presents descriptive statistics for the key study measures by the three subgroups that emerged from the question ‘How many times, on average, do you miss your oral contraceptive pill each month?’ There were no significant differences between the 3 groups in background and demographic factors apart from on the question asking whether participants had a fixed time for OCP use. Participants who didn’t have a fixed time for OCP use were much more likely to miss their OCP once or more (Odds ratio 3.62, 95% confidence interval: 1.34-9.79). The correlation between the continuous (MARS) and categorical measures (0, 1 or 2+ times per month) of Adherence to OCP was positive and large (r = 0.57, p<0.01) providing evidence of convergent validity for the two measures.
Study sample by ‘How many times, on average, do you miss your oral contraceptive pill each month?’
With regard to the psychological predictors there were significant differences between the three groups on all measures with the exception of anticipated regret and necessity beliefs. The significant differences observed were all in the predicted direction with those reporting never missing their oral contraception each month with higher intentions to use OCP, higher perceived behavioural control, lower concerns about OCP and higher levels of action and coping planning.
presents the correlation matrix together with the means and standard deviations and the Cronbach’s alphas for the main study measures. All of the psychological predictors with the exception of Necessity beliefs were significantly associated with adherence to OCP in the predicted direction. Effect sizes were mostly in the small to medium range (r >0.10 and r< 0.30), according to Cohen’s criteria
], with the exception of PBC which had a large association with OCP adherence i.e. r > 0.50. The Cronbach’s alpha for all scales was satisfactory (alpha > 0.70) with the exception of Necessity beliefs.
Descriptive statistics and correlation matrix for the main study measures
presents the hierarchical multiple regression of Adherence to OCP on the seven psychological predictors. Although action and coping planning were highly correlated multicollinearity was not a problem in this analysis according to the VIF values. The final model accounted for almost half of the variation in adherence to OCP. Although the final model in Step 3 did account for an additional 4% of the variability beyond that accounted for the initial model with Intention, PBC and regret, there was not a statistically significant change to the R2 in steps 2 (p=0.57) and step 3 (p=0.06). The strongest and only statistically significant predictors in the final model were perceived behavioural control and coping planning. Higher levels of perceived behavioural control and higher levels of coping planning were associated with better adherence to OCP. We re-ran the analysis with the problematic Necessity variable excluded and this did not significantly alter the overall pattern of findings. We also ran this model using unintentional (Item 1: I forget to take my oral contraceptive pill) and intentional MARS items (other 4 items) as dependent variables in a sensitivity analysis. The R2 for unintentional non-adherence was 0.46, p<0.001 while the R2 for intentional non-adherence was 0.35, p<0.001, with pbc and coping planning remaining as the strongest predictors in both models.
Hierarchical multiple regression analysis of Adherence to OCP on psychological predictors
A simple moderated regression analysis revealed an interaction between intentions and anticipated regret. The R2 increase due to the interaction was 0.09, p<0.001 and the beta for the interaction term was - 0.207, standard error = 0.056, p<0.001. The probed interaction is illustrated in Figure
. We conducted slopes analysis on the low and high anticipated regret lines to determine whether they differed from zero. This revealed that the line for low anticipated regret differed significantly from zero (Beta =−0.438, standard error= 0.081, p<0.001), but the line for high anticipated regret did not (Beta =−0.006, standard error =0.088, p =0.948).This shows that intention to use OCP and adherence to the OCP relationship was moderated by anticipated regret. In other words intentions to use the OCP were strongly predictive of non-adherence when anticipated regret was low but not when regret anticipated regret was high.
Interaction between anticipated regret (AR) and intention on OCP non-adherence (higher scores on the y axis relate to more non-adherence to OCP).