3.3,3 Procedure
3.3.4.1 Independent variables Motivation Phase
Appendix 3 is a copy of the full baseline questionnaire which includes all the questions measured at all three time points. The measures described below thus give references to this questionnaire by giving the section and numbers where each variable is situated in the questionnaire, e.g. (S2; 1,4) refers to Section 2, items 1 and 4.
3.3.4.1 Independent variables Motivation Phase
Name of
variable Description / examples of items No.of items Response Scale Rangeof possible scores Internal Reliability Cronbach a Threat: 8-392 Perceived susceptibility (S2; 1,4) X
My physical health makes it likely that
I will have a heart attack 2 7-point .72
Perceived Severity
(S2: 2,3,5,7)
Having a heart attack would ruin my
chances of future happiness) 4 7-point .69
Outcome expectancies'^ (83:1,3,5,6,7,8, 10,11,12)
Making some changes to my lifestyle would help protect me from heart disease
My family and close friends would encourage me to make some changes to my lifestyle
Making some changes to my lifestyle would help me feel better and more alive
9 7-point 9-63 .73
Self-efficacy^ (S3: 2,9)
I am confident that I would be able to make changes to my lifestyle if I wanted to
2 7-point 2-14 .87 Intention'^
(S3:13,14)
I intend to make a change in my lifestyle in the next two months to reduce my risk of heart disease
2 5-point 2-10 .95
^ indicates reverse coding of variable, i.e. high score indicates low extent of variable
Chapter 3
Action Phase
The action plans described below have been developed using the concepts of Schwarzer (1992), Bagozzi (1992) and Gollwitzer (1993).
Name of
variable Description / examples of items
No. of items
Response
Scale Range of possible scores Internal Reliability Cronbach a Decision to change (S3: 18)
Have you actually made a decision to make one or more changes in your lifestyle in the next two months to reduce your risk o f heart disease?
1 yes/no 0-1 — Wrote change
(S3: 20)
Subjects asked to write down what particular change they would like to make in the next two months. This variable rated whether they wrote something or not
1 yes/no 0-1
Evaluation of decision (S3: 21-27)
Thinking of the above decision to change, subjects were asked to rate how they felt about doing it:
How easy would this be for you to do? How much effort are you prepared to put into doing this, etc.
7 5-point 7-35 .74
Number of means (S3: 28, 1-3)
Thinking of their decision to change, subjects were asked to write down as many means as they could think of for making this change. 3 blanks were provided. 0-3 Specific means relating to a) exercise, b) diet and c) smoking
a computed variable - did any or all of the subject’s means (see above) relate to the specific behaviour change or not?
1 yes/no 0-1
No. of
implementatio n intentions
(S3: 28, 1-3)
Thinking of each of the means to change, subjects were asked if they had decided when they were going to start doing it. The number of yes’s were totalled. 3 yes/no 0-3 Specific implementation intentions relating to a) exercise, b) diet and c) smoking
Computed variable - did the subject make implementation intentions for means relating to the specific behaviour change
1 yes/no 0-1
There were three separate specific means variables computed, one for each of the outcome behaviours; diet, exercise and smoking. Each was a dichotomous variable relating to whether tlie subject did or did not write down something in tlie means section relating to the specific outcome behaviour.
There were three separate specific implementation variables computed, one for each of tlie outcome behaviours; diet, exercise and smoking. Each was a dichotomous variable relating to whether die subject did or did not make (an) implementation intention(s) for means which were related to the specific outcome behaviour.
Precaution Adoption Process (PAP) Stases (see Appendix 3. Section 1. 1-61 Subjects were assessed as to which stage they had reached in the PAP by asking them to agree or disagree with each of six statements relating to the stages of the model as shown in Chapter 1, Figure 5. The model implies that each stage leads on from the one before, e.g. if a subject agrees to statement 5, they should also have agreed to statements 1-4. Therefore, a new variable was computed called ‘stage’ which gave each subject a stage number from 1-7^^. A subject was allocated to a particular stage if they had agreed to the statement referring to that stage and all the preceding stages, but not to any subsequent stages.
3 3.4.2 Dependent Variables: Measures of health behaviour change
The measures of health behaviour change covered three main areas: diet, exercise and smoking which together encompass most of the behaviours which are
suggested to alter risk of cardiovascular disease. Behaviour change index (Appendix 3, Section 6 )
Subjects were asked to respond to a list of questions relating to each of smoking (4 items), food and drink (12 items) and exercise (4 items). The questions were based on items used in the Scottish Heart Health Study (see Smith et al, 1987). All the questions referred to the past week in order to help focus the subject and to be sensitive to change within the relatively short three-month period of the study.
Smokers were asked three different questions relating to how many cigarettes they smoked in a day (Section 6: A. 1-3). Two of the questions had 5-point (A2) and 7-point (A3) response scales respectively relating to different levels of smoking. The response to question Alb related to the actual number of
cigarettes smoked, i.e. During the past week how many did you smoke a day? The inter-correlations between these three items were high (r=.94, p<.001; r=.72, p<.001 and r=.74, p<.001) which suggested that the measure was reliable. Only
Although the PAP only has 6 stages as shown in tlie introduction, there were 7 stages defining this sample, allowing for an extra stage where a subject may be aware and personally engaged (e.g. may feel susceptible), but have not yet thought about what they could do to reduce their risk. This point will be
________________________________________ Chapter 3________________________________
one of the three questions was therefore used as the measure of extent of smoking. Alb, which was chosen for its greater sensitivity to change. This was decided to be more meaningful than a scale developed by adding the three items together, or taking an average of the three items.
The food and drink items consisted of three sections (Section 6, B):
1. The extent of consumption o f ‘health’ foods (B 1,8) (fruit and vegetables, cereals, whole grain breads, oily fish) (2 items)
2. The extent of consumption o f ‘unhealthy’ foods ( B2-7, 9-10) (high-fat foods and salt) (9 items)
3. The extent of alcohol consumption (Bll,12) (2 items).
Three scores were derived by adding up the items in each of sections 1 and 2^ and by multiplying the items in section 3 (how often did you drink alcohol in the
past week? x how much, on a\>erage, did you drink on each occasion?).
There were four exercise items (Section 6, C), two relating to how active the subjects perceived themselves to be, at work (including work around the home) (Cl) and in their leisure time (C3), added together to create an ‘active’ score, and two relating to how much physical activity they did in the week at work (C2) and in their leisure time (C4), added together to create an exercise’ score. By doing this addition of ‘work’ and ‘leisure’ scores it was possible to gain an idea of overall exercise.
These six ‘behaviour’ scores (defined in bold above) were computed for the subjects’ responses at time 1 (before screening) and at time 3 (2nd time-point after screening). By subtracting the scores at time 1 from the scores at time 3, behaviour change scores were created. The behaviour change scores were recoded into dichotomous variables. The two groups were those who changed for the better (in accordance with health promotion guidelines) and those who did not change for the better (stayed the same or changed for the worse). The scores were computed in such a manner mainly because the main point of interest was to
hi the ‘unhealtliy foods’ section the amount of milk drunk was multiplied by tlie kind of milk drunk (i.e. skimmed, full fat, etc.) and added to the rest of the items)
predict who changed for the better and who did not, but also to maximise the sizes of the groups for analyses.
Finally, a behaviour change index was computed by adding up the six
dichotomous behaviour change variables (coded 0 and 1), giving a maximum score of 6 (six changes for the better) and a minimum of 0 (no changes for the better). This index was used in the analyses as a measure of the number of different behaviour changes made.
Behaviour change ladders fSection 5. 1-3)
The second measure of behaviour change was based on work by Marcus & Owen (1992) and Booth et al (1993) which used the model developed by Prochaska and DiClemente (1983) (see Figure 4) to investigate the process of exercise behaviour change. Subjects were shown three sets of statements, one for exercise, one for diet and one for smoking. These statements were in the form of a ladder with 10 rungs, each rung representing a stage between not acting with no intention to change and having acted. Five of the ladder ‘rungs’ were labelled with statements relating to the five stages of the transtheoretical model (i.e. / currently do not exercise and do not intend to exercise in the next 2 months; I currently do not exercise but I am thinking about starting to exercise in the next 2 months; I currently exercise a little but not regularly; I currently exercise regularly but have only begun to do so in the last 2 months; I currently exercise regularly and
have done so for longer than 2 months). However, the subjects were encouraged
to circle any of the numbers fi"om 1-10, and not to feel restricted to the five statements if they felt their response lay somewhere in-between.
In a similar manner to the development of the behaviour change scores described above, a ladder change score was created for each subject by subtracting their ladder position at time 1 from that at time 3. Again, the ladder change scores were recoded into dichotomous variables because the main point of interest was whether the subjects changed for the better or not, in order to be able to predict
________________________________________ Chapter 3________________________________
those who had moved up the ladder (towards maintenance of positive behaviour change) and 2) those who stayed where they were or moved down the ladder. The resulting dichotomous ladder change scores, one for each of exercise, dietary and smoking behaviour, were used in the analyses as measures of specific
behaviour change. However, it should be pointed out that these ladder scores do not provide a definitive measure of actual behaviour change. The ladders include both cognitions (rungs 1-4) and self-reports of behaviour (rungs 5-10). So the resulting group who are reported to have changed for the better on the ladder may not actually have changed their behaviour if they have moved for example from rung 2-4. However, this does provide a measure of whether someone has moved upwards in the process of reaching the maintenance of behaviour change.
3.4 Results
Computer package
Data were analysed using SPSS for Windows on the PC. Data screening
Prior to analysis, all the variables were examined to check for accuracy of data entry, a plausible range of the variable and missing values. This was carried out using aspects of the FREQUENCIES programme in SPSS for Windows. All the ranges were found to be plausible given the variables’ construction and most of the missing values were seen to be randomly distributed throughout the sample and the variables. There was however a problem with missing values in the variables of the action phase of the HAPA. There was a shortage of numbers who completed the ‘action phase of the HAPA’ section (Section 3: 20-28) due to the basis of the measures of the action phase being whether or not they had made a decision to make changes in their lifestyle (Section 3:18) and what they had decided to do (Section 3: 19). If either of these questions was answered in the negative, the subject could not fill in Questions 20 to 28 and indeed were
instructed not to do so. It was however important to ask these questions (18 and 19) in order to focus the subject on the specific change they wished to make as
how they planned to go about it. However, most of the action variables were calculated for those who had not filled in this section. Subjects in this group were given zeros for decision to change, wrote change, number o f means, number o f
implementation intentions, specific means and specific implementation
intentions.
Dichotomous variables were all checked for equal split between the two
categories. There were no variables found with extreme splits, i.e. 90-10% which would have led to truncation of correlation coefficients with these and other variables. So there was no need to delete any of these variables.
Continuous/ordinal variables, i.e. intention, outcome expectancies, threat, self^ efficacy, evaluation o f decision, number o f means, number o f implementation
intentions and behaviour change index were examined to see if they fiilfilled the
assumptions of normality required for the use of parametric statistical tests. SPSS FREQUENCIES provided skewness and kurtosis values and their respective standard error scores for each of these variables at time 1 (before screening) and time 2 (after screening). By calculating z scores by dividing the skewness and kurtosis values by their respective standard error scores, it was possible to see if the distributions differed significantly from the normal distribution. None were found to differ significantly from zero at the 1% significance level, except for
number o f means and number o f implementation intentions which were skewed -
the majority of subjects scored 0 (they had not devised any means or made implementation intentions). These variables were thus dichotomised for further analyses as it was decided that transformation would have been too difficult to interpret. The score ‘0’ was given for no means and for no implementation intentions; the score ‘1 ’ was given for one or more means or implementation intentions. All the other variables can be described as not deviating significantly from the normal distribution, so parametric tests were used for these variables. Choice of statistical tests
Parametric and non-parametric tests were used. Non-parametric tests were chosen for dichotomous variables {decision to change, wrote change, specific
________________________________________ Chapter 3________________________________
means, specific implementation intentions, number o f meansand number o f
implementation intentions). These non-parametric tests employed were: Chi-
square (x^) and McNemar’s test to examine differences in proportions and point- biserial correlation (used when one variable is dichotomous, the other continuous) to examine relationships between variables. Parametric tests were chosen for all the other variables since they fulfilled the requirements necessary for parametric tests as shown above. Parametric tests used were t-tests to look for differences between means and Pearson’s product-moment (r) correlations to test for associations between variables.
Multivariate analyses techniques were used to assess relationships between one continuous dependent variable and several independent variables. Multiple regression analysis was used to predict intention fi-om the combination of the other variables fi'om the HAPA motivation phase. It was also used to predict behaviour change in terms of the number of changes made (behaviour change index). Variables intended for use in these multiple regressions were examined for multivariate normality, linearity and multivariate outliers. This was done using SPSS REGRESSION which examines residuals. Scatterplots of predicted values of the dependent variables {intention and behaviour change index) against residuals showed that the assumptions of linearity were met in both cases. A normal probability plot of residuals showed that the multivariate distributions appeared normal - the points fell along an approximately straight diagonal line. With the use of a p<.OOI criterion for Mahalanobis distance, no outliers among the cases were identified. Correlation matrices between the variables to be entered into the equation were checked for multicollinearity and singularity. There were no highly correlated independent variables, so none of them were redundant in the analyses.
Discriminant Function Analysis was chosen to assess the predictive relationship between variables of the HAPA motivation phase and the specific behaviour change variables which were dichotomous. Previous examination of variables for parametric and multiple regression assumptions revealed no particular threat to this analysis.