailoring intervention, the following recommendations were chosen to be used: Eurodiet,23 D-A-CH Reference Values for nutrient intake,24 and World Health Organization recommendations.25

Most of the recommendations related to the intake of vitamins, minerals or trace elements include certain allowances to ensure that a corresponding intake meets the need of almost any healthy person.24 In such cases, recommendations were multiplied by 0.8 to obtain an estimate of the average nutrient intake requirement (assuming that the coefficients of variation of the nutrient requirements range from 0.1 to 0.15).

The optimized diets were calculated by means of quadratic programming and solutions were obtained using the Software LINGO Hyper (Release 10.0, LINDO Systems Inc., Chicago, IL, USA).

Design and subjects

Masters’ students in Social Medical Sciences of the Ghent University (n=11) were asked to approach five adolescents each between 13 and 17 years to participate in a set of validation studies related to food and physical activity. The questionnaires were completed at the pupils’ or the students’ home. Pupils who completed one or more questionnaires inaccurately (n=7) were removed from further analyses, such that dietary information of 48 students is available; 54% of the final samples were boys.

Results

The FFQ reported an average energy intake of 2679±718 kcal day−1.

The optimization calculations resulted for all cases in individual advice, for example, changes in a minimum of 36 and in a maximum of 88 items (mean: 61 items/respondent); recommendations for changes ranged from less than 1 g day−1 for individual food items up to 1660 g day−1. When minor changes (that is, less than 5 g day−1) are disregarded, a minimum of 16 and a maximum of 46 recommended changes were identified (mean: 31 items/respondent).

Table 1 gives an overview of the results obtained from the optimization calculations for the food items for which a change in consumption (recommended increase or decrease) was calculated for at least 75% of the respondents. For each item, descriptive statistics on recommended increases of intake are presented in the left part of the table, whereas corresponding information on recommended decreases can be found on the right.

Table 1 Descriptive statistics on dietary advice obtained by nutrition optimization for food items that were recommended to be changed in at least 36 out of 48 respondents (75%)

In almost all cases a higher intake of fresh fruit, prepared vegetables without sauce and raw vegetables, was recommended by the programme. In 46 of the 48 cases (96%) a higher intake of fresh fruit was computed as recommendable varying between 1 and 1474 g day−1 with a median of 52 g day−1, and an average recommendable increased intake of 158 g day−1. However, for two cases the software advised to decrease the intake by 33 g and 87 g day−1, respectively. The intake of raw vegetables and prepared vegetables without sauce should be increased in 92 and 94% of the cases varying between 0.3 and 266 g day−1 for raw vegetables and between 1 and 690 g day−1 for raw vegetables without sauce.

Items that were most recommended to be decreased were other biscuits, chocolates and candy bars in 40 (83%) and 39 (81%) respondents with up to 29 and 100 g day−1, respectively. For one respondent an increase of 5 g day−1 of chocolate and candy bars was advised.

Other notable results are a total of 42 participants (88%) who are encouraged to drink more fruit and vegetable juices up to an increase of 853 g day−1.

In seven cases (15%) drinking water should be reduced between 6 and 600 g day−1, in 23 (48%) participants the intake of brown bread should be reduced by 1–189 g day−1 and in 8 (17%) cases the intake of soup should be reduced by 22–477 g day−1.

An increased intake of pizza was to be recommended to 40 participants (83%) ranging from 0.2 to 142 g day−1, an increase in the consumption of crisps for 37 (77%) participants (up to 80 g day−1), an increase of the intake of croissants and buns was recommended in 35 cases (73%) cases (up to 115 g day−1), an increase in meat and fish salads (containing a lot of mayonnaise) in 31 (65%) participants (up to 74 g day−1), an increase in soft drink consumption in 25 (52%) participants (up to 936 ml day−1) and an increase in French fries in 14 (29%) cases (up to 21 g day−1).

Discussion

The use of the public health optimization approach opens new perspectives for tailoring interventions as it provides the basis for nutritional advice regarding the complexity of the dietary needs of a person and taking into account their personal food preferences. In aiming at meeting the recommendations, this approach differs from the fuzzy logic approach for diet optimization in which the recommendations are met ‘as well as possible’.17 No other interventions for adolescents or adults are known using the public health diet optimization approach. Using the diet optimization approach avoids the necessity of developing and implementing several separate, possibly competing interventions on different aspects of healthy eating. As the development of tailoring interventions is labour intensive and expensive and as the adolescent population is difficult to motivate to change their eating behaviours for health purposes, generating one intervention tool tailored to their specific deviations from the recommendations might be an advantage. As importantly, an optimization approach has the advantage regarding the multitude of nutritional needs simultaneously, such that it avoids the risk of introducing new problems when solving others (for example, by advising a greater intake of milk products to meet the calcium recommendations, the intake of saturated fat can be increased). This risk exists when the intervention is only dealing with one or a few aspects of the diet (for example, fruit and vegetables, fat intake).

However, the feasibility study of using the optimization approach revealed several problems.

Nutrient intake-based advice

The current diet optimization process is an exclusively nutrient-based nutrition optimization approach. Such an approach has—almost by definition—three potential weaknesses: (1) the assessment of usual food intake, (2) the determination of food composition and (3) the definition of recommendable nutrient intakes. The three aspects will be discussed below.

For assessing the participants’ usual food intake the FFQ has been used. FFQs are designed to assess usual eating habits over a certain period of time. In the present study the recall period was the previous month and the nutrients of interest were energy, fat, fiber and calcium. The main advantages of FFQs are their ease and uniformity of administration, low cost and their use with samples, which are geographically widespread (for example, different European countries). A disadvantage is the limited number of nutrients that can be assessed with adequate precision. Therefore, one may question the validity of the questionnaire related to those nutrients that are originally not included in the questionnaire (for example, vitamin D) taking into consideration that the nutrition optimization process considered a much wider range of food constituents. But there are no alternatives available.

The validity of any indirect approach to assessing nutrient intake (through the assessment of food intake) relies on the choice of a corresponding food composition database. In the present study, national food composition databases were used (Germany, the Netherlands and Belgium); therefore, comparisons with other European countries could introduce errors.26 In the current diet optimization tool the food composition database primarily used was the official German food composition database, known as BLS. Such an approach, however, might introduce country-specific errors, when dietary advice is provided in different countries. The pizza, for instance, listed in Table 1, was assumed to be a vegetarian pizza margarita, which would be a relatively good source of folate associated with relatively low amounts of unfavourable constituents. If the usual choice of pizza in an area (or a country) was not in agreement with that assumption, then the computed dietary recommendations for the participants of the feasibility study would not be valid. Therefore, it seemed advisable to follow the general rule that the use of local food composition tables is to be favoured, if accurate nutrient information is available.27

The diet optimization approach used here is based on a number of constraints exclusively based on nutrient intake recommendations. The purpose of these nutrient recommendations is to maintain and to promote health and quality of life (DACH).25 The World Health Organization and the Food and Agriculture Organization mentioned that the aim of reference values is to ensure the vital metabolic, physical and psychological functions in nearly all healthy individuals of the population.28 Historically, the main objective of these recommendations was to prevent deficiency disorders and not to maintain good health, preventing obesity or major chronic diseases. However, science is not yet able to prove whether a single nutrient or a combination of nutrients and non-nutrients are the sources of benefit for health; therefore, food-based dietary guidelines are introduced as reference values.29 On the basis of a strictly nutrient-related model, the optimization approach used in this feasibility study did not take these recent developments into account. Nevertheless, the optimized diets of the study participants are in accordance with certain frequently stated food-based recommendations, such as a high intake of fruit and vegetables. For instance, the optimized intake of fruit and vegetables averaged to 280 g day−1 (with a median=165 g day−1) and 264 g day−1, respectively (median=237 g day−1) with a total of 544 g day−1 (median=507 g day−1). Furthermore, food-based intake recommendations with satisfying evidence could easily be included into the optimization approach by enhancing the model with corresponding constraints.

Complexity of output and nutrition communication

On account of the distinct interpretation of the nutrient intake recommendations, the optimization calculations usually suggest changes in intake for almost all food items consumed, even though the changes are often quite small (a few grams). Such an output cannot immediately be used for the purposes of nutrition communication, as it includes too much information and does not distinguish between important and irrelevant messages. Simply shortening the output, for instance, would not solve the problem, as quantitatively minor changes in the intake of several food items might sum up to considerable effects that should not be neglected. Thus, translating the output of the optimization calculations into useful feedback to consumers requires either the development and evaluation of specific algorithms for identifying the messages to be passed on or a modification in the optimization procedure itself, such that the intake of only a limited number of food items would be affected and immediately usable information for feedback would be generated.

The complexity of the actual output might be reduced by reducing the number of nutrients in the model. In the model used, 30 nutrients and energy were included and treated as equally important. In future developments of the model, the number of nutrients might be reduced based on evidence concerning the importance of the nutrients for the main nutritional problems of adolescents or weights might be included in the model to reflect the importance of the different nutrients.

A second drawback of the actual output is that respondents deviating most from the recommendations do not get stepwise advice; the model tries to solve all the problems at the same time, which might not be realistic. A stepwise approach might be more motivating and feasible for the respondents.

Until now no feedback module was developed, as the mathematical model itself needs to be further developed; once a satisfactory model has been found in further experiments a feedback module will be developed.

Individual output in relation to general food-based dietary guidelines

The purpose of the chosen optimization procedure is to determine an individual dietary pattern that meets a set of nutrient intake recommendations with as little deviation from an existing pattern as possible. Consequently, the optimized diet vastly depends on the given dietary pattern. Persons, for instance, indicating in the FFQ that they are vegetarian would not be advised to increase their meat intake to reach the recommended levels of iron intake. And other persons stating they do not drink anything else but soft drinks, might even be told to increase the corresponding intake for meeting the recommendations of water intake, provided that other recommendations (for example, regarding sugar intake) are not violated. These examples demonstrate the conflict between the assumed acceptability of dietary advice on an individual basis and the accordance of dietary advice with general recommendations given on a population basis. The approach chosen here gives more attention to acceptability and would tolerate deviations from general food-based recommendations, as long as nutrient intake recommendations are met.

Such recommendations, however, conflict with other messages adolescents get in the school context, through the media or other channels. This can lead to confusion or, even worse, to doubts about the credibility of the individual advice or the general nutrition recommendations.

Conclusion

Using the optimization approach is a step forward in nutrition tailoring interventions, but the model used in the present feasibility study still needs to be refined.

References

  1. 1

    Dwyer JT . Diet and nutritional strategies for cancer risk reduction-focus on the 21st century. Cancer 1993; 72: 1024–1031.

    CAS  Article  Google Scholar 

  2. 2

    Gong E, Heald F . Diet, nutrition and adolescence. In: Shils M, Olsen J, Shike M (eds). Modern Nutrition in Health and Disease. Lea and Febiger: Philadelphia, 1994, pp 7759–7769.

    Google Scholar 

  3. 3

    Nicklas TA, Johnson CC, Myers L, Farris RP, Cunningham A . Outcomes of a high school program to increase fruit and vegetable consumption: Gimme 5—a nutrition concept for students. J Sch Health 1998; 68: 248–253.

    CAS  Article  Google Scholar 

  4. 4

    Lyttle LA, Murray DM, Perry CL, Story M, Birnbaum AS, Kubrik MY et al. School-based approaches to effect adolescents’ diets: results form the TEENS study. Health Educ Behav 2004; 31: 270–287.

    Article  Google Scholar 

  5. 5

    Haerens L, De Bourdeaudhuij I, Maes L, Vereecken C, Brug J, Deforche B . The effects of a middle school healthy eating intervention on adolescents’ fat and fruit intake and soft-drink consumption. Public Health Nutr 2007; 10: 443–449.

    Article  Google Scholar 

  6. 6

    Kroeze W, Werkman A, Brug J . A systematic review of randomized trials on the effectiveness of computer-tailored education on physical activity and dietary behaviours. Ann Behav Med 2006; 3: 205–223.

    Article  Google Scholar 

  7. 7

    De Vries H, Brug J . Computer-tailored interventions motivating people to adopt health promoting behaviours: introduction to a new approach. Patient Educ Couns 1999; 36: 99–135.

    CAS  Article  Google Scholar 

  8. 8

    Baker LE, Kirschenbaum DS . Self-monitoring may be necessary for successful weight control. Behav Ther 1993; 24: 377–394.

    Article  Google Scholar 

  9. 9

    De Bourdeaudhuij I, Brug J, Vandelanotte C, Van Oost P . Differences in impact between a family- versus an individual-based tailored intervention to reduce fat intake. Health Educ Res 2002; 17: 435–449.

    Article  Google Scholar 

  10. 10

    Frenn M, Malin S, Brown R, Greer Y, Fox J, Greer J, et al. Changing the tide: an Internet/video exercise and low fat diet intervention with middle school students. Appl Nurs Res 2005; 18: 13–21.

    Article  Google Scholar 

  11. 11

    Haerens L, Deforche B, Maes L, Brug J, Vandelanotte C, De Bourdeaudhuij I . A computer tailored fat intake intervention among adolescents: results of a randomised controlled trial. Ann Behav Med 2007; 34: 253–262.

    Article  Google Scholar 

  12. 12

    Haerens L, Deforche B, Maes L, Cardon G, Stevens V, De Bourdeaudhuij I . Evaluation of a 2-year physical activity and healthy eating intervention in middle school children. Health Educ Res 2006; 21: 911–921.

    Article  Google Scholar 

  13. 13

    Nigg CR, Allegrante JP, Ory M . Theory-comparison and multiple-behaviour research: common themes advancing health behaviour research. Health Educ Res 2002; 17: 670–679.

    Article  Google Scholar 

  14. 14

    Schuit AJ, Van Loon JM, Tijhuis M, Ocke MC . Clustering of life style risk factors in a general adult population. Prev Med 2002; 35: 219–224.

    Article  Google Scholar 

  15. 15

    Moreno LA, Gonzalez-Gross M, Kersting M, Molnar D, De Henauw S, Beghin L et al., on behalf of the HELENA Study Group. Assessing, understanding and modifying nutritional status, eating habits and physical activity in European adolescents. The HELENA Study. Public Health Nutr 2008; 11: 288–299.

    CAS  Article  Google Scholar 

  16. 16

    Brug J, Campbell M, van Assema P . The application and impact of computer-generated personalized nutrition education: a review of the literature. Patient Educ Couns 1999; 36: 145–156.

    CAS  Article  Google Scholar 

  17. 17

    Gedrich K, Hensel A, Binder I, Karg G . How optimal are computer-calculated optimal diets? Eur J Clin Nutr 1999; 53: 309–318.

    CAS  Article  Google Scholar 

  18. 18

    Vandelanotte C, Matthys C, De Bouredeaudhuij I . Reliability of a computerized questionnaire to measure fat intake in Belgium. Nutr Res 2004; 24: 621–631.

    CAS  Article  Google Scholar 

  19. 19

    The Federal Food Code and Nutrient Data Base (BLS II.2). Conception, Structure and Documentation of the Data Base Blsdat. BgVV: Berlin, 1998.

  20. 20

    NEVO Tabel, Nederlands Voedingsstoffenbestand (Dutch Food Composition Table). NEVO Foundation: Zeist, 2001.

  21. 21

    Belgische Voedingsmiddelentabel, tweede uitgave (Belgian Food Composition Table, 4th edn). NUBEL, Ministerie van Volksgezondheid: Brussels, 2004.

  22. 22

    Steinel M . Models for determining optimal nutrition in private households. Z Ernahrungswis 1993; 32: 79–92.

    CAS  Article  Google Scholar 

  23. 23

    Kafatos A, Codrington CA . Nutrition and diet for healthy lifestyles in Europe; the EURODIET project. Public Health Nutr 2001; 4 (S2A): 265–336.

    Google Scholar 

  24. 24

    German Nutrition Society, Austrian Nutrition Society, Swiss Society for Nutrition Research, Swiss Nutrition Association. D-A-CH Reference Values for Nutrient Intake. Umschau/Braus: Frankfurt/M, 2002.

  25. 25

    WHO. Diet, nutrition and the prevention of chronic diseases. Report of the joint WHO/FAO expert consultation. WHO Technical Report Series, No. 916 (TRS 916), 2003.

  26. 26

    Leclercq C, Valsta LM, Turrini A . Food composition issues-implications for the development of food-based dietary guidelines. Public Health Nutr 2001; 4: 677–682.

    CAS  PubMed  Google Scholar 

  27. 27

    West CE, van Staveren WA . Food consumption, nutrient intake, and the use of food composition tables. In: Margetts B, Nelson M (eds). Design Concepts in Nutrition Epidemiology, 2nd edn. Oxford University Press: Oxford, 1997, pp 107–122.

    Google Scholar 

  28. 28

    FAO/WHO. Preparation and use of Food-Based Dietary Guidelines. Report of a Joint FAO/WHO Consultation, World Health Organisation, Geneva, WHO Technical Report Series 880, 1998.

  29. 29

    Food and Agriculture Organization/World Health Organization. Preparation and Use of Food-based Dietary Guidelines. Report of a Joint FAO/WHO Consultation. WHO Technical Report Series no. 880. WHO: Geneva, 1998.

Download references

Acknowledgements

The HELENA Study takes place with the financial support of the European Community Sixth RTD Framework Programme (Contract FOOD-CT-2005-007034). The content of this study reflects only the authors’ views and the European Community is not liable for any use that may be made of the information contained therein. Carine Vereecken is postdoctoral researcher funded by the Research Foundation Flanders (FWO).

Author information

Affiliations

Authors

Consortia

Corresponding author

Correspondence to L Maes.

Additional information

Conflict of interest

The authors state no conflict of interest.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Maes, L., Vereecken, C., Gedrich, K. et al. A feasibility study of using a diet optimization approach in a web-based computer-tailoring intervention for adolescents. Int J Obes 32, S76–S81 (2008). https://doi.org/10.1038/ijo.2008.186

Download citation

Keywords

  • adolescence
  • nutrition
  • diet optimization
  • tailoring
  • HELENA

Further reading

Search