Cardiovascular diseases (CVD) are still the leading cause of death in many western countries even though major advances in treatment have contributed to higher rates of surviving cardiac events [1-3]. It is also evident that not all segments of society benefit from the major advances and disparities in CVD rates persist between the majority and ethnic minorities [4,5].

In order to prevent recurrent cardiovascular events, patients are advised to gradually resume their regular life routines while also changing lifestyle habits that are major risk factors for CVD [6,7]. Smoking (tobacco addiction) is the most significant of the modifiable cardiovascular risk factors; thus smoking cessation is considered a high priority in the management of patients with cardiovascular disease [8]. However, most patients find it difficult to adhere to smoking cessation [9]. Again, Non-adherence to recommended cardiac rehabilitative and preventive actions is particularly higher in patients from minority groups [10].

In Israel, ethnic disparities are mainly assessed as those between the Jewish majority and the Arab minority citizens. The Arabs compose 20% of the population, of which 86% are Muslims. Their socioeconomic status (education, income and employment) is consistently lower than that of the Jewish population, and the disparities in health status are significant, with lower life expectancy and higher mortality rates in the Arab population [11]. Similar differences are found in cardiac morbidity and mortality rates [12-14]. Smoking rates in the general population are also higher among the Arabs: the rates are 28% and 45% for Jewish and Arab men, respectively, (but 16% and 4% for Jewish and Arab women, respectively) [15]. The higher rates of smoking among Arab men are sustained among cardiac patients, even when smoking is conceptualized as a proximate health risk, i.e., risk of recurrent cardiac events: smoking cessation was almost twice as low for Arab than for Jewish cardiac patients [16]. Additionally, others showed that the rate of Arab patients’ participation in cardiac rehabilitation programs is significantly lower than that of Jewish patients [17], minimizing their chances of being exposed to smoking cessation interventions. Therefore, it is a priority to recognize and understand the factors that feed into the ethnic gaps in smoking behavior, in order to address the needs of these populations and to implement culturally sensitive health-promoting interventions [5]. Yet, theory-based assessment of possible factors that may account for the disparities in smoking behavior among cardiac patients, beyond socio-demographic determinants per-se, is extremely limited.

Efforts to explain health behavior change, including smoking cessation, may benefit from the well-established theoretical models of health behavior. These frameworks attest to the fact that besides sociodemographic characteristics and knowledge, the way people perceive the situation and particularly their ability to control it is an important determinant of health promoting behavior [18-20]. Indeed, perceived behavioral control is a central component of a handful of theories which predict overall behaviors, and health behaviors in specific, such as: the Social Cognitive Model [21], Theory of Planned Behavior [22], and Health Belief Model [23].

Perceived behavioral control (PBC) is defined as the extent to which a person feels able to enact behaviors [24]. PBC is shaped by life experiences as well as by one’s cultural background [22]. PBC has two aspects: the first aspect consists of how confident a person feels about being able to perform or not perform the behavior, whereas the second aspect taps beliefs about the power of both situational and internal or external factors to inhibit or facilitate the performing of the behavior [25]. In the current study, in our attempt to pinpoint putative psychological variables which may explain the differences in smoking cessation between Arab and Jewish cardiac patients in Israel, the first aspect was conceptualized by Bandura's concept of self-efficacy [26], and the second aspect, external perceived behavioral was conceptualized by the concept of fatalism.

Self-efficacy

According to Bandura's social learning theory [26,27] self-efficacy (SE) stands for one's confidence in his/her ability to perform different life tasks. With regard to disease prevention and health promotion, sense of efficacy was found to be a crucial regulator of one's motivations and behaviors [28]. SE was found to be a strong common determinant of behavioral change, especially in smoking behaviors [29,30], although the mechanisms linking SE to success in adherence to smoking cessation have not been well documented [9]. Yet, it is proposed that low SE undermines smokers’ ability to initiate or maintain efforts to cope with high-risk situations that may lead to smoking (e.g., negative affect states, exposure to smoking cues) or to reduce smokers’ likelihood of recovering from lapses once they do occur [31]. A qualitative review of 21 studies examining predictors of long-term [six month] cessation outcomes found that self-efficacy also predicted relapse among smokers [32].

The studies which explored self-efficacy among ethnic minorities validated its predictive value with regard to various health behaviors. Low self-efficacy was an important determinant of low rates of mammography screening among South Asian women living in Canada [33] as well as of smoking cessation among Afro-American women of low socio-economic status [34]. SE was the strongest predictor of nonsmoking among both Caucasians and African Americans [24]. Data on self-efficacy as a determinant of health behavior among Arabs and Jews in Israel is scarce, suffers from inconsistency in its measurements, and is contradictory. For example, in one study no association was found between self-efficacy and mammography screening among Arab women [35], whereas in another study, selfefficacy predicted physical activity among both Arab and Jewish college students in Israel [36].

Fatalism

Fatalism is defined as a person's faith or position regarding his destiny [37]. Fatalism refers to a general certainty that destiny cannot be altered and that one has very little control over his/her life [38].

A fatalistic person believes that each incident of his/her life is determined by God or fate or some other metaphysical entity; therefore any struggle against this power is futile [37]. This perception may lead to a tendency toward learned helplessness and prevent patients from acting in favor of their health [38]. Such fatalistic views of cancer were among the main barriers to mammography screening among Hispanic minority women [39]. Fatalism is an integral part of the Arab culture [40,41]. For most Arabs, fatalism consists of the belief that God possesses an ultimate power over one's life. If something goes wrong, people can release themselves from blame or do nothing to improve the situation by assigning the cause to God's will. Too much self-assurance about controlling events can even be considered a sign of arrogance to the point of blasphemy [42]. Indeed, Arab women in Israel who believed that God or fate was the cause of cancer refrained from having a mammogram [43].

In light of the overall goal of the current study to explore putative psychological variables of the differences in smoking cessation, the aim of the study was to assess the contribution of internal (self-efficacy) and external (fatalism) perceived behavioral control to smoking behavior among Arab and Jewish cardiac patients in Israel.

• Our hypothesis was as follows: Arab patients will be less likely than Jewish patients to cease smoking.

• Higher levels of SE and lower levels of fatalism will be detected among the Jewish patients compared with Arab patients.

• Higher scores of SE will be associated with higher rates of smoking cessation.

• Higher scores of fatalism will be associated with lower rates of smoking cessation.

• SE and fatalism will explain the differences: they will mediate the association between ethnicity and smoking behavior.

Participants and procedure

The study population was defined as all patients hospitalized in the Carmel Hospital Cardiology Department between January 1, 2006 and April 30, 2008 for first or recurrent cardiac event. Eligible patients were Arab and Jewish Israelis, aged 40-69 years, who had Been diagnosed with a major cardiac event and/or undergone a relevant medical procedure i.e., Myocardial infarction (MI); Percutaneous Trans luminal coronary angioplasty (PTCA); coronary artery bypass graft (CABG), and who suffered from no other severe chronic illness. The current study was a part of a larger study examining couples' relationships, therefore all eligible patients were married. Data were collected in 2007-2008 and a total of 630 patients were found eligible for inclusion in the study. The study was approved by the Carmel Hospital Review Board.

Names of all eligible patients were transferred to Arab or Jewish interviewers who had been specifically trained for the study. The interviewers first contacted the patients by telephone and after receiving an oral consent, a face-to-face interview at the patient's home was scheduled. After signing an informed consent form, the patients were interviewed in Arabic or Hebrew. Of the eligible 630 patients, 463 were approached (every second name and then repeating the list) until reaching the required quota of 200 patients (100 Arabs and 100 Jews). Of the 463, 163 (97(60%) Jews; 66 (40%) Arabs) were not located (due to incorrect telephone numbers, no answer at the correct telephone numbers, etc.), and 100 (71 Jews and 29 Arabs) refused to take part in the study, reaching a 66% response rate.

Measures

All questionnaires were prepared in Hebrew and in Arabic by the research team which consisted of both Arab and Jewish researchers, and assessed for clarity. Minor revisions were conducted following a pre-test of 20 patients.

Smoking cessation: Participants described their smoking behavior prior to the index cardiac event and their current smoking: smoked before and continue to smoke; smoked and ceased; never smoked; ceased before the index event. In order to tap smoking cessation, two Categories of smoking behavior were defined: (0) smoked and continues to smoke, and [1] smoked and ceased smoking after the index event. Naturally, only patients who had smoked prior to their index event were included in the analysis.

Self-efficacy (Internal PBC): Self-efficacy was measured by Bandura's self-efficacy scale. It was specifically adapted to meet the particular challenges facing cardiac patients. The scale taps 10 items such as: 'I believe in my ability to function at work despite my cardiac illness'; 'I believe in my ability to do whatever the rehabilitation program demands.' Patients were asked to rate each item on a 7-point Likert scale (1=not at all, 7=very much). The final score, a mean of all the items, represents the overall feeling of self-efficacy, in which higher scores represented greater self-efficacy. Cronbach's alpha in the current study was 0.93.

Fatalism (External PBC): This questionnaire was specifically designed for the current study. It is based on writings of both Arab and Jewish scholars in the field of Judaism and Islamic thought [37,44-47] and on consultations we had conducted with scholars in the field of Judaism and Islamic thought. The scale consists of seven items addressing beliefs in an outer force which governs one's life. Participants were asked to rate each item on a 5-point Likert scale (1=completely disagree, 5=completely agree). Higher scores represented higher levels of fatalistic beliefs. After applying principal component exploratory factor analysis with Oblimin rotation, two factors emerged, explaining 62% of the total variance. However, three items had very high loadings on both factors and an additional three items had low loadings on both factors; therefore, they were omitted. The final instrument therefore consisted of four items integrated into one factor which now explained 62.5% of the total variance. The items were: 'God controls one's life;' 'no matter what you do for your own health, eventually it is God's verdict;' 'praying to God may be helpful in times of illness;' and 'whatever happens to a person is in God's hands.' Cronbach's alpha was 0.86.

Health related variables: Three measures were used to assess patients' health status: a) Illness severity was measured by the patient’s self-report of any additional Major Acute Coronary Event (MACE) since the time of their index event. The responses were divided into high severity (1- recurrent MI/ recurrent CABG/ recurrent PTCA) versus low severity (0- no additional events, or elective diagnostic angioplasty only). b) Self-rated general health was assessed by a single question, using a 5-point scale (1=very bad, 5=very good). c) Length of time since the first cardiac event was assessed by a single self-report question (in years).

Socio-demographic background: This questionnaire included items on population group (1=Arab, 0=Jewish); age (in years); education level, measured on a 5-point scale (1=less than eight years of formal schooling, 5=university degree); Perceived economic status, selfassessed economic state, measured on a 5-point scale (1=very bad, 5=very good); and level of religiosity measured on a 5-point scale (1=not practicing any religious commandments, 5=practicing religious commandments to a great extent).

Statistical analysis

Multiple imputation analysis was applied in order to deal with the issue of missing data. According to Enders [48], the multiple imputation technique uses a regression- based procedure to generate multiple copies of the data set, each of which contains different estimates of the missing values. In the current analysis, we applied the SPSS 0.20 MI procedure, and 10 copies of the data set were generated. Briefly, for the imputation of each scale's missing items we used the other scales relevant for the analysis as temporary auxiliary variables. Therefore, an iterative imputation process repeatedly filled in the item scores from one subset while using the scale scores from the remaining subsets as auxiliary variables.

After completing the imputation process for each item subset, we computed a new set of composite scores from the filled-in item responses.

After creating the complete data sets, we ran the statistical analyses on each filled-in data set and subsequently used [49] formulas to combine the parameter estimates and standard errors into a single set of results. Since interactive effects were hypothesized, we followed Ender's advice to preserve interaction effects in the imputation model. Because the variables of perceived health, perceived socio-economic status and education level consisted of five levels, and their distribution was very close to normal distribution, these variables were treated as continuous independent variables throughout all analyses.

To examine the contribution of ethnicity, socio-demographics, health status and psychological variables to the variance in smoking behaviors, we initially applied Spearman correlations analyses, due to the dependent variable being dichotomous. Then, the data were analyzed using a four-step hierarchical logistic regression. In this regression, the predicted variable was smoking behavior (ceased/ continues). The first step of the regression included the patient population group (Arab/Jew). Step 2 included the relevant demographic variables (only those which were found to significantly differentiate between Arabs and Jews or to be associated with smoking behavior). Step 3 included self-efficacy and fatalism. The two-way interactions of fatalism, self-efficacy, and each of the demographic variables with population group were entered in Step 4, which consisted of the product of the centered scores of these measures (Population group x Fatalism; Population group x self-efficacy; Population group x age; etc).

Sample characteristics

Of the 200 participants, 78 Arabs (78% of the Arab sample) and 62 Jews (62% of the Jewish sample) were smoking before the index event, and therefore comprise the sample for our analyses. Half (50%, n=39) of the Arab patients continued to smoke after the event, compared with less than a fifth (19.4%, n=12) of the Jewish patients, χ²=14.01, df=1; p <0.000.

Table 1 presents the means and standard deviations of the sample background and study variables, for each of the population groups. It can be seen that the Arab patients were significantly younger, less educated, more religious, and perceived themselves to be of a lower socio-economic status than the Jewish patients. Significant differences were also found with regard to the two psychological variables. Arab patients were found to hold higher levels of fatalism and lower levels of self-efficacy than their Jewish counterparts. In addition (data not shown), though no differences were found between the population groups regarding perceived health status, the objective measure of illness severity shows that the majority of Arab patients (79.5%, n=62) were classified as having a severe illness compared with only half (54.8%, n=34) of the Jewish patients, χ² (1)=9.74, p<0.01.

Table 1: Frequencies, means and standard deviations of the study variables by group [N=140].

Associations with smoking cessation

Table 2 presents the Spearman correlations among the study variables. As can be seen, in addition to the lower education and SES level of the Arab patients compared to the Jewish patients, they also reported higher levels of religiosity, severity of illness and, as hypothesized, lower levels of SE and higher fatalism. Smoking cessation was significantly.

Table 2: Pearson’s correlations of the study’s variables [N=140].

Lower among the Arabs and was correlated with level of education, level of religiosity and fatalism but not with self-efficacy. The less educated, more religious and more fatalistic one was, the more one continued to smoke.

The binary logistic regression model predicting the probability of continuation versus cessation of smoking is presented in Table 3. In the current analysis, only the variables found to significantly differentiate between Arabs and Jews or to be associated with smoking behavior were entered into the regression model in addition to population group: age, perceived socioeconomic status, level of education, religiosity level, illness severity, self-efficacy and fatalism. The first step of the regression showed a significant fit of the model (χ²=14.58, p<0.001). Arab patients were about four times less likely than Jews to stop smoking. The demographic and illness variables entered in the second step did not add significantly to the explained variance beyond ethnicity (Δχ²=5.93, p=0.31, df=5), nor did self-efficacy and fatalism entered in step 3 (Δχ²=1.41, p=0.49, df=2), and none of the variables was significantly associated with smoking cessation. Finally, the inclusion of the interactions in step 4 did not add significantly to the explained variance (Δχ²=6.94, p=0.44, df=7) and none of the interactions reached statistical significance (data not shown in Table 3).

Table 3: The contribution of population group, socio-demographics, illness severity, fatalism, and self-efficacy to smoking behavior (N=140).

Sample characteristics

Of the 200 participants, 78 Arabs (78% of the Arab sample) and 62 Jews (62% of the Jewish sample) were smoking before the index event, and therefore comprise the sample for our analyses. Half (50%, n=39) of the Arab patients continued to smoke after the event, compared with less than a fifth (19.4%, n=12) of the Jewish patients, χ²=14.01, df=1; p <0.000.

Table 1 presents the means and standard deviations of the sample background and study variables, for each of the population groups. It can be seen that the Arab patients were significantly younger, less educated, more religious, and perceived themselves to be of a lower socio-economic status than the Jewish patients. Significant differences were also found with regard to the two psychological variables. Arab patients were found to hold higher levels of fatalism and lower levels of self-efficacy than their Jewish counterparts. In addition (data not shown), though no differences were found between the population groups regarding perceived health status, the objective measure of illness severity shows that the majority of Arab patients (79.5%, n=62) were classified as having a severe illness compared with only half (54.8%, n=34) of the Jewish patients, χ² (1)=9.74, p<0.01.

Table 1: Frequencies, means and standard deviations of the study variables by group [N=140].

Associations with smoking cessation

Table 2 presents the Spearman correlations among the study variables. As can be seen, in addition to the lower education and SES level of the Arab patients compared to the Jewish patients, they also reported higher levels of religiosity, severity of illness and, as hypothesized, lower levels of SE and higher fatalism. Smoking cessation was significantly.

Table 2: Pearson’s correlations of the study’s variables [N=140].

Lower among the Arabs and was correlated with level of education, level of religiosity and fatalism but not with self-efficacy. The less educated, more religious and more fatalistic one was, the more one continued to smoke.

The binary logistic regression model predicting the probability of continuation versus cessation of smoking is presented in Table 3. In the current analysis, only the variables found to significantly differentiate between Arabs and Jews or to be associated with smoking behavior were entered into the regression model in addition to population group: age, perceived socioeconomic status, level of education, religiosity level, illness severity, self-efficacy and fatalism. The first step of the regression showed a significant fit of the model (χ²=14.58, p<0.001). Arab patients were about four times less likely than Jews to stop smoking. The demographic and illness variables entered in the second step did not add significantly to the explained variance beyond ethnicity (Δχ²=5.93, p=0.31, df=5), nor did self-efficacy and fatalism entered in step 3 (Δχ²=1.41, p=0.49, df=2), and none of the variables was significantly associated with smoking cessation. Finally, the inclusion of the interactions in step 4 did not add significantly to the explained variance (Δχ²=6.94, p=0.44, df=7) and none of the interactions reached statistical significance (data not shown in Table 3).

Table 3: The contribution of population group, socio-demographics, illness severity, fatalism, and self-efficacy to smoking behavior (N=140).