OUP user menu

Level of education and risk of heart failure: a prospective cohort study with echocardiography evaluation

Stefan Christensen, Rasmus Mogelvang, Merete Heitmann, Eva Prescott
DOI: http://dx.doi.org/10.1093/eurheartj/ehq435 450-458 First published online: 9 December 2010

Abstract

Aims With increasing socioeconomic disparity in cardiovascular risk factors, there is a need to assess the role of socioeconomic factors in chronic heart failure (CHF) and to what extent this is caused by modifiable risk factors.

Methods and results In a prospective cohort of 18 616 men and women without known myocardial infarction or CHF examined in 1976–78, 1981–83, 1991–94, and 2001–03 in the Copenhagen City Heart Study, we studied the effect of education on CHF incidence. During a median follow-up of 21 years (range 0–31), 2190 participants were admitted to hospital for CHF. Age-adjusted hazard ratio (HR) for intermediary (8–10 years) and high level of education (>10 years) with low (<8 years) as reference was 0.69 (0.62–0.78) and 0.52 (0.43–0.63), respectively, with similar associations in men and women. After adjusting for updated cardiovascular risk factors, corresponding HRs were 0.75 (0.67–0.85) and 0.61 (0.50–0.73). In a random subset of the population examined with echocardiography in 2001–03 (n = 3589), education was associated with left ventricular (LV) hypertrophy, LV dilatation, reduced LV ejection fraction, and severe diastolic dysfunction (P for trend, all <0.05), whereas no association was found for mild diastolic dysfunction (P for trend, 0.61). With the exception of LV hypertrophy, significant associations persisted after adjustment for potential mediating factors.

Conclusion In this cohort study, the level of education was associated with cardiac dysfunction and predicted future hospital admission for CHF. Only a minor part of the excess risk was mediated through traditional cardiovascular risk factors. Strategies to reduce this inequality should be strengthened.

  • Heart failure
  • Epidemiology
  • Morbidity
  • Echocardiography
  • Socioeconomic status

Introduction

It is well accepted that socioeconomic deprivation is associated with coronary heart disease (CHD) but much less is known of the link with the development of chronic heart failure (CHF). Although CHD and CHF share several risk factors, it has been estimated that <50% of CHF is caused by CHD.1 Despite improved treatment, CHF remains a significant health problem with increasing prevalence and associated high health costs.2 With growing disparity in the distribution of cardiovascular risk factors such as smoking, obesity, and physical inactivity, it is important to determine to which extent the risk of CHF is linked with socioeconomic factors and whether this association is mediated by modifiable risk factors. In the few prospective studies that have addressed this issue, methodology and results have differed and several have included only men.37 Moreover, no study has taken changes in risk factor distribution in recent decades into account.

The aim of the present study, the largest to our knowledge, is to examine the longitudinal relationship between socioeconomic factors and risk of hospital discharge with a diagnosis of CHF and to determine whether any associations are mediated by potentially modifiable risk factors, while adjusting for changes in these risk factors during follow-up. In addition, since any analysis of hospital admissions may exaggerate the effect of social deprivation because of a possible lower threshold for hospital admission in the relatively deprived, association with early echocardiography signs of heart failure is explored.

Methods

Study population

The data used in this study are from the Copenhagen City Heart Study (CCHS), which is an ongoing population study in which a random sample of the population living in a area of Copenhagen are invited to participate at regular intervals. Details of the enrolment and examination have been described elsewhere.8 Briefly, the original sample included in 1976–78 comprised 14 223 participants (response rate 74%). In 1981–83, 1991–94, and 2001–03, participants were re-examined and new, primarily young subjects were invited in an attempt to have a study population with representatives from all age groups. A total of 18 974 subjects participated in one or more of the examinations (see Figure 1 for flowchart). Participants reported previous myocardial infarction (MI) at baseline. In addition, all subjects with a hospital admission of MI or CHF according to the Danish National Patient Register from this was established in 1977 until study inclusion were excluded. A total of 358 subjects with previous MI or CHF were thus excluded leaving 18 616 subjects eligible for analyses. The median duration of follow-up was 21 years (range 0–31). Cardiovascular risk factors were assessed at each of the four examinations using a self-administered questionnaire, a physical examination, and paraclinical tests. The Ethics Committee for the Copenhagen area approved the study (KF 100.2039/91).

Figure 1

Flow diagram of the study population. The entire study sample consisted of persons participating in at least one of the four examinations (i.e. some persons participated in multiple examinations) in the Copenhagen City Heart Study who were free of previous MI or CHF at their first examination in the study. The Echocardiography sub study consisted of randomly selected participants from the 4th examination. MI, myocardial infarction; CHF, chronic heart failure.

Socioeconomic variables

Two indicators of socioeconomic position were available: duration of education (<8, 8–10, and >10 years, corresponding to lower primary school, higher primary school, and secondary school, respectively) and household income (categorized into three groups: low, medium, and high). Household income does not accurately reflect the socioeconomic position after retirement, whereas educational attainment is a stable indicator of the socioeconomic position from relatively early in the life course. Education was therefore used as the primary exposure variable throughout.

Covariates

Potentially modifiable cardiovascular risk factors regarded as mediators were measured as follows: smoking habits were categorized into never smokers, ex-smokers, and current smokers of 1–15, 15–24, and >24 cigarettes per day. Physical activity was measured as self-reported leisure time activity in four categories: sedentary, moderate activity 2–4 h/week, moderate activity >4 h/week, and strenuous activity >4 h/week. Heart rate, included as an indicator of physical fitness, and systolic blood pressure were measured in a sitting position after 5 min rest. Treatment for hypertension was self-reported. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2). Alcohol consumption was categorized as abstainers, monthly, weekly, and daily intake. Blood samples were drawn non-fasting and analysed for lipids and glucose. Total cholesterol was available at all examinations, whereas HDL-cholesterol was not available at the first examination and LDL-cholesterol and triglycerides not available at the second. Family history of CHD and diabetes were self-reported.

Echocardiography

In the 2001–03 survey, 6237 subjects were included (response rate 50%) and a random sample consisting of 3654 subjects (59%) underwent an echocardiography. Proportion that underwent echocardiography did not differ by educational attainment, age, gender, or any of the baseline characteristics, thus minimizing the likelihood that any bias has been introduced. Persons with atrial fibrillation, significant valvular stenosis or regurgitation, or missing data on level of education (n = 40) were excluded as were participants with previous MI (n = 25), resulting in the inclusion of echocardiography from 3589 study participants.

Three experienced echo technicians performed echocardiography. Details of recording and analysing have been described elsewhere.9 Systolic dysfunction was defined as left ventricular ejection fraction (LVEF) < 50%. Left ventricular mass index was calculated as the anatomic mass10 divided by body surface area. Left ventricular hypertrophy was defined as LV mass index ≥104 g/m2 for women and ≥116 g/m2 for men.11 Left ventricular dilatation was considered present if the diameter of the LV at end-diastole/height was ≥3.3 cm/m.12 Pulsed-wave Doppler at the apical position was used to record mitral inflow between the tips of the mitral leaflets. Peak velocities of early (E) and atrial (A) diastolic filling and deceleration time of the E-wave (DT) were measured and the E/A ratio was calculated. Mild diastolic dysfunction was defined as E/A < 1 and DT > 240 ms. Severe diastolic dysfunction was defined as DT < 140 ms and E/A<50 years >2.5, E/A50–70 years >2, or E/A>70 years >1.5.13 An abnormal echocardiography examination identified subjects with LV hypertrophy, dilatation, ejection fraction <50%, or mild or severe diastolic dysfunction.

Endpoints

The primary endpoint of this study was first-ever hospital admission with a diagnosis of CHF (ICD8 codes 425.99, 427.09–11, 427.19, and 428.99 until 1 January 1994 and ICD10 codes I11.0, I25.5, I42.0, I42.6, I42.9, and I50.0–9 from 1994 onwards) from the Danish National Patient Register. Follow-up was until first admission for CHF, death, emigration or end of follow-up (9 July 2007), whichever occurred first. Follow-up was more than 99.5% complete.

Analysis strategy

Main exposure variable was level of education. Age, gender, and family history were considered confounders, whereas all other covariates were considered mediating factors. Comparisons across groups were done with one-way analysis of variance (ANOVA), Pearson's χ2, and tests for trend by linear and logistic regression. A two-sided P-value of 0.05 was considered significant. The contribution of education to dichotomized echocardiography findings was analysed by multivariable logistic regression.

Hospital admission data were first analysed by constructing cumulative survival curves for CHF according to exposure categories by the Kaplan–Meier method. For multiple-adjusted survival analyses, we used the Cox proportional hazard model with age (in days) as the underlying time scale with delayed entry (thus ensuring optimal adjustment for age) to examine the effects of education on the risk of CHF. We further adjusted all analyses for changes in admission rates over time by splitting observation time into 10-year periods beginning from 1 January 1976 and adjusting for time period effect.

Analyses were performed by using updated information on education and potential mediating factors whenever these were available (time-dependent variables). As an example, a person could enter the study in 1976 being a smoker. At re-examination in 1992, the person had quit smoking. The person then died in 1997. This person would contribute 16 years of time-at-risk as a smoker (1976–92) with no endpoint and then 5 years as an ex-smoker (1992–97) before being censored at death in 1997. All initial analyses were gender-specific. Univariable Cox's regression analyses, adjusted for age as described, were used to evaluate potential mediators of the association between education and outcome and were analysed as categorical or continuous variables as indicated. Evidence for non-linear trends in increases in the risk of hospital admission was tested by comparing models of co-variables on a continuous scale with models of the variable in quintiles or by adding the squared term. All variables listed above were considered potential confounders or mediators and tested in the multivariable model if they met the criteria of P < 0.15 in the initial age-adjusted model. The effect of education and mediating variables was tested with respect to interaction with gender by a nested log-likelihood test, comparing a model containing the variables as single terms with a model also including the interaction terms.

The assumption of proportional hazards was tested formally by Schoenfeld's residuals. The assumption was violated with regard to the effect of education in both men and women: during follow-up, 1089 women and 1101 men were admitted to hospital with a diagnosis of CHF. In both genders, higher level of education was associated with better survival free of CHF admission in the younger. However, as illustrated by the Kaplan–Meier curve of survival free of hospital admission for CHF (Figures 2 and 3), risk associated with education could not be uniformly described for all age groups. Survival analyses were therefore restricted to age below 80, i.e. all study participants aged 79 or below at the time of their first study participation were included in the analyses and contributed with time-at-risk until reaching an endpoint, death, emigration, or age 80, whichever came first. With this restriction, model assumptions were not violated. One hundred and thirty subjects were aged 80 or above at inclusion. Censoring at age 80 reduced the number of participants from 18 616 to 18 486, person-years of follow-up from 375 609 to 349 492, and the number of hospital admissions from 2190 to 1473.

Figure 2

Survival free of hospital admission for chronic heart failure by level of education (men).

Figure 3

Survival free of hospital admission for chronic heart failure by level of education (women).

Data analyses were conducted using Stata version 10.0 (Stat Corp., College Station, TX, USA).

Results

Baseline characteristics

Of the 13 930 participants free of previous MI or CHF examined in 1976–78, information on level of education was available for 13 902. Table 1 shows the distribution of risk factors among these 7611 women and 6291 men according to level of education. As anticipated, both male and female participants with lowest level of education were older and had significantly more adverse cardiovascular risk profile. Similar associations were seen for risk factors at the subsequent rounds of examination (results not shown).

View this table:
Table 1

Risk factor profile by level of education in 6291 men and 7611 women free of previous myocardial infarction or chronic heart failure in the Copenhagen City Heart Study examined at baseline in 1976–78.

CharacteristicLevel of educationP-value (test for trend)
Low (<8 years)Medium (8–10 years)High (>10 years)
Menn = 2984 (47.4%)n = 2233 (35.5%)n = 1074 (17.1%)
 Age (years)55.1 (11.1)51.3 (12.1)47.7 (14.4)<0.0001
 Low income885 (29.9%)319 (14.5%)145 (13.6%)<0.0001
 Systolic BP (mmHg)141.0 (21.0)140.3 (20.2)137.8 (20.3)<0.0001
 Treated hypertension157 (5.3%)129 (5.8%)48 (4.5%)0.28
 Diabetes146 (5.2%)78 (3.7%)30 (2.9%)0.002
 BMI (kg/m2)26.3 (3.9)25.7 (3.6)24.5 (3.2)<0.0001
 Sedentary722 (24.2%)391 (17.5%)150 (14.0%)<0.0001
 Current smoker2169 (72.7%)1562 (70.0%)677 (63.0%)<0.0001
 Daily alcohol intake1282 (43.2%)889 (39.9%)282 (26.3%)<0.0001
 Cholesterol (mmol/L)6.00 (1.17)5.97 (1.14)5.73 (1.10)<0.0001
 Triglycerides (mmol/L)2.16 (1.53)2.10 (1.45)1.79 (1.13)<0.0001
 Heart rate (b.p.m.)77.3 (13.5)77.8 (12.9)79.2 (13.5)<0.0001
 Family history of MI851 (32.3%)657 (31.2%)314 (30.0%)0.36
 CHF admission532 (17.8%)294 (13.2%)136 (12.7%)<0.0001
Womenn = 3694 (48.5%)n = 2945 (38.7%)n = 972 (12.8%)
 Age (years)54.8 (10.3)51.2 (11.6)46.4 (13.8)<0.0001
 Low income1523 (44.4%)602 (21.4%)206 (21.7%)<0.0001
 Systolic BP (mmHg)137.2 (22.6)133.9 (21.7)128.3 (22.0)<0.0001
 Treated hypertension307 (8.3%)181 (6.2%)47 (4.8%)<0.0001
 Diabetes94 (2.7%)40 (1.4%)11 (1.1%)<0.0001
 BMI (kg/m2)25.4 (4.8)24.2 (4.0)23.0 (3.5)<0.0001
 Sedentary923 (25.0%)450 (15.3%)116 (12.0%)<0.0001
 Current smoker2239 (60.6%)1667 (56.6%)504 (51.9%)<0.0001
 Daily alcohol intake218 (5.9%)170 (5.8%)66 (6.8%)0.49
 Cholesterol (mmol/L)6.41 (1.24)6.21 (1.28)5.84 (1.36)<0.0001
 Triglycerides (mmol/L)1.59 (0.95)1.43 (0.83)1.29 (0.79)<0.0001
 Heart rate (b.p.m.)77.3 (12.2)77.9 (12.4)78.1 (12.3)0.02
 Family history of MI1271 (37.5%)1067 (37.8%)329 (34.6%)0.17
 CHF admission559 (15.1%)363 (12.3%)60 (6.2%)<0.0001
  • Values are number (%) or mean (SD) as indicated. P-value for linear trend for data on a continuous scale from linear regression and for categorical data from logistic regression.

Hospital admission for chronic heart failure

Both educational attainment and household income predicted admission for CHF with stronger associations seen for education (Table 2). Risk for CHF in the group with highest education was approximately half with similar associations in men and women. In further analyses, men and women were pooled. The socioeconomic gradient did not differ over time (test for interaction between time period in four groups and level of education in three groups: P = 0.13). In a model including income and education, both were significantly associated with risk of CHF: hazard ratio (HR) was 0.55 (0.46–0.68) and 0.81 (0.68–0.96) for high vs. low level of education and income, respectively, after adjusting for age, gender, and time period. There was no statistical interaction, i.e. HR associated with low income was similar in the three strata of education.

View this table:
Table 2

Hazard ratios for hospital admission for chronic heart failure by level of education and household income in 18 486 participants below age 80 and free of myocardial infarction and chronic heart failure at baseline in the Copenhagen City Heart Study

Women (n = 9994)Men (n = 8492)
No. of endpointsaHR95% CINo. of endpointsaHR95% CI
Education
 <8 years3751Ref.4791Ref.
 8–10 years2150.720.61–0.852530.670.58–0.78
 >10 years440.500.37–0.69920.530.42–0.66
Household income
 Low3591Ref.3251Ref.
 Medium2020.720.65–0.923490.750.64–0.88
 High620.670.51–0.891480.660.54–0.81
  • Results from Cox's regression analyses with adjustment for age and time period.

  • aNumber of endpoints does not add up to 1473 because of missing data on level of education or household income in some subjects.

Table 3 gives HRs for hospital admission for CHF by educational level with multivariable adjustments for cardiovascular risk factors updated during follow-up. The modifiable risk factors systolic blood pressure, treatment for hypertension, BMI, physical inactivity, and smoking were all associated with risk of developing CHF, as were gender, diabetes, and family history whereas alcohol consumption, heart rate and serum lipids (triglycerides, total-, HDL-, and LDL-cholesterol) were not (results not shown). Association between educational attainment and subsequent risk of developing CHF was attenuated by adjustment but remained highly significant: HR for high vs. low level of education was 0.52 (0.43–0.63) before adjustment and 0.61 (0.50–0.73) after adjustment. Repeated analyses including death from CHF in endpoints did not alter results. To ensure that a socioeconomic gradient in re-participation and thus in updating of risk factors did not bias results, Cox's regression analyses were repeated based only on the 13 902 subjects examined in 1976–78 without updating of risk factors: results were similar (not shown).

View this table:
Table 3

Level of education and risk of hospital admission for heart failure in 18 486 participants below age 80 and free of myocardial infarction and chronic heart failure at baseline in the Copenhagen City Heart Study

No. of endpointsAdjusteda HR95% CIMultivariable adjustedb HR95% CI
Education
 <8 years8541Ref.1Ref.
 8–10 years4680.690.62–0.780.750.67–0.85
 >10 years1360.520.43–0.630.610.50–0.73
  • Results from Cox's regression analysis with multivariable adjustment for CVD risk factors and changes in CVD risk factors during follow-up.

  • aAdjusted for age, gender, and time period.

  • bAdjusted for age, gender, time period, systolic blood pressure, medical treatment for hypertension, diabetes, BMI, smoking, physical inactivity, and interaction between smoking and gender (P = 0.001). Alcohol consumption, heart rate, and plasma lipids (triglycerides, total-, LDL-, and HDL-cholesterol) were not associated with CHF in the multivariable adjusted model.

To address the issue of possible interim coronary events during follow-up, analyses were repeated after excluding 400 participants with an MI after study inclusion but prior to admission for CHF. Results were similar: age-, gender-, and time period-adjusted HRs for medium and high level of education were 0.69 (0.60–0.78) and 0.54 (0.44–0.67), respectively. Similarly, after excluding further 1205 participants who suffered an MI at any time during the follow-up results were also unaffected: corresponding HRs were 0.69 (0.61–0.79) and 0.52 (0.42–0.64), respectively.

Echocardiography

Results of echocardiography by level of education in 3589 subjects after excluding subjects with previous MI from the survey in 2001–03 are shown in Table 4. Overall, for each of the indices of cardiac abnormalities defined, proportion increased with decreasing educational attainment, resulting in one or more abnormalities found in 27.7% of the lowest educated vs. 10.5% of the highest (P < 0.001). However, there were also large age differences between groups. In logistic regression adjusting for age and gender, educational attainment remained associated with all indicators of abnormal echocardiography with the exception of mild diastolic dysfunction (Table 5). After further adjustment for potential confounders and mediators, associations were attenuated but remained statistically significant for LV dilatation, LVEF, severe diastolic dysfunction, and any abnormality.

View this table:
Table 4

Abnormal echocardiography findings according to level of education in 3589 study participants free of myocardial infarction and chronic heart failure in the fourth examination in the Copenhagen City Heart Study

Level of educationP-value
Low (<8 years, n = 842)Medium (8–10 years, n = 1296)High (>10 years, n = 1451)
Age69.6 (8.8)62.5 (13.3)47.3 (16.1)<0.001
Echocardiography
 LV hypertrophy142 (16.9%)151 (11.7%)100 (6.9%)<0.001
 LV dilatation81 (9.6%)66 (5.1%)43 (3.0%)<0.001
 LVEF < 50%17 (2.0%)12 (0.9%)3 (0.2%)<0.001
 Mild diastolic dysfunction60 (7.1%)64 (4.9%)33 (2.3%)<0.001
 Severe diastolic dysfunction19 (2.3%)11 (0.9%)6 (0.4%)<0.001
 Abnormal echocardiography233 (27.7%)243 (18.8%)153 (10.5%)<0.001
  • Values are number (%) or mean (SD) as indicated. P-value from one-way ANOVA or χ2. LV, left ventricular; EF, ejection fraction. Abnormal echocardiography defined as one or more of the following: LV hypertrophy, LV dilatation, LVEF < 50%, and mild or severe diastolic dysfunction.

View this table:
Table 5

Association between level of education and abnormal echocardiography findings in 3589 participants free of myocardial infarction and chronic heart failure in the fourth examination in the Copenhagen City Heart Study

AdjustedaMultivariable adjustedb
HR95% CIHR95% CI
LV hypertrophy
 <8 years1Ref.1Ref.
 8–10 years0.770.59–1.000.810.61–1.07
 >10 years0.670.49–0.910.810.58–1.13
 Test for trendc0.010.19
LV dilatation
 <8 years1ref1Ref.
 8–10 years0.52(0.37–0.74)0.620.43–0.90
 >10 years0.34(0.21–0.53)0.470.29–0.76
 Test for trendc<0.0010.001
LVEF < 50%
 <8 years1Ref.1Ref.
 8–10 years0.550.26–1.180.670.31–1.47
 >10 years0.200.05–0.750.260.07–0.96
 Test for trendc0.0090.04
Mild diastolic dysfunction
 <8 years1Ref.1Ref.
 8–10 years0.930.64–1.360.960.66–1.40
 >10 years0.890.56–1.420.890.56–1.42
 Test for trendc0.610.63
Severe diastolic dysfunction
 <8 years1Ref.1Ref.
 8–10 years0.470.22–1.000.440.20–0.94
 >10 years0.420.16–1.140.350.13–0.98
 Test for trendc0.040.02
Any abnormal echocardiography
 <8 years1Ref.1Ref.
 8–10 years0.730.58–0.920.790.62–1.01
 >10 years0.610.47–0.800.750.56–0.99
 Test for trendc<0.0010.04
  • LV, left ventricular; EF, ejection fraction. Abnormal echocardiography defined as one or more of the following: LV hypertrophy, LV dilatation, LVEF < 50%, and mild or severe diastolic dysfunction.

  • aAdjusted for age and gender.

  • bAdjusted for age, gender, systolic and diastolic blood pressure, medical treatment for hypertension, diabetes, BMI, smoking, alcohol consumption, plasma lipids (triglycerides, total-, LDL-, and HDL-cholesterol), physical inactivity, and family history of MI.

  • cLog-likelihood test for linear trend.

Discussion

The main finding of this study was the relationship between educational level and hospital admission for CHF with an almost 50% lower risk with the highest level of education compared with the lowest in both men and women. Participants with low level of education in general had a poorer risk factor profile but this explained only a minor part of the excess risk and a statistically significant stepwise decrease in the risk of CHF with higher levels of education persisted after multivariable adjustment. Correspondingly, in cross-sectional data, early stages of cardiac dysfunction assessed by echocardiography were associated with educational level.

We have identified four prospective studies of the effect of socioeconomic factors on population risk of CHF, which all report results similar to ours.57,14 In the NHANES, in which ∼13 000 subjects were studied in the period from 1971 to 1992, less than high school education conveyed a relative risk (RR) of 1.22 (1.04–1.42) of hospital admission or death from CHF after multivariable adjustment.5 In the Renfrew/Paisley study, which followed ∼15 000 middle-aged individuals for 20 years after 1972–76, the most deprived individuals had an RR of CHF admission within 20 years of baseline screening of 1.39 (1.04–2.01) compared with the most affluent.7 Neither of these studies found gender differences in this socioeconomic gradient. In a Swedish study comprising 6999 middle-aged men followed for 28 years, unadjusted RR comparing highest with lowest occupational class was 1.92 (1.50–2.45) and adjusted 1.72 (1.34–2.20).14 Another Swedish study of 50-year-old men with adjustment for interim MI found adjusted HR of 1.98 (1.07–3.68) for lowest vs. highest education.6 Our results are consistent with these other studies and further complement them by showing a similar gradient in echocardiography indicators of cardiac dysfunction. We found that educational attainment and income were independently associated with heart failure development with strongest associations seen for education after mutual adjustment. This may partly reflect imprecision in the two measures, particularly household income as discussed above, but indicates that estimating risk based on one socioeconomic measure may underestimate the true socioeconomic gradient.

There are several ways in which socioeconomic deprivation may contribute to excess CHF morbidity. Excess risk is clearly related to uneven distribution of cardiovascular risk factors as also seen in the present study. A large proportion of CHF is caused by CHD and there is a well-known socioeconomic gradient in the risk of CHD. Similar to previous studies,35,7,1423 we found male sex, hypertension, diabetes, smoking, and obesity to be important risk factors for CHF, but adjustment for these risk factors did not account for the excess risk. In an attempt to determine whether the socioeconomic gradient was partly caused by differences in changes in cardiovascular risk factors during follow-up, we adjusted risk estimates for changes in these factors over the follow-up period. However, adjustment only attenuated HR from 0.52 to 0.61 for the highest and from 0.69 to 0.75 for intermediary educational attainment, indicating that only a minor part of the excess risk in the socially deprived was mediated by difference in these cardiovascular risk factors. Other possible explanations include increased risk of developing disease through poorer preventive measures, lower medical compliance, and lower threshold for hospital admission. Previous studies have suggested that lower educational levels are related to limited access to higher-quality healthcare and poor adherence to treatment of cardiovascular diseases and CHF. Although the Danish national health system has universal coverage and there are no private hospitals or clinics that admit patients for treatment for heart failure, there are informal barriers and cost of medication that may contribute to inequity in disease management. Poorer treatment adherence and compliance in CHD including poorer treatment of acute coronary syndrome, later presentation of MI, and less use of secondary prevention treatment may cause more deprived patients to have higher risk of re-MI and subsequent development of CHF. Poor compliance may also lead to more frequent hospital admission due to poor disease control once CHF has developed. Other possibilities include higher threshold for taking action on symptoms in the least educated group, leading to patients delay and worsening of symptoms. Psychosocial factors that were not measured may also play a role.

Prevalence of both systolic and diastolic dysfunction in the present study was lower than reported in a previous study.24 This is likely to be caused both by differences in the underlying population, selection of study participants, age and gender distribution, and echocardiography evaluation methods employed. Conventional echocardiography, as employed in the present study, can be used to identify mild and severe diastolic dysfunction. Tissue Doppler imaging, as well as pulmonal venous flow and Valsalva manoeuvre, can (sometimes) be helpful in the identification of pseudo-normalization. Without these measures, we were unable to differentiate between Grade 2 (pseudo-normal) and normal diastolic function. However, we feel that it is unlikely that many of our participants were pseudo-normalized without concomitant structural heart disease or decreased LVEF (abnormal echocardiography).

The number of subjects with abnormal echocardiography findings was limited particularly in the group with high education. Nevertheless, recordings show a socioeconomic gradient in all of the markers of systolic and diastolic dysfunction with the exception of mild diastolic dysfunction.

The results presented imply that the socioeconomic gradient in CHF is not only present in advanced stages of disease as measured by hospital admission or death but also present already at subclinical stages. Although it is has not been proven that systolic dysfunction leads to systolic heart failure and diastolic dysfunction to heart failure with preserved ejection fraction (HFPEF), systolic and diastolic dysfunction have been shown to be markers of increased mortality in population-based studies.24,25 Thus, the mechanisms responsible for the socioeconomic gradient should probably not only be sought in threshold for hospital admission or poor medicine compliance but also in the early development of both systolic heart failure and HFPEF. We know of no previous study examining socioeconomic gradient in early structural changes of the heart in the general population and these findings add significantly to our knowledge.

Strengths and limitations

The strengths of this study include the prospective design, large number of outcomes, detailed registration of cardiovascular risk factors, updating of risk factors, complete follow-up regarding hospital admission, and the availability of echocardiographic recordings in a subset. Limitations include that our case ascertainment relied solely on hospital records and we did not identify milder cases treated outside of hospital. The diagnosis of CHF from our national hospital records have been shown to have low sensitivity (29%) but high specificity (99%) when validated against criteria based on symptoms, clinical signs, and echocardiography evaluation.26 Thus, we probably underestimated the true incidence of CHF but internal validity is high. Diagnostic inaccuracy may have biased results if this is related to socioeconomic status but we find this unlikely. Another limitation is the lack of information on the type of CHF at admission (preserved vs. low ejection fraction), or cause (valvular heart disease, hypertension, CHD, alcoholic, metabolic, inflammatory) or other clinical data. In a recent Swedish study, however, risk associated with occupational class did not differ when distinguishing between CHF with or without ischaemic origin.14 The socioeconomic gradient found in both early abnormalities related to systolic and diastolic dysfunction corroborates this.

In our study, neither education nor income predicted hospital admission for CHF in the oldest subjects and these were excluded from analyses. Another published study of risk factors for incident CHF in 1749 subjects aged 65 or older (mean age 75) found no association between education and CHF.3 This is likely to be caused by selection mechanisms and competing mortality risks in the very old.

Conclusion

In summary, we show that level of education is associated with early systolic and diastolic dysfunction assessed by echocardiography and is a strong predictor of future hospital admission for heart failure. Only a minor part of the excess risk was mediated by differences in potentially modifiable cardiovascular risk factors. Further studies are needed to explain the remaining excess just as strategies to reduce this inequality should be strengthened.

Funding

This work was supported by unrestricted grants from The Danish Heart Foundation (06-10-B381-A1313-22314F) and The Velux Foundation. The authors are independent from funders.

Conflict of interest: none declared.

References

View Abstract