Mediators of gender effects on depression among cardiovascular disease patients in Palestine

This cross-sectional hospital-based study was conducted on consecutive patients aged 30–80 years admitted to the cardiology and cardiac surgery departments at one of the four main hospitals in Nablus, Palestine, during their stay in the hospital (1–7 days) between March 2017 and November 2017. They were included in the study if they were diagnosed with CHD, ST elevation or non-ST elevation MI, angina, HF, cardiac arrhythmia, valve disease or any other cardiac disease. Patients with a normal cardiac catheterization (CATH), an acute or past stroke, end-stage kidney disease (including dialysis patients), peripheral vascular disease, psychological disorders, cognitive deficit, neurological disorders (dementia, Alzheimer’s disease, epilepsy, Parkinson’s disease), use of antidepressant or any other condition affecting the quality of their responses were excluded from the study. A total of 1092 patients were approached and met initial eligibility requirements, and 1053 agreed to participate in the study. The analytic sample included 1022 subjects. The study was approved by the Ethics Committee of Nordwest-und Zentral Schweiz (EKNZ) in Basel, Switzerland, and by the Institutional Review Board (IRB) committee at An-Najah National University in Nablus, Palestine.

Eligible patients were approached for private in-person interviews using hospital registries and written informed consent was obtained from those who agreed to participate. Prior to completion of the interview, socio-demographic and clinical characteristics including: age, gender, marital status, residence, education level, occupation, income; current diagnosis, previous cardiac diagnoses, years with cardiac disease, cardiac treatment (at admission), co-morbidities, medications and family history of CVD were retrieved from patients’ medical records and were ascertained via patient interview. In-person interviews consisted of a sequence of validated instruments assessing physical and psychosocial factors and questions assessing lifestyle factors.

Depressive symptoms were assessed by the Cardiac Depression Scale (CDS), a disease-specific, 26-item questionnaire used to accurately measure depression in patients with CVD. The CDS is scored based on a Likert scale ranging from strongly disagree (1) to strongly agree (7). Total scores are calculated as the sum of the scores of each item and range from 26 to 182 [38]. The CDS can be used as a continuous measure, where higher scores indicate more severe depressive symptoms, or as an ordinal indicator of possible depression using previously established cut-off points which include: < 90 for no depression; 90–100 for mild to moderate depression; and >100 for severe depression. These cut offs were used in the current study to categorize depressive symptoms. The CDS has a sensitivity of 88% and a specificity of 84% for severe depression when a cutoff of > 100 is used, and 84% sensitivity and 78% specificity for cutoff scores of 90–100 [39]. Further instruments used include the Post-Traumatic Stress Disorder Checklist (PTSD-PCL-S) [40]; Enhancing Recovery in Coronary Heart Social Support Instrument (ESSI) [41, 42]; Resilience Scale-14 (RS-14) [43]; Single-Item Self-Esteem Scale (SISE) [44]; Short-form 12 Health Survey (SF-12-PCS; SF-12-MCS) [45] and the Patient Health Questionnaire-15 (PHQ-15) [46]. The instruments have been described in detail elsewhere [47].

A structural equation model (SEM) was built to assess direct and indirect (i.e. mediated) effects of gender on depression. Mediators are variables which transfer part of the effect of an independent variable (IV) on a dependent variable (DV). In our analysis, the IV was gender (used as a binary variable, with “being female” coded as 1), the DV was depression (CDS score, used as a continuous variable), and the potential mediator variables were socio-demographic, clinical, psychosocial and lifestyle factors [48]. First, mediator groups (i.e. socio-demographic, clinical, psychosocial and lifestyle factors) were assessed in separate models. In a second step, the final model was built (reference model, N = 1022), including education, income, comorbidities, somatic symptoms, quality of life, PTSD, resilience, self-esteem, smoking, physical activity. These potential mediators were chosen based on 1) their statistical significance in the preceding analyses, using a cut-off level of p < 0.10 2) the magnitude of their association 3) whether the associations were significant between gender and the potential mediator, as well as between the potential mediator and depression (CDS score). The SEM was estimated using the maximum likelihood method and included continuous, binary and ordinal variables which were all treated as continuous. Inclusion of a larger number of potential mediators reduces bias in estimating the true direct effect of gender on depression (CDS score). Potential confounding factors assessed were age, marital status, occupation, social support, and lifestyle factors (i.e. BMI, alcohol consumption). A further SEM-analysis was then conducted stratifying by age, running separate models for two age groups (those aged below and above the median age of 59 years; see Additional file 1: Figure S1a and b). Moderation by age was assessed by testing the differences between corresponding age-specific effects using Chi-squared tests. In further analyses, two additional SEM’s were computed to assess whether resilience and smoking could be potential mediators and/or moderators of the effect of PTSD on depression, adding the respective factor as mediator between PTSD and depression and testing its interaction with PTSD to assess potential moderation. This additional analysis was based on general conceptualization and literature on psychological comorbidity of PTSD and depression [36, 49]. Analysis was conducted using the STATA Statistical Software Release 15 (StataCorp., College Station, U.S.A.).

The pooled SEM of the direct effect of gender on depression (CDS score) and the indirect effects when mediators were added was conducted for the overall sample. Figure 2 illustrates the pathways between gender and depression with the tested mediators. Parameters reported in the figure include coefficients and confidence intervals (CI). They demonstrate the associations of gender with each potential mediating factor and between each potential mediating factor and outcome variable, after adjusting for the effect of gender. Results from the SEM showed both positive and negative mediators, where positive mediators mediating the gender difference on depression are associated both with being female and higher levels of depression. In the pooled SEM of gender and depression, there was no direct effect of gender on CDS score (coefficient: 0.0, 95% CI: − 2.2-2.2, p = 1.0). There were significant direct effects of gender on the mediators as shown in Fig. 2. On the one hand, female gender was positively associated with higher PTSD symptoms, comorbidities, somatic symptoms and income. On the other hand, female gender was associated with lower resilience, self-esteem, quality of life, education, less smoking and lower physical activity. Negative indirect effects of gender on depression were observed through resilience, self-esteem, quality of life, (SF-12-PCS), education, income and physical activity, whereas positive indirect effects of gender on depression were observed through PTSD, comorbidities, somatic symptoms, and smoking. Quality of life, education, and income were not significant mediators for the indirect effect of gender on depression. No evidence of moderation was found for the direct effect of gender on CDS score, the effects of gender on mediators and the effects of mediators on CDS score.

There was no indication of confounding or moderation by age. In the pooled model, age was not significant when a direct effect of age on CDS score was added to the model. There were also no significant differences between corresponding effect estimates after SEM’s were conducted separately for the two age groups < 59 years and ≥ 59 years. These results are presented in Additional file 1: Figure S1a and b.

The SEM conducted to assess whether resilience could also be a mediator and/or moderator between PTSD and depression (CDS score) revealed a significant direct association between PTSD and depression and a small positive moderating effect of resilience on this association was observed. There were negative yet significant associations between PTSD and resilience and between resilience and depression (see Additional file 2: Figure S2). Thus, in this SEM, resilience appeared both as a mediator of the total effect of PTSD on depression and a moderator of the respective direct effect.

Figure 3 illustrates the interaction between PTSD and resilience, in which the difference in predicted depression (CDS score) between those with low and high resilience decreases slightly with increasing PTSD symptoms. This suggests that resilience shows a clearer protective role at lower levels of PTSD, becoming less protective at higher levels of PTSD.

No mediation by smoking of the association between PTSD and depression (CDS score) was found (data not shown).

This study had several strengths. First, the high participation rate decreased the likelihood of selection and participation bias. Second, our sample of women was relatively large, compared to other studies, which provided sufficient statistical power for testing gender differences. Third, the CDS used in the study is the only psychometric scale suitable for the comparative depression assessment in heart disease patients, subjected to different interventions and has excellent properties for the diagnosis of severe depression, a score > 100 having a sensitivity of 88% and a specificity of 84%. Finally, the findings of our study are likely generalizable as our sample was recruited from four main hospitals of Nablus, Palestine, which provide cardiac care to a large percentage of individuals in the area.

The limitations of our study are due to several sources. First, depression was not formally diagnosed by a psychological interview and was rather assessed using the CDS, thus we cannot conclude that our sample had clinical depression. However, the identification of certain mediating factors of gender on depression among cardiac patients can serve as a basis for developing the best intervention targeting depression among women and men with cardiac disease. Second, some patients were recruited at the time of admission and for the most part before receiving their intervention or diagnosis, and thus were under stress and anxiety. As some of the anxiety observed may have been transient, the level of depression and anxiety could have been overestimated. Finally, due to the cross-sectional design of the study the direction of associations cannot be derived with certainty and causality cannot be inferred.