Elsevier

Social Science & Medicine

Volume 69, Issue 8, October 2009, Pages 1272-1280
Social Science & Medicine

The relative contributions of hostility and depressive symptoms to the income gradient in hospital-based incidence of ischaemic heart disease: 12-Year follow-up findings from the GLOBE study

https://doi.org/10.1016/j.socscimed.2009.07.031Get rights and content

Abstract

There is evidence to support the view that both hostility and depressive symptoms are psychological risk factors for ischaemic heart disease (IHD), additional to the effects of lifestyle and biomedical risk factors. Both are also more common in lower socioeconomic groups. Studies to find out how socioeconomic status (SES) gets under the skin have not yet determined the relative contributions of hostility and depression to the income gradient in IHD. This has been examined in a Dutch prospective population-based cohort study (GLOBE study), with participants aged 15–74 years (n = 2374). Self-reported data at baseline (1991) and in 1997 provided detailed information on income and on psychological, lifestyle and biomedical factors, which were linked to hospital admissions due to incident IHD over a period of 12 years since baseline. Cox proportional hazard models were used to study the contributions of hostility and depressive symptoms to the association between income and time to incident IHD. The relative risk of incident IHD was highest in the lowest income group, with a hazard ratio of 2.71. Men on the lowest incomes reported more adverse lifestyles and biomedical factors, which contributed to their higher risk of incident IHD. An unhealthy psychological profile, particularly hostility, contributed to the income differences in incident IHD among women. The low number of IHD incidents in the women however, warrants additional research in larger samples.

Introduction

Ischaemic heart disease (IHD) is the number one cause of death worldwide (Murray and Lopez, 1997, WHO, 2003). There is a strong socioeconomic gradient in IHD morbidity and mortality in the Western European countries disfavouring the lower socioeconomic groups (Dalstra et al., 2005, Huisman et al., 2005, Mackenbach et al., 2003). There may be gender differences in this gradient, although these differences seem to depend on the outcome as the gradient for cardiac morbidity seems to be steeper in women (Loucks et al., 2007, Thurston and Kubzansky, 2007), and that for cardiac mortality has been found to be steeper in men (Mackenbach et al., 2003).

In order to allow interventions aimed at reducing heart disease inequalities to be tailored, research has focused on explaining this gradient for several decades. Evidence points in the direction of classical risk factors for heart disease, such as lifestyle factors (smoking, alcohol consumption and physical inactivity) and biomedical factors (obesity, hypertension and diabetes), mediating the socioeconomic gradient in heart disease. However substantial the contribution of these classical risk factors may be (reported contributions vary from 16% to as much as 60%) (Ferrie et al., 2005, Lynch et al., 1996, Van Lenthe et al., 2002), a moderate part of the gradient remains unexplained. Psychological risk factors, which have been the subject of numerous studies in the last decade, might shed further light upon the pathways by which socioeconomic status gets under the skin.

Specifically, hostility and depressive symptoms have emerged as potential contributing factors to the socioeconomic gradient in IHD. Both are considered important psychological risk factors for incident heart disease, although the evidence is stronger for depressive symptoms and depression than for hostility (Kuper et al., 2002, Miller et al., 1996, Penninx et al., 1998, Rugulies, 2002). Also, both depressive symptoms and hostility are more prevalent in lower than in higher socioeconomic groups (Carroll et al., 1997, Lorant et al., 2003, Scherwitz et al., 1991, Stansfeld et al., 2003), perhaps due to a long-term exposure to adverse circumstances.

Few studies have investigated the influence of depressive symptoms on the socioeconomic gradient in heart disease and results have so far been inconclusive. Whereas Lynch and colleagues (Lynch et al., 1996) found that psychological and social risk factors, including depression, attenuated the association between income and cardiovascular mortality in a population of middle-aged Finnish men, Thurston and colleagues (Thurston, Kubzansky, Kawachi, & Berkman, 2006) concluded that depression did not appear to mediate the relation between educational attainment and incident CHD. The contribution of hostility in a general population has, to our knowledge, only been examined longitudinally by Schrijvers and colleagues (Schrijvers, Bosma, & Mackenbach, 2002), who found that a substantial part of the educational gradient in general perceived health could be ascribed to the intermediary effects of hostility in both men and women.

Although hostility and depressive symptoms differ conceptually, in the sense that hostility is an outward-focused negative emotion, directed at others and therefore reactive by nature, whereas depressive symptoms are more inward-focused, directed at oneself (De Vogli et al., 2007, Suarez et al., 1998), pathways by which they affect IHD might be similar. Both are responses to stressful circumstances, which are more common in lower SES groups, and both evoke a sustained physiological reactivity characterised by activation of the sympathetic nervous system and hypersecretion of the stress hormone cortisol (Rozanski, Blumenthal, & Kaplan, 1999). This, in turn, may have direct long-term adverse effects on the immune and cardiovascular systems (Adler and Ostrove, 1999, Kristenson et al., 2004). Depressive symptoms and hostility might also affect cardiovascular health indirectly through lifestyle as depressed and hostile individuals tend to engage in less healthy behaviours, such as smoking and heavy drinking (Laitinen, Ek, & Sovio, 2002).

Using 12-year longitudinal data from the Dutch GLOBE study, we examined the relative contributions of hostility and depressive symptoms to the income gradient in IHD (in addition to the contribution of the classical risk factors) and the possible gender differences therein. Given the above-mentioned associations, we hypothesised that individuals with lower incomes would be at higher risk of IHD, because they are more likely to suffer from depressive symptoms or to have hostile cognitions than their higher income counterparts. The aim of the study was to explore the complex pathways underlying socioeconomic inequalities in incident IHD and thereby to identify opportunities for prevention and intervention aimed at reducing these inequalities.

Longitudinal data were gathered between 1991 and 2003 as part of the prospective cohort study called GLOBE, a Dutch acronym for ‘Health and Living conditions of the Population of Eindhoven and Surroundings’. The GLOBE study aims to explain socioeconomic inequalities in health. The design and rationale of the study have been published elsewhere (Mackenbach, van de Mheen, & Stronks, 1994). Briefly, a postal survey was conducted in 1991 among 27,070 non-institutionalised individuals aged 15–74 years, with a Dutch nationality. With a response rate of 70.1%, 18,973 respondents completed this baseline questionnaire. There were no significant differences in non-response by socioeconomic position, age, sex, marital status or degree of urbanisation (Mackenbach et al., 1994). Two sub-samples were drawn to gather more detailed risk factor information by means of questionnaires and interviews between 1991 and 1997. One sub-sample was randomly drawn and consisted of 2802 respondents (response rate 79.4%) whereas in the other sub-sample, individuals with chronic diseases were oversampled (n = 2867, response rate 72.3%). Because the non-random sampling in the latter sample compromises its representativeness of the source population and possibly the estimation of effect sizes, weight factors were calculated for all respondents to re-establish the representativeness of the sample (Mackenbach, Simon, Looman, & Joung, 2002).

The interview and questionnaire data were linked to data from the municipal register on mortality and addresses (or changes of address), and with hospital admission data from five hospitals in the catchment area of the GLOBE study between 1991 and 2003. The catchment area of the hospitals was defined as the area around the city of Eindhoven, encompassing the villages where at least 90% of the population would be admitted to one of the five hospitals, when requiring hospital admission. Linkage was based on date of birth, gender and postal code. If identical combinations of these key variables were found, patient numbers, hospital codes and information on health insurance were checked to determine whether the admission was a re-admission or whether different individuals had been admitted. When re-admissions where found, only the first admissions were selected (Van Lenthe et al., 2002). Data on hospital admission were complete for all but 19 respondents.

The present study is based on the questionnaire and interview data from the two sub-samples. A total of 4109 (72.5%) respondents completed the 1991 baseline and 1997 follow-up measurements, which included detailed risk factor information. We excluded (a) individuals who were included in the sample because of known heart disease (chronic diseases were oversampled in one of the sub-samples) and those who reported heart disease at baseline or in the 5 years prior to baseline (n = 962), (b) respondents whose hospital admission data or municipal register data were missing (n = 19), (c) respondents with missing data for the risk factors under study (n = 345) and (d) respondents with missing income data (n = 409). The final sample comprised 2374 individuals.

Hospital-based incident ischaemic heart disease (referred to below as incident IHD) was determined according to ICD-9 classification codes 410–414. An incident IHD event was defined as a first hospital admission diagnosis of IHD after baseline. Fatal events were considered cases if death occurred after admission. Pre-hospital IHD deaths were not considered cases, since data on cause-specific mortality were not available. Incident IHD was measured continuously from baseline until 22 December 2003.

All other measures were assessed at baseline in 1991, except for hostility, which was measured in 1997 only.

Income and education were used to determine SES. Net household income was adjusted for the number of adults and children who were part of the household (Hagenaars, Vos, & Zaidi, 1994) and classified into thirds (based on tertiles). Three levels of education were distinguished: higher secondary education, higher professional education or university education (level 1), senior secondary vocational education or general secondary education (level 2), and lower secondary vocational education or primary education only (level 3).

Hostility was measured by the 6-item version of the hostility sub-scale of the Aggression Questionnaire (Buss and Perry, 1992, Meesters et al., 1996, Schrijvers et al., 2002) in 1997. This questionnaire has been proven reliable and valid (Buss and Perry, 1992, Felsten and Hill, 1999, Gerevich et al., 2007, Harris, 1997, Meesters et al., 1996). A seven-month test-retest stability of 0.67 has been reported by Harris and colleagues (Harris, 1997). An example item is ‘I wonder why sometimes I feel so bitter about things’. Items were rated on a 5-point Likert scale, ranging from 1 (‘totally disagree') to 5 (‘totally agree'). Items were summed to obtain an overall score (6–30) with higher scores indicating more hostility. Internal consistency, as measured by Cronbach's α, was 0.71. The scores were categorised into tertiles, the highest tertile reflecting the greatest hostility.

The 9-item emotional reactions sub-scale of the Nottingham Health Profile (NHP) (Erdman et al., 1993, Hunt et al., 1986) was used as a proxy for depressive symptoms. The NHP has been found to be valid and reliable (Coons, Rao, Keininger, & Hays, 2000). Comparison with the General Health Questionnaire (GHQ) has shown that a high NHP-ER score is indicative of depressed mood as measured with the GHQ, at least in stroke patients (Ebrahim, Barer, & Nouri, 1986). An example item is ‘I wake up feeling depressed’. Using predefined item weights, the dichotomous items were converted into a continuous scale score ranging from 0 to 100 (Hunt et al., 1986). Cronbach's α of the continuous scale was 0.78. Based on the skewed distribution (81% scored 0), we dichotomised this score (0 = no depressive symptoms, > 0 = depressive symptoms).

Classical risk factors were self-reported smoking, alcohol consumption, physical activity, hypertension, diabetes and obesity. Smoking status was indicated as never smoker, former smoker, pipe/cigar smoker, current smoker  20 cigarettes a day or current smoker > 20 cigarettes a day. Individuals were categorised into five groups for alcohol consumption: total abstainers, light, moderate, excessive and very excessive drinkers (Droomers, Schrijvers, Stronks, van de Mheen, & Mackenbach, 1999). Physical activity was based on the number of hours spent per week on sports activities and on gardening, cycling and walking. It was classified into inactivity, low, moderate or high physical activity (Droomers, Schrijvers, & Mackenbach, 2001). Hypertension and diabetes were self-reported and dichotomised into ‘never had the condition’ versus ‘(ever) had the condition’. Body Mass Index (BMI) was calculated by dividing self-reported weight (kg) by the square of self-reported height (m), and two groups were distinguished with obesity defined as BMI  30.

Age, sex and marital status (married, single, divorced, widowed) were the basic confounders included in the analyses.

First, the association between all variables and sex was examined by means of t-testing and chi-square testing, depending on the variable measurement levels. Cox proportional hazard models were used to study the association of income with incident IHD between 1991 and 2003. Survival time was defined as the time from baseline to the date of (a) first hospital admission due to diagnosed IHD, (b) death, (c) emigration outside the catchment area or (d) complete follow-up (22 December 2003). All total group analyses were adjusted for age, sex and marital status. The interaction between income and gender for the income differences in IHD was not significant. However, because of significant interactions between gender and income for the income differences in hostility (p = 0.01) and between gender and hostility for the association between hostility and IHD (p = 0.00), all results were reported for men and women separately. There were no significant interactions with age. The reference model included income and the basic confounders. Survival curves, indicating the cumulative proportion of participants without an admission during the follow-up interval, were estimated for men and women.

Second, we examined the associations of income with hostility and depressive symptoms by means of logistic regression and of hostility and depressive symptoms with incident IHD by means of Cox regression.

Third, we compared the basic model including income and the basic confounders with models including classical risk factors (model 2) depressive symptoms (model 3), depressive symptoms and classical risk factors (model 3a), hostility (model 4), hostility and classical risk factors (model 4a), depressive symptoms and hostility (model 5) and classical risk factors, depressive symptoms and hostility (model 6). The contribution of each of the added factors was expressed as the reduction in hazard ratio (HR) for income after inclusion of the additional factors (Baron & Kenny, 1986). This reduction was calculated as follows: (HRrestricted model  HRextended model)/(HRrestricted model  1) (Lynch et al., 1996). In order to verify the robustness of the results, the analyses were repeated with education as an SES measure, and with 1997 as the baseline year, as hostility was measured in 1997 only.

Section snippets

Results

A total of 106 respondents suffered incident IHD between 1991 and 2003. Men were more likely to suffer an IHD event (n = 75) than women (n = 31) (Table 1). Further, men were older, had higher incomes, smoked more and consumed more alcohol, and were more likely to have diabetes, compared to women. Women were more likely to report depressive symptoms (20.1% versus. 17.0% in men), although the difference was not significant.

Lower income levels were associated with a higher risk of incident IHD in both

Discussion

To our knowledge, this study is the first to stratify for gender in examining the contributions of both hostility and depressive symptoms to income differences in IHD. The results of the GLOBE study show that the lowest income group had a significantly higher risk of incident IHD than the highest income group. Persons in the lowest income group were more hostile and reported more depressive symptoms. The results indicate that there were gender differences in the contributions of the classical

Methodological limitations

First, hospital-based incident IHD as the outcome excludes silent myocardial ischaemia (minor asymptomatic infarctions) and pre-hospital deaths due to IHD. Data on cause-specific mortality were not available for this study. However, we assume that a substantial proportion of pre-hospital deaths is caused by cardiac events (IHD being the number one cause of death). Pre-hospital deaths due to IHD are generally more common in lower SES groups (Morrison, Woodward, Leslie, & Tunstall-Pedoe, 1997).

Conclusion

Although potentially unhealthy psychological profiles, as characterised by hostility and depressive symptoms, were more prevalent in the lower income groups, our findings suggest gender differences in pathways through which socioeconomic adversity gets under the skin. An unhealthy psychological profile, particularly hostility, contributed to the income differences in incident IHD in women only whereas the classical risk factors explained more of the income differences in incident IHD in men.

References (67)

  • J.P. Mackenbach et al.

    A prospective cohort study investigating the explanation of socio-economic inequalities in health in The Netherlands

    Social Science & Medicine

    (1994)
  • C. Meesters et al.

    Psychometric evaluation of the Dutch version of the aggression questionnaire

    Behaviour Research and Therapy

    (1996)
  • C.J. Murray et al.

    Mortality by cause for eight regions of the world: global burden of disease study

    Lancet

    (1997)
  • Y. Okura et al.

    Agreement between self-report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure

    Journal of Clinical Epidemiology

    (2004)
  • B.W. Penninx et al.

    Cardiovascular events and mortality in newly and chronically depressed persons > 70 years of age

    American Journal of Cardiology

    (1998)
  • R. Rugulies

    Depression as a predictor for coronary heart disease. A review and meta-analysis

    American Journal of Preventive Medicine

    (2002)
  • F.J. Van Lenthe et al.

    Material and behavioral factors in the explanation of educational differences in incidence of acute myocardial infarction: the Globe study

    Annals of Epidemiology

    (2002)
  • N.E. Adler et al.

    Socioeconomic status and health: what we know and what we don't

    Annals New York Academy of Sciences

    (1999)
  • C. Bakx et al.

    First myocardial infarction in a Dutch general practice population: trends in incidence from 1975–2003

    British Journal of General Practice

    (2005)
  • R.M. Baron et al.

    The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations

    Journal of Personality and Social Psychology

    (1986)
  • A.T. Beck et al.

    Psychometric properties of the Beck Depression Inventory: twenty-five years of evaluation

    Clinical Psychology Review

    (1999)
  • H. Bosma

    Socio-economic differences in health: are control beliefs fundamental mediators?

  • H. Bosma et al.

    Socioeconomic inequalities in mortality and importance of perceived control: cohort study

    British Medical Journal

    (1999)
  • H. Bosma et al.

    Job control, personal characteristics, and heart disease

    Journal of Occupational Health Psychology

    (1998)
  • A.H. Buss et al.

    The aggression questionnaire

    Journal of Personality and Social Psychology

    (1992)
  • D. Carroll et al.

    The relationship between socioeconomic status, hostility, and blood pressure reactions to mental stress in men: data from the Whitehall II study

    Health Psychology

    (1997)
  • S.J. Coons et al.

    A comparative review of generic quality-of-life instruments

    Pharmacoeconomics

    (2000)
  • P.T. Costa et al.

    Professional manual for the NEO PI-R and NEO-FFI

    (1992)
  • J.A. Dalstra et al.

    Socioeconomic differences in the prevalence of common chronic diseases: an overview of eight European countries

    International Journal of Epidemiology

    (2005)
  • M. Droomers et al.

    Educational level and decreases in leisure time physical activity: predictors from the longitudinal GLOBE study

    Journal of Epidemiology and Community Health

    (2001)
  • S. Ebrahim et al.

    Use of the Nottingham Health Profile with patients after a stroke

    Journal of Epidemiology and Community Health

    (1986)
  • R.A. Erdman et al.

    The Dutch version of the Nottingham Health Profile: investigations of psychometric aspects

    Psychological Reports

    (1993)
  • J.E. Ferrie et al.

    Self-reported economic difficulties and coronary events in men: evidence from the Whitehall II study

    International Journal of Epidemiology

    (2005)
  • Cited by (10)

    View all citing articles on Scopus

    The GLOBE study is supported by grants from the Dutch Ministry of Public Health, Welfare, and Sports, and the Dutch Prevention Fund. The study is conducted in close collaboration with the Public Health Services of the Dutch city of Eindhoven and the South-East Brabant region. The authors thank Roel Faber for carefully constructing the database and Casper Looman for designing the statistical procedures involved in the weighting of the data, and for his valuable comments on these procedures.

    View full text