The number of secondary-school students with overweight and obesity is increasing worldwide(Reference Summerbell, Water and Edmunds1). During the years 2002–2004, in a survey of 15-year-old Dutch students, 20·1 % of female and 16·8 % of male students suffered from overweight, whereas 4·7 % and 3·9 %, respectively, were diagnosed with obesity(Reference van den Hurk, van Dommelen and de Wilde2). Overweight and obesity in young people are diagnosed using the international BMI curves for age and sex(Reference Bulk-Bunschoten, Renders and van Leerdam3).
The increase in overweight and obesity is a worrying development because overweight, and particularly obesity, can have detrimental effects both at an early age and in the long term. Effects at a young age may include psychosocial stress, low self-esteem, skeletal abnormalities, fatigue and glucose intolerance. In the long term, the consequences may be CVD, diabetes, arthritis, high blood pressure and cancer(Reference Breetvelt4).
Overweight and obesity are caused by insufficient physical activity and unhealthy dietary habits in more than 90 % of cases(Reference Verheul-Koot5, Reference Renders, Seidell and van Mechelen6). One of the most unfavourable trends in the dietary behaviour of secondary-school students, a trend that exacerbates the problem of overweight, is the decreasing consumption of fruit and vegetables(Reference Renders, Seidell and van Mechelen6). Less than half of secondary-school students in the Netherlands consume the recommended daily amount of fruit and vegetables(Reference van Kooten, de Ridder and van Dorsselaer7). Habits that develop during the secondary-school phase have an important role in the development of health-related behaviour in adulthood(Reference Frenn, Malin and Bansal8). Cognitive motivational processes strongly influence secondary-school students’ eating habits, whereas younger children choose foods mainly on the basis of their preferences(Reference Contento, Koch and Lee9).
Self-efficacy has a central role in the cognitive regulation of motivation(Reference Bandura10). Self-efficacy proves to be an important predictor of many health-related phenomena, such as smoking, weight control, diet, alcohol use and AIDS prevention(Reference Bandura11–Reference Yalow and Collins17). The term ‘self-efficacy’ was used for the first time in 1977 by the American psychologist Albert Bandura(Reference Bandura18). Bandura described self-efficacy as ‘…people's judgement of their capabilities to organise and execute courses of action required to attain designated types of performances’(Reference Bandura11). This description shows that people's self-efficacy is not of a general nature, but is related to specific situations(Reference Shortridge and van der Bijl19). A person can judge himself as competent to eat fruit at home, whereas he may not judge himself as competent to eat fruit at school.
The higher the level of self-efficacy, the higher the likelihood that people will be able to display the desired behaviour(Reference Bandura11). The level of self-efficacy can be influenced by targeted interventions that are developed on the basis of the theoretical background. However, before interventions can be developed to increase self-efficacy with regard to fruit and vegetable consumption, it is important to know whether self-efficacy is actually a predictor of fruit and vegetable consumption among secondary-school students. A literature review proves that this cannot be ascertained yet. A significant positive relationship was found in three(Reference Zabinski, Daly and Norman20–Reference Young, Fors and Hayes22) out of five studies(Reference Zabinski, Daly and Norman20–Reference Martens, van Assema and Brug24) among secondary-school students in which the relationship between self-efficacy and consumption of fruit and vegetables was studied. Neumark-Sztainer et al.(Reference Neumark-Sztainer, Wall and Perry21) are the only ones reporting a correlation coefficient. Between the ‘self-efficacy for making healthy dietary choices’ and ‘fruit and vegetable consumption’, a correlation of 0·18 was found.
Reliable and valid instruments are required to expand research on self-efficacy as a predictor of fruit and vegetable consumption and thus gain more insight into the relevance of developing interventions for increasing self-efficacy. No instruments were found in the literature for measuring self-efficacy for vegetable consumption among Dutch secondary-school students, and only one instrument was found for measuring the self-efficacy for fruit consumption. This instrument was made up of a single item(Reference Martens, van Assema and Brug24). Thus, valid and reliable instruments for measuring ‘fruit self-efficacy’ (FSE) and ‘vegetable self-efficacy’ (VSE) among secondary-school students are lacking in the Netherlands. Because of the specificity of the self-efficacy concept, separate instruments must be developed for FSE and VSE(Reference Mailvil, Lawson and Horwath25).
The first aim in the present research was the development of two measurement instruments for the Dutch situation: one instrument to measure FSE among secondary-school students in the Netherlands, another to measure VSE among this target group. The second aim was to test the reliability and validity of the developed instruments.
Method
Design
The present research can be characterised as methodological research(Reference Polit and Beck26). The Lynn procedure(Reference Lynn27) was followed in the development of the two self-efficacy instruments to achieve sufficient content validity. This procedure entailed that the dimensions of the concept that was to be measured were identified in the development phase. Items were then formulated for these dimensions. In the final phase, the content validity of the items and instruments were assessed.
After the instruments were developed, both instruments were assessed for construct validity (using factor analysis), internal consistency, temporal stability and predictive validity.
Developing the instruments
Dimensions of self-efficacy
Bandura(Reference Bandura28) describes how self-efficacy can be measured in three dimensions, namely ‘generality’, ‘strength’ and ‘level’. ‘Generality’ refers to the degree to which self-efficacy is positively related within a behavioural domain, between different behavioural domains or in relation to time(Reference Maibach and Murphy13). ‘Strength’ refers to how certain a person is of being able to perform a certain activity(Reference Bandura28). ‘Level’ refers to the number of activities with an increasing difficulty level that a person thinks he/she can carry out(Reference Bandura28). All three dimensions were measured in the present research, whereby the dimension ‘generality’ was measured within and between the behavioural domain ‘fruit’ and the behavioural domain ‘vegetables’.
To ensure a reliable assessment of the ‘strength’ dimension, the secondary-school students scored their own self-efficacy on a 4-point Likert scale. An asymmetrical 4-point scale was chosen for the present research: 1 = ‘uncertain’, 2 = ‘almost certain’, 3 = ‘certain’, 4 = ‘very certain’. This simple scale was chosen because of the target group of secondary-school students who could have poor reading skills. Furthermore, this asymmetrical answer scale has relatively more positive answers. Research has shown that, when people are asked to grade their self-efficacy, they are more likely to think that they ‘will’ be able to do something rather than that they ‘will not’(Reference Schaalma29).
In English, self-efficacy items are typically formulated with the phrase ‘How confident are you at…’. In Dutch, various formulations are used, because the Dutch language does not have an exact equivalent of the concept ‘confident’. Commonly used phrases in the Netherlands are: ‘I think I'm able to…’, ‘Do you succeed in…’, ‘To what extent do you estimate it realisable to…’, ‘I'm sure/certain that I can…’, or ‘I'm convinced I can…’(Reference van der Bijl and Shortridge30). Six secondary-school students (preparatory secondary vocational education) were asked for their preference. The phrase ‘I'm certain that I can…’ was chosen on this basis.
Generating items
When measuring self-efficacy, difficult situations for the students to eat fruit or vegetables must be included in the instrument to ensure a reliable assessment of the dimension ‘level’. Sufficient difficult situations must be included to preclude ceiling effects(Reference Bandura31). To reliably assess the dimension ‘generality’, the behavioural factors related to the behavioural domain ‘fruit’ and the behavioural domain ‘vegetables’ must be included(Reference Bandura31), such as buying and eating fruit and vegetables.
Two existing American instruments were used to identify behavioural factors and difficult situations that have a role in fruit and vegetable consumption(Reference Thombs and Heatey32, Reference Baranowski, Domel and Gould33). On the basis of these instruments, twenty-three ‘fruit items’ and twenty-three ‘vegetable items’ were selected. To gain more insight into the difficult situations for secondary-school students in the Netherlands, these items were assessed in a pilot study for twenty-three secondary-school students using an opportunity sample (average age = 14·6 (sd 1·7) years). These students were asked to score the difficulty of carrying out the behaviour for each item using a 5-point Likert scale, from 1 = easy to 5 = difficult. One item was then removed from each of the instruments for which all the respondents gave the answer ‘easy’. These two items had no discriminatory value.
Students were also asked to make suggestions for improvement of each item, and they could indicate whether items were lacking. Subsequently, on the basis of the students’ feedback about the Dutch situation, three fruit and two vegetable items were removed, three fruit and two vegetable items were added and five fruit and six vegetable items were modified, as a result of which both the FSE and VSE instruments consisted of twenty-two items.
Assessing content validity
Thirteen Dutch and Belgian researchers were approached who had all published on self-efficacy in relation to school students’ dietary habits. These experts were asked to score the twenty-two items of the FSE instrument and the twenty-two items of the VSE instrument for relevance in relation to the concept of ‘self-efficacy’ on a 4-point scale, from 1 = not relevant to 4 = very relevant(Reference Lynn27). The experts were also able to make suggestions for improvement of each item, and they could indicate whether items were lacking.
Testing the instruments
Research participants
Twenty-three secondary schools throughout the Netherlands were approached to participate in the research. Seven schools were willing to cooperate, and were located in four cities in the middle and one city in the south of the Netherlands. The size of the schools ranged from about 650 students to 1850 students.
Eight teachers at the seven secondary schools were asked to participate. Students at these schools were recruited using an opportunity sample. The teachers were able to indicate which class or classes they considered most suitable for participation. This increased the chances of an individual teacher being willing to participate. The research strove for as much variety as possible in class levels and years. A total of twenty-one classes participated. Both self-efficacy instruments were administered among these secondary-school students. The students simultaneously completed an ‘alcohol self-efficacy’ instrument for a similar research project. The order in which the students completed the three self-efficacy instruments varied. Students from four teachers started with the FSE and VSE instruments, students from the other four teachers started with the ‘alcohol self-efficacy’ instrument. This was to reduce the influence of completion order on the results; the students might, after completing one or two instruments, become less motivated to fill in a third.
The determining factor for the sample size was the factor analysis within the framework of the construct validation. The common rule of thumb, ‘at least ten times as many respondents as items’, was applied(Reference Stokking34). Because the final versions of the two self-efficacy instruments consisted of seventeen items each (see ‘Content validity’ under ‘Results’ section), the minimum sample size was 170 students.
Fruit and vegetable consumption questionnaire
Fruit and vegetable consumption was measured to determine the predictive validity of the self-efficacy instruments. Consumption was measured by questioning students on how often they ate fruit and vegetables during a normal week. The seven response options of both questions ranged from ‘never’ to ‘more than once per day’. These two questions were taken from the international questionnaire prepared by the Health Behaviour in School-aged Children organisation.
The test–retest reliability of the consumption questions was assessed in another study and was 0·68 for fruit and 0·59 for vegetables. The correlation between the above-mentioned questions and a 7 d diary was 0·34 for fruit and 0·48 for vegetables. A detailed description of the reliability and validity can be read elsewhere(Reference Vereecken and Maes35).
Procedure
The researcher and/or the students’ teacher was present while the instruments were being completed in the classroom. Teachers could choose between a written and a digital version. This was to increase the chance of teachers participating.
To determine the temporal stability of both self-efficacy instruments, after 28 d the students were asked to complete the self-efficacy instruments again within a period of 2 weeks. The students of four classes completed the instruments in the classroom a second time. The students in the other seventeen classes were allowed to decide for themselves where and when they would complete the instruments digitally, thus limiting the amount of time these schools had to invest.
Ethical aspects
All secondary-school students and the parents of students up to the age of 16 years received written information about the research. They could decline to cooperate before the research started. Students and parents were reassured that the anonymity of the students would be guaranteed. The Medical Ethical Assessment Committee of Utrecht University Medical Center gave permission for conducting the present research.
Statistical analysis
All data were analysed using the Statistical Package for the Social Sciences statistical software package version 16·0 (SPSS Inc., Chicago, IL, USA).
Content validity
Both the ‘item-content validity index’ (I-CVI) and the ‘scale-content validity index’ (S-CVI) were calculated. The I-CVI was determined by the proportion of experts who assessed an item as content valid (score: 3 or 4). The S-CVI was determined by the average proportion of items that scored 3 or 4(Reference Polit and Beck36).
Construct validity
Data were analysed using the Principal Components Analysis (PCA). The following criteria were used for determining the number of factors: factors must have an eigenvalue of >1·00; factors are included up to the point that their eigenvalue curve flattens (calculated with a screeplot)(Reference Stokking34); and factor loadings must have an absolute value of >0·40(Reference Polit and Beck26).
Internal consistency
Cronbach's α was calculated to determine internal consistency.
Temporal stability
Stability was determined using the single-measure intra-class correlation coefficient (ICC; one-way random model). According to Cicchetti's guidelines(Reference Cicchetti37), an ICC is poor when it has a value <0·40, fair when it is between 0·40 and 0·60, good >0·60 and excellent >0·75.
Predictive validity
Spearman's rank correlation was used to calculate the correlations between self-efficacy and consumption, as measured at the ordinal level.
Results
Content validity
Four out of thirteen experts scored the relevance of the items. According to Lynn's method(Reference Lynn27), if there are four experts, then all of them must score an item as 3 or 4 to be able to class that item as content valid. On this basis, eight items in the FSE instrument and seven items in the VSE instrument could not be classed as content valid. Of these, six items were removed from both instruments that, according to the experts, were not relevant to fruit or vegetable consumption, or were not related to self-efficacy. The other three non-content valid items were reviewed, and three items were added to each instrument on the basis of the experts’ suggestions. Both instruments were then sent to the experts for re-assessment. After this re-assessment, both instruments still contained two items that could not be classed as content valid. The experts indicated that these two items were not relevant to fruit and vegetable consumption and these items were removed. The S-CVI of both instruments at the end of this procedure was 1·0, which is a reliable content validity score(Reference Polit and Beck36). The final version of each instrument contained seventeen items.
Results of testing the instruments
Research participants
In the present research, 466 students took part (Table 1). Only one student did not take part because his parents objected to his participation.
Table 1 Demographic characteristics of the research participants for time 1 (n 466) and time 2 (n 106)
PSVE, preparatory secondary vocational education; SGSE, senior general secondary education; PUE, pre-university education.
†Some items are not equal to n 466 or n 106 because of missing values.
‡Forty ethnic groups.
§Eleven ethnic groups.
The students were 11–19 years of age, and 50·6 % were male students. All secondary-school levels recognised in the Netherlands were included. The majority of respondents were ethnic Dutch (80·7 %). This percentage is in line with the percentage of ethnic Dutch secondary-school students in the 2007–2008 school year (79·2 %). However, other widely represented ethnic minorities in the Netherlands were strongly under-represented in this sample. These include Surinamese, Moroccan, Turkish, Antillean and Aruban students(38).
Both self-efficacy instruments were completed a second time by 106 out of 466 students. Fifty-eight students completed the instruments again in the classroom, forty-eight students completed them at their own initiative.
Construct validity
The PCA of the FSE and VSE instruments produced an eigenvalue of >1·00 for the first two factors (Tables 2 and 3). However, for the first factor, all factor loadings had a value of >0·40. Moreover, only one item (in the VSE instrument) had a higher factor loading for factor 2 (0·57) than for factor 1 (0·54). On the basis of screeplots, one factor proved to be a suitable choice for both instruments. Both instruments can therefore be characterised as unidimensional scales, which means that both instruments measure a single construct. The first factor accounts for 50·3 % of the variance in the FSE instrument and 53·9 % of the variance in the VSE instrument.
Table 2 Factor structure and item scores for the fruit self-efficacy instrument
†Response options were 1 = uncertain, 2 = almost certain, 3 = certain, 4 = very certain.
Table 3 Factor structure and item scores for the vegetable self-efficacy instrument
†Response options were 1 = uncertain, 2 = almost certain, 3 = certain, 4 = very certain.
Internal consistency
The Cronbach's α of the FSE instrument was 0·94 and of the VSE instrument 0·95. On the basis of the analysis of items at the item level, according to the ‘α if item deleted’ procedure, removing an item did not have a great influence on the α level (maximum increase of 0·001).
Further research on the inter-item correlation is required to properly interpret the high α. A rough indication of the internal consistency of a unidimensional instrument is that the average inter-item correlation must be considerably higher than 0·20, and Cronbach's α must be at least 0·75(Reference de Jonge and Hamers39). The average inter-item correlation of the FSE instrument was 0·46 and of the VSE instrument 0·50.
Temporal stability
Because the conditions in which the students completed the instruments for the second time varied, subgroups were formed for calculating the ICC. Students completed the instruments the second time in the classroom (Table 4) or at their own initiative (Table 5). A distinction was also made in the order in which the self-efficacy instruments were completed. As with the first test, students started either with the FSE and VSE instruments or with the ‘alcohol self-efficacy’ instrument. The ICC of the two self-efficacy instruments varied between 0·33 and 0·84 (P < 0·05) in the subgroups.
Table 4 Test–retest reliability of the FSE and VSE instruments (completed in the classroom; ICC)
FSE, fruit self-efficacy instrument; VSE, vegetable self-efficacy instrument; ICC, intra-class correlation coefficient.
*P < 0·05; **P < 0·01.
†Questionnaire started with the FSE and VSE instruments and ended with the ‘alcohol self-efficacy’ instrument.
‡Questionnaire started with the ‘alcohol self-efficacy’ instrument and ended with the FSE and VSE instruments.
Table 5 Test–retest reliability of the FSE and VSE instruments (completed at the students’ own initiative; ICC)
FSE, fruit self-efficacy instrument; VSE, vegetable self-efficacy instrument; ICC, intra-class correlation coefficient.
*P < 0·05; **P < 0·01.
†Questionnaire started with the FSE and VSE instruments and ended with the ‘alcohol self-efficacy’ instrument.
‡Questionnaire started with the ‘alcohol self-efficacy’ instrument and ended with the FSE and VSE instruments.
Predictive validity
Spearman's rank correlations between FSE and consumption and between VSE and consumption varied between 0·26 and 0·41 (P < 0·01; Table 6).
Table 6 Predictive validity of the FSE and VSE instruments (Spearman's rank correlation)
FSE, fruit self-efficacy instrument; VSE, vegetable self-efficacy instrument.
**P < 0·01.
†Refers to the level in which self-efficacy is positively correlated within the fruit behavioural domain or the vegetable behavioural domain (‘generality’).
‡Refers to the level in which self-efficacy is positively correlated between the fruit behavioural domain and the vegetable behavioural domain (‘generality’).
Discussion
The purpose of the present research was to test the reliability and validity of the developed FSE and VSE instruments for secondary-school students.
Reliability
Reliability is determined by calculating internal consistency and temporal stability. Both the FSE and VSE instruments display good internal consistency. The fact that the Cronbach's α of both instruments is very high (>0·90) could indicate that a number of items pose the same question in a slightly different way(Reference Streiner and Norman40). Therefore, a reduction in the number of items is worth the benefit of lowering respondent burden. For example, the content of items 3, 7 and 14 of the FSE instrument (Table 2) seems to overlap with that of item 6, and the content of items 5, 9 and 11 with that of item 1. If these six items are removed, α is 0·90. The content of items 3 and 17 of the VSE instrument (Table 3) seems to overlap with that of item 14, and that of items 7, 9 and 10 with item 1. If these five items are removed, α is 0·93. The final decision regarding the definite deletion of items of both instruments should be taken by the expert panel.
This is the first time that the stability of an FSE or VSE instrument is being studied for secondary-school students or for young people in this age group. The order in which the self-efficacy instruments were completed seems to have had a larger influence on stability than the method used to gather the data (Tables 4 and 5). The students who completed the ‘alcohol self-efficacy’ instrument first scored lower for the stability of the FSE and VSE instruments than the students who started with the FSE instrument. A possible explanation is that the students started to tire of the questionnaires, and were unable to concentrate properly on the last instruments(Reference Polit and Beck26). However, it is important to mention here that there were not many students in the subgroups, and thus conclusions must not be drawn too hastily.
On the basis of the guidelines of Cichetti(Reference Cicchetti37), the ICC of the self-efficacy instruments used in the present research are fair to excellent for students who started by completing the FSE (0·70 and 0·84) and VSE instruments (0·57 and 0·70). The ICC is poor to fair for students who completed the FSE (0·35 and 0·41) and VSE instruments (0·33 and 0·47) last.
Validity
Content validity, construct validity and predictive validity were all calculated. The content validity of both instruments was thoroughly examined by experts with knowledge of the concept of self-efficacy and fruit and vegetable consumption among young people. The content validity of the research would thus appear to be sufficiently safeguarded.
Research on the factors used in an FSE or VSE instrument among secondary-school students or young people in this age group has not been described previously in the literature. In previous research of 10–11-year-old primary-school students(Reference Thompson, Bachman and Baranowski41) and adults(Reference Ma, Betts and Horacek42, Reference de Vet, de Nooijer and de Vries43), the FSE and VSE instruments were characterised as unidimensional, just as in the present research.
The predictive validity of both instruments is moderate. In comparison with the research by Neumark-Sztainer et al.(Reference Neumark-Sztainer, Wall and Perry21) – the only study of secondary-school students that describes the correlation coefficient – the correlations in the present research are higher. In the research by Neumark-Sztainer et al.(Reference Neumark-Sztainer, Wall and Perry21), a correlation of 0·18 was found between the ‘self-efficacy for making healthy dietary choices’ and fruit and vegetable consumption.
It is striking that the present research is the first on secondary-school students in which self-efficacy is measured specifically for vegetables, and only the second in which it is measured specifically for fruit. Other studies measured such phenomena as ‘self-efficacy for making healthy dietary choices’. It is possible that a relationship with fruit and vegetable consumption was not always found in earlier research among secondary-school students because of this.
The correlations in the present research are comparable with research on 10–12-year-old primary-school students(Reference Thompson, Bachman and Baranowski41, Reference Vereecken, van Damme and Maes44–Reference Wind, de Bourdeaudhuij and te Velde46). The correlation between FSE and fruit consumption was 0·41 in the present research, whereas among primary-school students it varied between 0·19 and 0·42. The correlation between VSE and vegetable consumption was 0·32 in the present research, whereas among primary-school students it varied between no correlation in one study and 0·33. These results suggest that the self-efficacy of secondary-school students does not have, or hardly has, more influence on fruit and vegetable consumption than the self-efficacy of primary-school students. These results also lead one to surmise that self-efficacy has an influence on fruit and vegetable consumption at an earlier stage than at secondary-school level. Another salient finding is that the relationship between FSE and fruit consumption is stronger than that between VSE and vegetable consumption. The explanation for this may be that young people have more influence on what they consume between meals, such as fruit, than on what they consume at main meal times, such as vegetables(Reference Vereecken, van Damme and Maes44).
On the basis of the poor correlation found between self-efficacy and fruit and vegetable consumption in the present research, it is debatable whether there is any value in designing interventions for increasing FSE and VSE. Self-efficacy increasing interventions may well have only a slight effect on fruit and vegetable consumption.
Limitations
Because the instruments were completed in the classroom, it is conceivable that the students influenced each other's choices. The students were spaced apart to prevent this as much as possible. During the completion of the instruments a second time, this ‘separation’ was not possible for seventeen classes. The students were allowed to decide for themselves where and when they would complete the instruments digitally. This has also influenced the number of responses negatively.
Furthermore, alongside the FSE and VSE instruments, the students also had to complete an ‘alcohol self-efficacy’ instrument. Because some students may have become demotivated after completing one or two instruments, this may have influenced the results.
Another limitation is that many ethnic minorities widely represented in the Netherlands were strongly under-represented.
Practical implications
Both instruments can be used for researching self-efficacy as a determinant of fruit and vegetable consumption, and for evaluating the effects of nutrition education programmes on secondary-school students. The two instruments can be used independently or in combination. It is important not to offer too many instruments to the students at the same time, as this could influence the reliability of the results.
Conclusion and recommendations
The FSE and VSE instruments are sufficiently reliable and valid. The instruments can be used to measure FSE and VSE among Dutch secondary-school students. Further research, in which the predominant ethnic minorities in the Netherlands are sufficiently represented, is required to confirm this conclusion.
Despite the fact that increasing FSE and VSE can have a small effect on fruit and vegetable consumption, more research on the effect of self-efficacy increasing interventions is required to understand this better. Such experimental research can provide more evidence of a causal relationship than can cross-sectional research.
Furthermore, following research should preferably measure fruit and vegetable consumption by means of observation or ‘24 h recall’ questionnaires(Reference van Kooten, de Ridder and van Dorsselaer7). This would allow more accurate measurement of consumption, and the relationship between consumption and self-efficacy could be more precisely studied.
Acknowledgements
The present research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. The authors have no conflict of interest. R.B. designed the study and carried out all aspects of data collection and statistical analyses; J.J.v.d.B. assisted with all aspects of the study. The authors thank the teachers and experts for their participation in the present research.
