Individual and collaborative implementation intentions and the promotion of breast self-examination

Abstract Implementation intentions, plans relating to when and where behaviours will be performed, have been effective in increasing health behaviour. Two studies are reported that test the impact of this strategy in promoting breast self-examination (BSE), a behaviour shown to aid the early detection of tumours. In study 1, 457 participants were randomly allocated to either implementation intention or control conditions. Implementation intentions significantly increased the likelihood of BSE at one month. The effect of the intervention was marginally significant at six months. Study 2 (N = 101) tested the efficacy of a collaborative implementation intention intervention that required female participants to plan, with a partner, collaborative BSE performance. Results indicated that both implementation intentions and partner involvement were associated with BSE performance at one month, whilst the collaborative implementation intention intervention showed a 100% success rate. Collaborative implementation intentions may reduce forgetfulness. Interventions that encourage partner involvement in planning and enacting behaviour appear to enhance implementation intention utility. Keywords: Implementation intentionscollaborative implementation intentionsbreast self-examinationpartner support Notes 1 d = 2Φ/(√(1 − (Φ)2)). 2 Homogeneity of variance assumption was violated. Corrected values are reported. 3 d = M1 − M2/σpooled. 4 As different behavioural measures were taken at baseline compared to those taken at one-month and six-months a mixed (time × group) ANOVA was inappropriate. However, relevant ANCOVAs with past behaviour as a significant covariate one month: F(1, 149) = 48.32, p < 0.0005; six months: F(1, 116) = 63.74, p < 0.0005 were performed with a significant effect of group (implementation intention vs. control) at one-month F(1, 149) = 8.76, p < 0.005 but no effect at six months F(1, 116) = 1.23, p = 0.05. 5 Contingency coefficient value is reported as an index of effect size. 6 Contingency coefficient value is reported as an index of effect size. 7 d = 2Φ/(√(1 − (Φ)2)). 8 Fisher's exact test reported as 50% of cells had expected count less than 5. 9 When logistic regression is performed on data with an empty cell the results often become unstable. The inflated standard errors on the final step of regressions A and B reflect this and might have been responsible for the failure to detect a significant interaction between implementation intentions and partner involvement. However, running the equivalent linear regression failed to identify such an interaction, therefore, the lack of significant interaction might be more appropriately attributed to a ceiling effect arising from the large impact of partner involvement on BSE.