The theory of planned behaviour and breast self-examination: Assessing the impact of past behaviour, context stability and habit strength

Abstract

This study applied the theory of planned behaviour (TPB) to the prediction of breast self-examination (BSE) intentions and behaviour, and tested whether the frequency of past behaviour and context stability moderates intention–behaviour and habit–behaviour relations. Seventy-seven females completed measures of the TPB, frequency of past behaviour, context stability and habit strength (Self-Report Habit Index). BSE behaviour was assessed at 1-month follow-up (n = 66). The TPB explained 33% of the variance in BSE intentions and 11% of the variance in time 2 BSE. The frequency of past behaviour moderated the intention–behaviour relationship such that the intention was only positively related to time 2 BSE behaviour when the frequency of past behaviour was low. Context stability and the combination of the frequency of past behaviour × context stability moderated the habit–behaviour relationship such that habit strength was only positively related to time 2 BSE behaviour when context stability and the combination of frequency of past behaviour × context stability were high. The results are consistent with the proposal that behaviours that are performed frequently in stable contexts are predominantly under the control of habitual processes, whereas behaviours that are performed infrequently in unstable contexts are predominantly under the control of intentional processes.

Keywords:

• theory of planned behaviour
• past behaviour
• habit
• breast self-examination

Notes

Notes

1. In order to assess the possible impact of multicollinearity among the independent variables, collinearity statistics (i.e. tolerance, variance inflation factors) and collinearity diagnostics (i.e. condition index, variance proportions) were computed for the regression analyses. None of the tests indicated that multicollinearity was a cause for concern.

2. A power analysis, with power set at 0.80 and alpha set at 0.05, was conducted to estimate the sample sizes that would be required to detect typical effect sizes reported in the literature. Armitage and Conner (2001) in their meta-analysis of TPB studies reported average correlations with intention of 0.49, 0.34 and 0.43, for attitude, subjective norm, and perceived behavioural control, and a multiple R of 0.63. At least 22, 49, 29 and 22 participants are required to detect these effect sizes. For behaviour, the average correlations were reported as 0.47 and 0.37 for intention and perceived behavioural control, and the multiple R was 0.52. At least 24, 41 and 30 participants are required to detect these effect sizes. Ouellette and Wood (1998) reported an average correlation of 0.39 between measures of past behaviour and future behaviour in their meta-analysis, and Conner and Armitage (1998) reported that, on average, past behaviour explained an additional 13% of the variance in behaviour over and above the TPB. At least 36 and 55 participants are required to detect these effect sizes. In the current study, four measures of past behaviour were assessed. Assuming a similar increment in R ², at least 85 participants would be required to detect such an effect.

3. The regression analysis was repeated by only entering the frequency of past behaviour at step 2, to allow comparisons with other TPB studies. The addition of the frequency of past behaviour led to a non-significant increment in the amount of variance explained in time 2 BSE behaviour, ΔR ²= 0.04, ΔF(1,62) = 1.60, p = 0.21.

4. An additional moderated regression analysis was performed to examine the interaction between intention and habit strength when predicting time 2 BSE. The main effects for intention, β = 0.24, p = 0.10, and habit strength, β = 0.16, p = 0.26, were non-significant at step 1, although together they explained a significant proportion of the variance, R ²= 0.12, F(2,63) = 4.32, p = 0.02. The addition of the interaction term failed to increase the amount of variance explained, ΔR ²= 0.001, F(1,62) = 0.04, p = 0.84; the intention × habit strength interaction term was non-significant, β = −0.03, p = 0.84.

5. It could be argued that the significant interactions involving habit strength might have occurred because the SRHI includes some items that refer to the frequency or stability of past behaviour. The moderated regression analyses were therefore re-run after removing such items (i.e. ‘…I do frequently’, ‘…that belongs to my (daily, weekly, monthly) routine’, ‘…I have been doing for a long time’). The pattern and significance of the results remained unchanged. As in the initial analyses, context stability moderated the habit–behaviour relationship, β = 0.30, p = 0.020, as did the combination of frequency of past behaviour frequency × context stability, β = 0.82, p = 0.008.

6. The likely impact of attrition on the study findings was also assessed by re-running the time 1 analyses for the reduced sample (n = 66) who completed questionnaires at both time points. The pattern and significance of the correlations between the variables remained unchanged. Only one difference was found in the regression analysis predicting intention to perform BSE; namely, habit strength became non-significant. This is consistent with the idea that the restricted range of the past behaviour/habit measures may have reduced their predictive power.