This study received funding in January 2015 (trial registration NCT02778074) . Participant recruitment took place between January 2017 and February 2018, and the main findings were published in 2019 [15]. Thus, this study is a secondary analysis of the data that were collected in an RCT evaluating the effect of a 9-week iCBT program on depressive symptoms in patients with CVD. A total of 144 CVD patients with at least mild depressive symptoms (ie, Patient Health Questionnaire-9 score>5) were included and randomized to 9 weeks of the iCBT (n=72) or the ODF (n=72) [15]. The iCBT program consisted of 7 modules, namely goal setting, psychoeducation, problem-solving, behavioral activation, part 1 (2 weeks) and part 2 (2 weeks), and a summary module. The program also included weekly homework assignments with feedback provided by nurses. The ODF consisted of 9 discussion topics moderated by a nurse. The discussion took place in writing.

The regional ethical review board of Linköping in Sweden approved the study (reference number 2016/72-31).

Data for this analysis were collected at baseline and at follow-up at the end of the 9-week intervention period. All data were collected through questionnaires on the study website [15].

Descriptive statistics was used to describe the study population. Continuous variables were described as means and SDs, and categorical variables were described as numbers and percentages. All analyses were performed on original data, thus including participants with complete data at the 9-week follow-up (n=127). Analysis of covariance was used to evaluate if there was a significant difference in self-efficacy scores at the 9-week follow-up between the iCBT and ODF groups after adjustment for self-efficacy scores at baseline. The Cohen d was calculated for evaluating the effect of iCBT on self-efficacy. The associations among the changes in self-efficacy, depressive symptoms, and physical activity were explored by structural equation modeling (SEM). For these analyses, we calculated scores representing changes in self-efficacy, depressive symptoms, and physical activity between the baseline and 9-week follow-up measurements. To explore the influence of iCBT on these associations, we also added iCBT as an independent variable in the final SEM model. The associations/relationships obtained by SEM were described using their standardized coefficients (β). The chi-square value, the root mean square error of approximation (RMSEA), and comparative fit index (CFI) were used as goodness of fit indices of the SEM models. An insignificant chi-square would indicate a good model fit. An overall RMSEA below 0.06 indicates a good fit whereas a CFI ≥0.95 is considered a good fit, indicating that at least 95% of the covariation in the data is reproduced by the model. P<.05 indicates a significant value. Statistical analyses were performed with SPSS version 25.0 (IBM Corp) and the LISREL8 software (Scientific Software International) [23].

The mean age of the study population was 63 years. Among the 144 participants, 90 (62%) were males; almost 29 (20%) lived alone, and another 29 participants (20)% described their financial situation as problematic. Few participants were current smokers (ie, 4/144, 3%) or drank more than 10 units of alcohol per week (ie, 7/144, 5%). The median number of medications was 5 and approximately 48 participants had more than 1 comorbidity. Table 1 presents the characteristic of the study participants. The mean and SD values at baseline for self-efficacy, depressive symptoms, and physical activity as measured by the GSES, MADRS-S, and physical activity factor were 27.2 (6.3), 17.8 (6.7), and 4.8 (5.3) respectively. None of the variables presented in Table 1 differed significantly between those randomized to iCBT or the ODF.

At the 9-week follow-up, the mean self-efficacy scores and SD values for the iCBT group and the ODF group were 29.9 (5.8) and 28.2 (6.4), respectively. After adjustment for baseline scores, analysis of covariance showed a significant difference in the increase of self-efficacy in favor of iCBT (1.67, P=.04). The Cohen d was 0.27, indicating a small significant effect of iCBT on self-efficacy.

For analyzing the associations among the changes in self-efficacy, depressive symptoms, and physical activity, we first explored an SEM model based on the mediation model reported by Maeda et al [13]. They initially analyzed the association between depression and medical adherence and then that between depression and self-efficacy, which were both statistically significant. After adding the association between self-efficacy and medical adherence to the model, the significant association between depressive symptoms and medical adherence disappeared. Thus, self-efficacy served as a full mediator between depression and medical adherence. Our model showing the relationships among depression, physical activity, and self-efficacy (Figure 1) had a perfect fit (χ²₀=0, P>.99, CFI=1, and RMSEA=0), which indicates that the model completely explained the correlations. We found a significant direct effect (β=–.24, P<.01) among changes in depression, physical activity, and self-efficacy (β=–.64, P<.001). However, the association between the change in self-efficacy and change in physical activity was weak and not statistically significant (β=.11, P=.5). Consequently, in our model, self-efficacy was not a mediator between changes in depressive symptoms and physical activity.

However, the findings reported by Maeda et al [13] may also be interpreted as a simplex model, assuming that a change in depressive symptoms leads to a change in self-efficacy, which in turn leads to a change in physical activity. Given this assumption, self-efficacy could serve as a mediator between depressive symptoms and physical activity (Figure 2). This model had a perfect fit (χ²₀=0, P>.99, CFI=1, and RMSEA=0). A summary of the associations is shown in Table 2. The model shows that a change in depressive symptoms has a significant relationship with the change in self-efficacy (β=–.64, P<.001), and a change in self-efficacy is significantly related to a change in physical activity (β=.49, P<.02). There was also an indirect association between the change in depressive symptoms and change in physical activity in which the change in self-efficacy served as a mediator (β=–.31, P<.01).

Furthermore, adding iCBT to our simplex model (Figure 3) showed that as expected, the iCBT had a direct and significant association with the changes in depressive symptoms (β=.37, P<.001) (χ²₂=1.49, P=.47, CFI=1, and RMSEA=0 ). Table 3 summarizes these associations. The iCBT was also significantly and indirectly associated with the changes in self-efficacy (β=.23, P<.001) and physical activity (β=.12, P=.03). We also explored a reversed model in which the iCBT had a direct effect on the change in self-efficacy, which in turn influenced the changes in depressive symptoms and physical activity. The fit of the model was poor (χ²₂=10.35, P=.005, CFI=0.82, and RMSEA=0.196), and the modification indices of the LISREL analysis strongly supported the model presented in Figure 3.

Performance accomplishment has been discussed as an important precursor of self-efficacy; if a person is successful at tasks, the feeling of efficacy will increase [6,7], which may indicate that an increase in self-efficacy leads to an increase in physical activity, which in turn increases self-efficacy. We added such a recursive association to the model presented in Figure 3. The fit of the model was good (χ²₂=1.49, P=.47, CFI=1, and RMSEA=0) and showed that the relationship between self-efficacy and physical activity was significant (β=.5, P<.001) whereas the relationship between physical activity and self-efficacy was not (β=.27, P=.11). Accordingly, a significant recursive association was not found.

The main findings of this study were that iCBT improved self-efficacy. However, the influence of iCBT on the improvement in self-efficacy and physical activity was mediated by improvements in depressive symptoms. Thus, self-efficacy was a mediator between improvements in depressive symptoms and physical activity.

In this study, we investigated the influence of self-efficacy on the changes in depressive symptoms and self-care through physical activity (ie, an aspect of autonomous self-care) [24]. To our knowledge, there is a lack of such studies involving CVD patients. In one of the few studies, Shen et al [14] showed that changes over 6 months in depressive symptoms (β=–.15, P=.05) and self-efficacy (β=.34, P<.001) were associated with medical adherence in patients with heart failure. However, in that study, only approximately 44% of the participants showed increased levels of depressive symptoms, and this may underestimate the results reported. Another study involving patients with heart failure reported that depression was negatively indirectly associated with poorer self-care through poorer self-efficacy [25], but this was a cross-sectionally designed study with no possibility to detect changes. Our study was performed on CVD patients with at least mild depression, and we found that a change from the baseline depressive symptoms was directly associated with the change in self-efficacy; this was indirectly associated with changes in physical activity, and thus self-efficacy had a mediating role in the relationship between changes in depressive symptoms and changes in physical activity. This highlights the importance of improvements in depressive symptoms in CVD patients. Such improvements are likely to be followed by improvements in self-efficacy and self-care, such as physical activity.

However, for such improvements to take place, an intervention is most likely needed. There have been requests for behavioral interventions that promote improvement of depressive symptoms, self-efficacy, and self-care in CVD patients [14]. In previous studies, we have found that an iCBT program for depression in CVD patients can improve depressive symptoms and increase physical activity [15,16]. This study now adds that iCBT can improve self-efficacy in depressed CVD patients as well. However, as observed in Figure 3, the impact of iCBT on self-efficacy is indirectly mediated by improvements in depressive symptoms. In Figure 3, the total effect of self-efficacy on physical activity is β=.5, but most of this is explained by the change in depressive symptoms (β=–.32); thus, the unique effect of self-efficacy on physical activity is β=.18 This suggests that only a minor part of the mediating role of self-efficacy between depressive symptoms and physical activity is related to self-efficacy itself. The model in which we analyzed if iCBT first led to changes in self-efficacy and then to changes in depressive symptoms was not valid, suggesting that improvements in depressive symptoms precede the increase in self-efficacy and physical activity. Moreover, we also explored if a change in self-efficacy leads to a change in physical activity, which in turn leads to changes in self-efficacy (ie, performance accomplishment). However, changes in physical activity had no significant effect on changes in self-efficacy (β=.27, P=.11).

Most digital interventional studies in cardiac rehabilitation focus on physical activity and counseling, with less focus on the core components such as psychological management [26]. However, our study suggests that in CVD patients with depressive symptoms, a digital intervention designed to improve self-efficacy and self-care such as physical activity should primarily focus on psychological management of depressive symptoms, for example, using iCBT. This does not contradict that the fact that such an intervention also includes physical activity counseling in the management of CVD and depression, and this can motivate the patients to perform physical activity when being involved in the iCBT program.

A limitation of this study is that this was a secondary analysis of data, and it did not primarily intend to investigate the effect of iCBT on self-efficacy or to explore the associations among changes in depressive symptoms, self-efficacy, and physical activity. Furthermore, data regarding physical activity were collected using only self-reports and not through objective measurements. This may have underestimated the level of physical activity measured [27] and may explain the nonsignificant association between self-efficacy and physical activity in Figure 1. However, in CVD patients with depression, there is a knowledge gap in the relationships among the changes in depressive symptoms, self-efficacy, and self-care such as physical activity and the influence of iCBT on these relationships. Therefore, we believe that despite the limitations, the results of this study provide interesting and useful information.

The iCBT program was more effective than the ODF in increasing self-efficacy in CVD patients. An increase in self-efficacy was the mediator between improvements in depressive symptoms and physical activity. Improvements in depressive symptoms mediated most of the influence of iCBT on the improvements in self-efficacy and physical activity. The findings are preliminary and replication in larger samples is needed.