Psychosocial Factors and Risk of Incident Heart Failure: The Multi-Ethnic Study of Atherosclerosis (MESA)

Abstract

Background

Heart failure is a major source of morbidity and mortality in the United States. Psychosocial factors have frequently been studied as risk factors for coronary heart disease, but not for heart failure.

Methods and Results

We examined the relationship between psychological status and incident heart failure among 6,782 individuals from the Multi-Ethnic Study of Atherosclerosis (MESA). Anger, anxiety, chronic stress, depressive symptoms, and hostility were measured using validated scales and physician reviewers adjudicated incident heart failure events. Cox proportional hazards models were used to adjust for relevant demographic, behavioral, and physiological covariates. Interactions by age, race, sex, and self-reported health were examined in exploratory analyses. During a mean follow up of 9.3 years, 242 participants developed incident heart failure. There was no association between psychosocial factors and heart failure hazard ratios (95% CI) for highest vs. lowest quartile: anger=1.14 (0.81-1.60), anxiety=0.74 (0.51-1.07), chronic stress=1.25 (0.90-1.72), depressive symptoms=1.19 (0.76-1.85), and hostility=0.95 (0.62-1.42). In exploratory analysis, among the participants reporting fair/poor health at baseline, those reporting high vs. low levels of anxiety, chronic stress, and depressive symptoms had 2-fold higher risk of incident heart failure, but there was no association for those with good/very good/excellent self-reported health.

Conclusions

Overall these psychosocial factors were not significantly associated with incident heart failure. However, for participants reporting poor health at baseline, there was evidence that anxiety, chronic stress, and depressive symptoms were associated with increased risk of heart failure. Future research with greater statistical power is necessary to replicate these findings and seek explanations.

Heart failure (HF) affects 5.1 million Americans age 20 and older, and given the aging U.S. population its prevalence is projected to increase 25% by the year 2030.¹ Because HF is a major source of morbidity and mortality, new potentially modifiable risk factors need to be identified.

Psychosocial risk factors have been frequently studied in relation to coronary heart disease. In observational studies, high levels of depression and anxiety have consistently been associated with incident coronary heart disease, while associations with hostility and social support have been mixed² However, psychosocial factors have been examined less often in HF. Among HF patients, depression is the most commonly researched psychosocial risk factor. Approximately 1 in 5 HF patients meet the criteria for major depression, with a higher prevalence in patients with more severe HF.³ Depression is also a predictor of repeated hospitalization in HF patients, and is an independent risk factor of cardiovascular and all-cause mortality in prevalent HF cases, though many early studies were cross-sectional or had short follow-up periods.^4-6 Fewer studies have examined whether depression is associated with incidence of HF.^{3, 7-12} Among those that have, the results have been mixed, though associations were stronger in populations at high-risk for HF.^{8, 9}

Few other psychosocial factors have been examined in relation to risk of incident HF. Anxiety has also been examined in association with incident heart failure, with mixed results, and anger has been modestly associated with incident heart failure in one study.^{10, 11, 13} Lack of social support has also been associated with greater incident HF risk, ^{3, 14} though as noted in a recent meta-analysis several studies had short follow-up periods and the majority were composed mostly of men.⁴ Whether psychosocial factors such as chronic stress and hostility are associated with HF incidence is presently unknown.

Several plausible pathways could explain the relationship between psychosocial factors and HF. One pathway suggests that people with adverse levels of psychosocial factors may be more likely to experience a variety of poor physiological effects that may lead to HF. Potential biologic pathways include inflammation, endothelial dysfunction, platelet activity, hormones, and brain-derived neurotrophic factor (BDNF).¹⁵ For example, depression causes heightened activation of the sympathetic nervous system^{16, 17}, which is involved in the pathogenesis of HF¹⁸, specifically left ventricular dysfunction and renal sodium retention. Additionally, stress has been associated with impaired left ventricular function.¹⁹ Also, depression and anxiety are associated with elevated inflammatory marker levels, which have been associated with HF.¹⁵ Another plausible pathway is that people with psychosocial problems are less likely to adhere to medical and behavioral guidelines, which makes them more likely to develop diseases such as HF.²⁰

Because results from previous research have been mixed and mostly focused on depression, additional data are needed to elucidate the association between psychosocial factors and incident HF, especially in diverse samples of individuals who were healthy at baseline. Therefore, the aim of this study is to determine whether psychosocial factors, namely anger, anxiety, chronic stress, depressive symptoms and hostility are associated with incident HF. Based on previous research, we hypothesized that higher levels of these psychosocial factors would be associated with greater risk of incident HF.

Methods

The Multiethnic Study of Atherosclerosis (MESA) is a prospective cohort study designed to evaluate risk factors for clinical and subclinical cardiovascular diseases in several racial/ethnic groups.²¹ The study began in July 2000 and recruited 6,814 adults free of clinical cardiovascular disease and aged 45-84 years from six field centers across the United States: Baltimore, MD; Chicago, IL; Los Angeles, CA; New York, NY; Saint Paul, MN; and Winston-Salem, NC. There have been five total examinations with similar protocols. MESA participants with data on psychosocial factors at baseline were included in the present analysis (n = 6,782). Local institutional review boards approved study protocols, and all participants gave written informed consent.

Exposures

Anger, anxiety,²² chronic stress,²³ and depressive symptoms²⁴ were measured via questionnaires administered at MESA Exam 1, which occurred in 2000-2002.Hostility²⁵ was measured at Exam 2, which occurred in 2002-2004. Descriptions of these measures can be found in Table 1. For all of the psychosocial measures, individual questions were summed to create scores, and higher scores represent more severe symptoms. Some of these measures have been found valid and reliable in older populations. ²⁶ For the primary analysis, the psychosocial measures were categorized according to their distribution, as shown in Table 1. For depression, we utilized the clinically relevant cut-point of 16 as an additional separate category. Exposures were also modeled continuously per interquartile range. Additionally, in exploratory analyses where we stratified according to self-reported health, dichotomous categorizations were used.

Table 1.

Psychosocial exposures: The Multi-Ethnic Study of Atherosclerosis^*

Construct	Measure	Range	Calculated	Modeled
Anger	Spielberger Trait Scale22	10-40	Summation, all responses on 4 point scale	Quartiles
Anxiety	Spielberger Trait Scale22	10-40	Summation, all responses on 4 point scale	Quartiles
Chronic Stress	Chronic Burden Scale23	0-5	Summation of domains with an ongoing difficulty	Tertiles
Depressive Symptoms	Center for Epidemiological Studies Depression Scale (CES-D)24	0-60	Summation, responses on 0-3 scale	5 categories: quartiles with highest category split at >16**
Hostility	Eight items from Cook-Medley Hostility Scale25	0-8	Summation	Quartiles

^*Measured at Exam 1 (2000-2002) except for hostility, which was measured at Exam 2 (2002-2004)

^**A score of 16 or higher on the CES-D is typically indicative of clinically significant symptomatology

Outcome

New cases of HF were adjudicated by physician reviewers according to standard MESA procedures, as has been described elsewhere^{21, 27}. Three criteria were used to determine HF events: 1) Physician-diagnosed HF and patient medical treatment, 2) Pulmonary edema or congestion as indicated by chest x-ray, 3) Dilated ventricle, poor left ventricular function, or evidence of left ventricular diastolic dysfunction, as indicated by echocardiography, radionuclide ventriculogram/multigated acquisition, or other contrast ventriculography.

Covariates

For most analyses all covariates came from Exam 1. The exception was analyses where hostility, which was measured at Exam 2, was the exposure of interest. When available, covariates for analyses of hostility came from Exam 2.Sociodemographic characteristics include self-reported age, race/ethnicity (White, Black/African-American, Hispanic, and Chinese American), sex, and field center. Behavioral factors included smoking status (current smoker, former smoker, never smoker) and moderate-vigorous physical activity in met-min/wk. HF risk factors included body mass index, systolic and diastolic blood pressure, HDL, LDL, triglycerides, diabetes status, C-reactive protein, fibrinogen, and albumin.

Analysis

Descriptive statistics were calculated for exposure variables at baseline. Cox proportional hazards models were used to model time to incident HF events. Person-time was calculated by using the time from the baseline examination until a HF event, death, loss to follow-up, or December 31, 2011. Exam 1 was baseline for chronic stress, depressive symptoms, anger, and anxiety, and Exam 2, for hostility. Exposures were categorized as indicated in Table 1, and also modeled continuously per interquartile range. The proportional hazards assumption was checked using interactions with time and graphs of the residuals, and no violations were detected.

We explored a series of models. The first model controlled for age, race/ethnicity, sex, and field center. Model 2 added smoking and physical activity. Model 3 additionally controlled for C-reactive protein, fibrinogen, and albumin, and model 4 further controlled for BMI, diabetes, systolic and diastolic blood pressure, HDL, LDL, and triglycerides. C-reactive protein and triglycerides were log-transformed to account for skewness.

Multiplicative interactions by sex and self-reported health [categorized dichotomously as good (excellent/very good/good) or poor (fair/poor)] were tested by including cross-product terms in the models. Stratified analyses were performed for all significant or suggestive interactions. For stratified analyses, psychosocial exposures were dichotomized since power was low due to lower numbers of events occurring within strata. Anger, anxiety, and hostility were all dichotomized at their respective medians. For chronic stress we combined the lower two tertiles, given the distribution of the data. Depressive symptoms were divided at the clinically relevant cut-point of 16. Exposures were also modeled continuously per interquartile range. Models 1 and 2 were run again for each exposure in the stratified analysis.

Results

There were 242 new cases of HF through a total of 63,584 years of follow-up. On average, participants were followed 9.3 years (SD=2.5), with a maximum follow-up time of 11.5 years. The incidence rate was 3.8 HF cases per 1000 person-years (95%CI 3.4-4.3) Of those who did not develop HF, 638 were censored due to death by other causes, 367 were lost to follow-up, and 5,535 were censored at the end of follow-up. The overall mean age at Exam 1 was 62.2 years (SD=10.2), 52.9% of the sample was female, and 10.0% reported fair/poor health. At Exam 1, the mean scores were 14.8 (SD=3.7) for anger, 15.9 (SD=4.5) for anxiety, 1.2 (SD=1.2) for chronic stress, and 7.6 (SD=7.6) for depressive symptoms. At Exam 2, the mean score was 2.7 (SD=2.3) for hostility.

Characteristics of MESA participants by HF status are provided in Table 2. On average, those who developed HF during follow-up were, at baseline, older, had fewer Met-minutes per week of physical activity, had higher BMIs, and had higher systolic blood pressure compared to those who did not develop HF during follow-up. Those who developed HF during follow-up were also more likely at baseline to have less than a high school education, be current or former smokers, have diabetes, and be on hypertension medication compared to those did not develop HF.

Table 2.

Baseline participant characteristics stratified by those who developed heart failure during follow-up versus those who did not: The Multi-Ethnic Study of Atherosclerosis (2000-2012)

	HF Cases	No HF	p-value†
N total	242	6567
Demographics
Age, mean years ± SD	68.7 ± 8.8	61.9 ± 10.2	<0.001
Race/Ethnicity, n (%) *			0.76
White/Caucasian	97 (40.1)	2522 (38.4)
Chinese American	14 (5.8)	789 (12.0)
Black/African American	79 (32.6)	1813 (27.6)
Hispanic	52 (21.5)	1443 (22.0)
Education, n (%) *			0.05
< High school	55 (28.5)	1170 (21.8)
High school	36 (18.7)	1072 (20.0)
> High school	102 (52.8)	3115 (58.2)
Psychosocial Exposures, mean ± SD
Anger	14.2 ± 3.5	14.8 ± 3.7	0.030
Anxiety	15.1 ± 4.2	15.9 ± 4.5	0.007
Chronic Stress	1.2 ± 1.2	1.2 ± 1.2	0.83
Depressive Symptoms	7.3 ± 6.9	7.6 ± 7.6	0.50
Hostility	3.0 ± 2.3	2.7 ± 2.3	0.08
Self-reported health	3.3 ± 0.9	3.6 ± 0.9	<0.001
Behavioral Characteristics
Moderate-Vigorous Physical Activity, Met-min/wk **	4804.4 ±5116.1	5785.9 ± 5922.2	0.001
Smoking n (%) *			0.013
Current	37 (15.4)	850 (13.0)
Former	104 (43.2)	2380 (36.4)
Never	100 (41.5)	3316 (50.7)
Physiologic Characteristics
BMI, kg/m2 ± SD	29.9 ± 6.0	28.3 ± 5.4	<0.0001
Prevalent diabetes, n (%) *	75 (31.0)	781 (11.9)	<0.0001
Systolic BP, mmHg ± SD	139.0 ± 23.3	126.1 ± 21.3	<0.0001
Hypertension medication use, n (%) *	148 (61.2)	2385 (36.3)	<0.0001
Total cholesterol, mg/dL ± SD	189.7 ± 35.4	194.3 ± 35.7	0.048
HDL cholesterol, mg/dL ± SD	48.6 ± 13.9	51.1 ± 14.9	0.012
LDL cholesterol, mg/dL ± SD	114.71 ±32.0	117.3 ± 31.4	0.126
Triglycerides, mg/dL ±SD **	140.9 ± 118.1	131.1 ± 87.4	0.242
C-Reactive Protein, mg/L ±SD **	5.0 ± 6.5	3.7 ± 5.9	<0.0001
Albumin, mg/dL ± SD **	15.3 ± 71.3	2.4 ± 12.6	<0.0001
Fibrinogen, mg/dL ± SD	370.5 ± 78.3	345.9 ± 73.6	<0.0001

^†T-tests were used for linear variables that were normally distributed, for categorical variables (designated with an asterisk) chi-square tests were used, and for non-normally distributed linear variables (designated with two asterisks) the Wilcoxon-Mann-Whitney test was used.

Tables 3 and 4 shows hazard ratios and 95% confidence intervals for each psychosocial factor in relation to risk of incident HF. In model 1, which controlled for age, sex, race, and study site, we found no significant association between any of the psychosocial factors and risk of incident HF. Compared to participants in the lowest level, hazard ratios for those categorized in the highest level of anger [HR=1.14 (95%CI: 0.81-1.60)], anxiety [HR=0.74 (95%CI: 0.51-1.07), chronic stress [HR=1.25 (95%CI: 0.90-1.72), depressive symptoms [HR=1.19 (95%CI: 0.76-1.85), and hostility [HR=0.95 (95%CI: 0.62-1.42) revealed no association with incident HF. When modeled continuously, an interquartile range increase in chronic stress was associated with a 27% increase in risk of incident HF. The addition of smoking status and physical activity in the model slightly attenuated the estimates. Results were similar in models adjusted for inflammatory markers (model 3) and additionally for traditional cardiovascular disease risk factors (model 4) (data not shown).

Table 3.

Adjusted hazard ratios (95% confidence intervals) for psychosocial factors and risk of incident heart failure: The Multi-Ethnic Study of Atherosclerosis (2000-2012)

Anger Quartile (Range)	1 (10-12)	2 (13-14)	3 (15-16)	4 (17-40)	p-trend categories		Continuous per IQR	p-trend continuous
Number of incident HF events	88	55	41	57			241
Total number of participants	2007	1703	1313	1754			6777
Model 1	1.00 (Referent)	0.90 (0.64-1.26)	0.94 (0.65-1.37)	1.14 (0.81-1.60)	0.80		1.02 (0.85-1.23)	0.80
Model 2	1.00 (Referent)	0.89 (0.64-1.26)	0.92 (0.63-1.35)	1.11 (0.78-1.56)	0.97		1.00 (0.83-1.20)	0.99
Anxiety quartile (Range)	1 (10-12)	2 (>12-15)	3 (>15-18.75)	4 (>18.75-37)	p-trend
Number of incident HF events	84	57	53	47			241
Total number of participants	1766	1803	1406	1795			6770
Model 1	1.00 (Referent)	0.78 (0.56-1.10)	1.04 (0.73-1.48)	0.74 (0.51-1.07)	0.49		0.93 (0.75-1.15)	0.49
Model 2	1.00 (Referent)	0.78 (0.55-1.09)	1.04 (0.73-1.47)	0.72 (0.50-1.05)	0.42		0.91 (0.74-1.13)	0.41
Chronic Stress Tertile (Range)	1 (0)	2 (1)	3 (2-5)	p-trend
Number of incident HF events	80	83	76				239
Total number of participants	2308	2108	2332				6748
Model 1	1.00 (Referent)	1.12 (0.82-1.53)	1.25 (0.90-1.72)	0.04			1.27 (1.01-1.59)	0.04
Model 2	1.00 (Referent)	1.13 (0.83-1.53)	1.23 (0.89-1.70)	0.05			1.25 (1.00-1.57)	0.05
Depressive Symptoms Quintiles (Range)	1 (0-2)	2 (3-5)	3 (6-10)	4 (11-15)	5 (16-60)	p-trend
Number of incident HF events	72	49	62	27	30
Total number of participants	1829	1549	1707	816	872
Model 1	1.00 (Referent)	0.85 (0.59-1.23)	1.01 (0.71-1.42)	0.97 (0.62-1.52)	1.19 (0.76-1.85)	0.35	1.07 (0.93-1.24)	0.35
Model 2	1.00 (Referent)	0.90 (0.60-1.35)	0.97 (0.66-1.43)	1.06 (0.65-1.74)	1.18 (0.72-1.93)	0.43	1.06 (0.91-1.22)	0.46
Hostility Quartile (Range)	1 (0)	2 (1-2)	3 (3-4)	4 (5-8)	p-trend
Number of incident HF events	51	37	69	50			207
Total number of participants	1440	1741	1544	1437			6162
Model 1	1.00 (Referent)	0.60 (0.39-0.91)	1.30 (0.90-1.89)	0.95 (0.62-1.43)	0.10		1.17 (0.97-1.41)	0.10
Model 2	1.00 (Referent)	0.58 (0.38-0.89)	1.28 (0.88-1.86)	0.93 (0.61-1.41)	0.13		1.16 (0.96-1.40)	0.13

Model 1: adjusted for age, sex, race, and field center

Model 2: adjusted for model 1 plus smoking and physical activity

Higher scores on the psychosocial measures represent more severe symptoms

Interquartile range (Q1, Q3)

Anger: (12, 17)

Anxiety: (12, 19)

Chronic stress: (0, 2)

Depressive symptoms: (2, 10)

Hostility: (1, 4)

Table 4.

Adjusted hazard ratios (95% confidence intervals) for psychosocial factors and risk of incident heart failure stratified by self-reported health: The Multi-Ethnic Study of Atherosclerosis (2000-2012)

	Good self-reported health			Poor self-reported health
Anger (Range)	Low (10-14)	High (15-40)	Per IQR	Low (10-14)	High (15-40)	Per IQR
Number of incident HF events	121	86	207	22	12	34
Total number of participants	3358	2742	6100	352	325	677
Model 1	1.00 (Referent)	1.16 (0.88-1.54)	1.09 (0.90-1.33)	1.00 (Referent)	0.78 (0.37-1.65)	0.73 (0.44-1.21)
Model 2	1.00 (Referent)	1.14 (0.86-1.52)	1.07 (0.88-1.31)	1.00 (Referent)	0.72 (0.34-1.53)	0.69 (0.42-1.14)
Anxiety (Range)	Low (10-15)	High (16-37)	Per IQR	Low (10-15)	High (16-37)	Per IQR
Number of incident HF events	129	78	207	12	22	34
Total number of participants	3286	2806	6092	283	395	678
Model 1	1.00 (Referent)	0.86 (0.65-1.15)	0.83 (0.65-1.05)	1.00 (Referent)	2.11 (1.00-4.47)	1.39 (0.86-2.25)
Model 2	1.00 (Referent)	0.85 (0.64-1.14)	0.82 (0.64-1.04)	1.00 (Referent)	2.03 (0.97-4.27)	1.31 (0.81-2.13)
Chronic Stress (Range)	Low (0-1)	High (2-5)	Per IQR	Low (0-1)	High (2-5)	Per IQR
Number of incident HF events	150	55	205	13	21	34
Total number of participants	4085	1987	6072	331	345	676
Model 1	1.00 (Referent)	1.01 (0.74-1.38)	1.10 (0.85-1.43)	1.00 (Referent)	2.25 (1.08-4.67)	2.12 (1.25-3.60)
Model 2	1.00 (Referent)	1.00 (0.73-1.37)	1.09 (0.85-1.42)	1.00 (Referent)	2.15 (1.04-4.47)	2.04 (1.21-3.44)
Depressive Symptoms (Range)	Low (0-15)	High (16-60)	Per IQR	Low (0-15)	High (16-60)	Per IQR
Number of incident HF events	187	19	206	23	11	34
Total number of participants	5411	685	6096	490	187	677
Model 1	1.00 (Referent)	1.01 (0.62-1.63)	1.02 (0.86-1.20)	1.00 (Referent)	2.15 (0.98-4.68)	1.21 (0.90-1.63)
Model 2	1.00 (Referent)	0.98 (0.61-1.59)	1.00 (0.85-1.19)	1.00 (Referent)	1.98 (0.90-4.34)	1.18 (0.88-1.58)
Hostility (Range)	Low (0-2)	High (3-8)	Per IQR	Low (0-2)	High (3-8)	Per IQR
Number of incident HF events	76	102	178	12	17	29
Total number of participants	3006	2589	5595	175	392	538
Model 1	1.00 (Referent)	1.63 (1.20-2.23)	1.25 (1.02-1.53)	1.00 (Referent)	0.60 (0.27-1.33)	0.73 (0.45-1.19)
Model 2	1.00 (Referent)	1.65 (1.20-2.25)	1.25 (1.02-1.53)	1.00 (Referent)	0.59 (0.26-1.31)	0.71 (0.43-1.17)

Model 1: adjusted for age, sex, race, and field center

Model 2: adjusted for model 1 plus smoking and physical activity

Higher scores on the psychosocial measures represent more severe symptoms

Interquartile range (Q1, Q3)

Anger: (12, 17)

Anxiety: (12, 19)

Chronic stress: (0, 2)

Depressive symptoms: (2, 10)

Hostility: (1, 4)

Interactions between sex and each psychosocial factor were tested, but were only significant for hostility (p=0.03).In sex-specific analyses, compared to the lowest quartile, the hazard ratio for incident HF in the highest quartile was 0.71 (0.40-1.26) among men and 1.39 (0.76-2.54) among women. Interactions between race/ethnicity and each psychosocial factor were also tested, but none was statistically significant (all p > 0.20).

Interactions were also tested between psychosocial factors (modeled dichotomously) and self-reported health (modeled dichotomously). Only hostility was significant at the .05 level, though anxiety and chronic stress were significant at the .10 level. Table 4 presents results stratified by dichotomous self-reported health categories. For anxiety, chronic stress, and depressive symptoms, those in the highest versus lowest categorization, and also self-reporting poor health at baseline, were at 2-fold greater risk of incident HF. For those with good self-reported health at baseline, there was no evidence of an association between these psychosocial characteristics and HF risk. The opposite was true for hostility, where HRs for the highest versus lowest categorization were larger among those with good self-reported health. For anger, associations were similar regardless of self-reported health status. Estimates were modestly attenuated when smoking status and physical activity were entered into the model.

Discussion

There was no association between several psychosocial factors (i.e. anger, anxiety, chronic stress, depressive symptoms, hostility) and risk of incident HF in this multiethnic population-based study. When modeled continuously, more chronic stress was associated with a slight increase in risk of incident HF. Though not statistically significant, there were some suggestions that the association between certain psychosocial factors and incident HF may differ by baseline self-reported health status. However, these results must be interpreted cautiously because of low power.

Relatively few previous studies have examined associations between psychosocial factors and incident HF, and most have only examined depression or depressive symptoms. Williams et al. found that high levels of depressive symptoms, as defined by CES-D scores greater than or equal to 21, were associated with greater risk of incident HF in an elderly population (mean age 74) from Connecticut, but this effect was only found in women, not men.⁷ Abramson et al. also found an independent association between depression and incident HF in an older sample of people (mean age 72) with isolated systolic hypertension.⁸ More recently, a study examined anxiety and depression in a very large VA population (mean age 63) using ICD-9 codes and found a small but statistically significant relationship between these factors and incident HF.¹⁰ Another recent European study found a positive association between depression and incident HF, but no association between anxiety and incident HF.¹¹ Although prior research has frequently reported associations between adverse levels of psychosocial factors and the development of stroke and coronary heart disease,^{28, 29} it is worthwhile to note that the onset of stroke and coronary heart disease is typically acute, while HF is a chronic condition that develops gradually. It is possible that pathways between psychosocial exposures and acute versus chronic cardiovascular outcomes may be different.

Although our study did not find significant overall results between psychosocial factors and HF, among those with poor self-reported health there was some evidence suggesting that those scoring higher on the anxiety, chronic stress, and depressive symptoms scales were at greater risk of incident HF. The idea that psychosocial factors may play a greater role in HF development among those with poor self-reported health – who likely have prevalent comorbidities – is supported by the existing literature. Like depression,⁵ poor self-rated health has been associated with increased risk of emergency department visits, hospitalization, and mortality among HF patients.^{30, 31} As noted above, in prior publications associations between depression, anxiety and incident HF were observed in study populations who were older^{7, 8} and hypertensive⁸, and therefore may also have been more prone to comorbidities and poor self-reported health. Additionally, those with lower self-reported health and adverse psychosocial profiles may be less likely to take medications as directed.

One major limitation of this study is the relatively small number of participants who developed HF, especially when we stratified on self-reported health. Interpretation of these results is difficult because it is unclear whether there is no overall association between psychosocial factors and incident HF, or there was not enough power to detect an association. Relatedly, given the limited number of HF cases our psychosocial factor categories were somewhat broad. Using finer categories, and therefore conducting more extreme comparisons, may have yielded different results. For instance, we employed the commonly used cut-point of 16 to define depressive symptoms. Prior work, however, has suggested that using this cut-point results in many individuals being falsely classified as having depressive symptoms, and that a higher cut-point should be employed.^{32, 33} Studies with a larger number of cases may be able to more precisely determine the magnitude of the relationship between psychosocial factors and risk of incident HF. Additionally, there is the potential for error in the measurement of the exposures and the outcome. For the exposures, we would expect this error to be unrelated to the outcome, because there were no prevalent HF cases at baseline and disease status is ascertained in the future. Because this misclassification is non-differential, estimates obtained from this study would likely be biased towards the null. For the outcome, because exposure status is determined before the outcome occurs, any error in the measurement of the outcome is likely to be unrelated to exposure and is expected to bias the results toward the null. This study also has several noteworthy strengths including a multiethnic representative population-based sample, ascertainment of multiple psychosocial factors, and adjudicated HF outcomes.

Overall, this study found no strong statistically significant relationships between psychosocial factors and incident HF. However, adverse levels of psychosocial factors may play a role in, or be an indicator of, HF development among those who perceive themselves as having poor health. Future research with greater power is necessary to reach more definitive conclusions.

Clinical Perspective.

Previous research has examined how psychosocial factors are related to coronary heart disease, but their association with heart failure is understudied. This paper investigated whether anger, anxiety, chronic burden, depression, and hostility are related to the development of the new heart failure cases. In the overall sample, we found no relationship between psychosocial factors and heart failure. However, among those reporting fair or poor health, several psychosocial factors (i.e. anxiety, chronic burden, and depression) may be related to heart failure risk. From a clinical perspective, although there is evidence of an association between psychosocial factors and other cardiovascular outcomes, the influence of psychosocial factors on incident heart failure appears to be mediated by baseline health status, with stronger influence in those with poorer baseline health.