Diet and Eating behaviour

4.2.5.1 Dietary Fat and Fibre

Levels of dietary fat and fibre were measured using a modified version of the Dietary Instrument for Nutrition Education (DINE: Roe et al 1994) This instrument was designed to be administered by interview, and some slight alterations were made to the language and format o f the questions in order to make it suitable for self administration by adolescents. The DINE produces a score for levels of dietary fat and fibre based on usual weekly consumption of key foods. A saturated fat score, which is also part o f the DINE, was omitted as piloting of the questionnaire suggested that it was unsuitable for use with adolescents, due to its emphasis on preparing food. The DINE has been validated against detailed four-day diet records (Roe et al 1994) producing acceptable correlations (0.51 and 0.46 for fat and fibre respectively). For some analyses fat and fibre scores were dichotomised using a median split.

Chapter 4. Baseline methodology and descriptive statistics

4.2.5.2 Fruit and vegetable intake

Daily fruit and vegetable intake were assessed with two questions; 'How many servings o f fruit (fresh frozen or tinned) do you eat in a usual day?' and 'How many servings o f vegetables (fresh frozen or tinned, not including potatoes) do you eat in a usual day?'. Response options were: Hess than one’, ‘l - 2 \ ‘3-4’, ‘5 or more ’, and 7 never eat fruit ’ or 7 never eat vegetables ’. The response options were given scores of 0.5, 1.5, 3.5, 5 and 0, to represent the estimated number of servings per day, and the fruit and vegetables scores were combined to provide a single daily fruit and vegetables score. Single item assessments of fruit and vegetable intake have been used in other studies o f dietary behaviour, and found to have good test-retest reliability (Wardle, Parmenter and Waller 2000; Baker and Wardle in press). For many analyses a dichotomised version of this variable was used, based on whether or not participants ate the recommended five portions of fruit and vegetables per day.

4.2.5.3 Healthv eating habits

Healthy eating habits were assessed using the Adolescent Food Habits Checklist (AFHC), which was developed for use in this research (Johnson et al in press). This 23-item scale assesses adolescents' involvement in specific healthy eating practices with regard to consumption of energy dense foods (e.g. I f I am buying crisps, I often choose a low-fat brand) and fruit and vegetables (e.g. I make sure I eat at least one serving o f fruit a day). It specifically addresses food choice situations in which adolescents are likely to have a high degree of personal choice. The validation of this new scale is detailed below and further details can be found in Johnson et al (in press).

Validation: Item selection Items were selected for the AFHC on the basis of findings from a pilot study carried out with 178 adolescent girls attending an independent girls’ school in the North West of England (mean age 15 years 10 months). A preliminary pool of seventy items for the AFHC were generated with reference to existing literature, and dietary health recommendations and in discussion with health psychologists and nutritionists. Participants were asked

Chapter 4. Baseline methodology and descriptive statistics

to reply "true ’ or "false ’ or "not applicable to me" with regard to whether they usually followed specific dietary practices. These practices included the purchase, preparation and consumption o f specific foods, as well as snacking habits. Items referred to both healthy and unhealthy behaviours. Participants were also asked to add any other things that they regularly did in order to make their diet more healthy.

Validation: analysis o f pilot data Responses to the pilot questionnaire were analysed first using factor analysis with varimax rotation in order to establish whether there was a multidimensional structure underlying the patterning o f food habits. Results from this analysis suggested a weak factor structure. A five- factor solution accounted for just 32 percent o f the variance, and intercorrelations between factors were high. Internal reliability for the item pool as a whole was good (Cronbach's a = .91). In light of the weakness of the factors, and other evidence that healthy eating patterns often do not form stable, replicable factors (Birkett and Boulet, 1995; Prewitt et al, 1997), items for the final scale were selected according to other criteria. It was decided to limit the scale to items pertaining to intake of fruit, vegetables and energy-dense foods. Items with a low item-total correlation (r <.20) and those which made reference to situations likely to be unfamiliar to adolescents were omitted fi"om the scale. Four items referring to general aims to eat a diet that is low in fat, low in sugar, high in fruit and vegetables, and healthy were retained. No items were added in response to comments from the pilot sample as no widely used practices emerged from these comments. This resulted in a 23-item scale, which had an internal reliability o f cronbach's a = 0.83 in the pilot sample. A true/false response format was selected to make the checklist easier to complete. Ten items also had an alternative response, equivalent to ‘not applicable’. Participants received one point for each ‘healthy’ response. The final score was adjusted for 'not applicable' and missing responses using the formula: AFHC score = no of 'healthy' responses * (23 / no of items completed).

Chapter 4. Baseline methodology and descriptive statistics

Validation: Test-retest reliability The test-retest reliability o f the 23-item AFHC was examined using a sample of 24 adolescents aged between 13-14 years (mean age 13 years 8 months). Participants completed the AFHC twice, with a delay of two weeks between the two completions. The correlation between score at T1 and score at T2 was very high (r = 0.90 p<0.001).

Validation: Internal reliability and external validity o f the AFHC. Internal reliability and external validity of the AFHC were assessed using data from the full sample of 1177 girls. Reliability o f the AFHC study was high (Cronbach's a =0.82), and similar to that found in the pilot sample. In order to assess the convergent validity of the AFHC, a number o f hypotheses were generated with regard to the associations between AFHC score and scores on other related measures. The measures used to validate the AFHC (which are described in more detail elsewhere in this section) are: Dietary fat and fibre intake (DINE: Roe et al 1994), Daily intake of fruit and vegetables (details in section 4.2.5.2), Dietary restraint (DEBQ-restraint: Van Strien et al 1986) and Nutrition Knowledge (NKQ: Parmenter and Wardle 1999).

Many o f the items of the AFHC refer to low-fat eating behaviours, and so a strong, negative correlation was expected between dietary fat intake and AFHC score. Similarly, the relationship between AFHC score and daily fruit and vegetable intake was hypothesised to be strongly positive. Items on the checklist are relevant to dietary fibre intake through questions on fruit and vegetable consumption, and so a positive but weaker correlation was also predicted between dietary fibre and AFHC score. The major role played by healthy eating in weight control (Nichter et al 1995) meant that AFHC was hypothesised to be positively associated with dietary restraint. Finally, nutrition knowledge has been linked with a more healthy diet in some studies (Wardle, Parmenter and Waller 2000), and such an association was predicted here.

Correlations between the AFHC and other measures are shown in table 4.7. As predicted a strong negative correlation was observed between AFHC score and

Chapter 4. Baseline methodology and descriptive statistics

levels of dietary fat. Similarly, daily fruit and vegetable intake and AFHC score was strongly associated. The correlation with dietary fibre was less strong but still highly significant and in the predicted direction. Dietary restraint was positively associated with healthy habits, and those participants who had a higher level of knowledge about dietary health and nutrition engaged in more healthy practices. In the light of these results the AFHC was considered to be a useful tool for examining aspects o f healthy eating linked to patterns of behaviours. AFHC scores were dichotomised for some analyses using a median split.

Table 4.7 Correlations between AFHC and validation measures

Scale r

N -1177

Fruit and Vegetable Intake 0.44 DINE - Dietary Fat -0.41 DINE - Dietary Fibre 0.18 DEBQ - Dietary Restraint 0.39 Nutrition Knowledge Questionnaire 0.14 All correlations significant at p<0.001