Article Text

Download PDFPDF

Original article
Psychological stress and short-term hospitalisations or death in patients with heart failure
  1. Romano Endrighi1,2,
  2. Andrew J Waters1,
  3. Stephen S Gottlieb3,
  4. Kristie M Harris4,
  5. Andrew J Wawrzyniak5,
  6. Nadine S Bekkouche1,
  7. Yisheng Li6,
  8. Willem J Kop7,
  9. David S Krantz1
  1. 1Department of Medical and Clinical Psychology, Uniformed Services University of the Health Sciences, Bethesda, Maryland, USA
  2. 2Division of Behavioral Science Research, Department of Health Policy and Health Services Research, Boston University, Boston, Massachusetts, USA
  3. 3University of Maryland School of Medicine, Baltimore, Maryland, USA
  4. 4Cardiopulmonary Behavioral Medicine Laboratory, The Ohio State University, Columbus, Ohio, USA
  5. 5Department of Psychiatry and Behavioral Sciences, University of Miami, Miami, Florida, USA
  6. 6Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, Texas, USA
  7. 7Department of Medical and Clinical Psychology, Tilburg University, Tilburg, The Netherlands
  1. Correspondence to Professor David S Krantz, Department of Medical and Clinical Psychology, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Rd., Bethesda MD 20814, USA; david.krantz{at}


Objective Standard predictors do not fully explain variations in the frequency and timing of heart failure (HF) adverse events (AEs). Psychological stress can trigger acute cardiovascular (CV) events, but it is not known whether stress can precipitate AEs in patients with HF. We investigated prospective associations of psychological stress with AEs in patients with HF.

Methods 144 patients with HF (77% male; 57.5±11.5, range 23–87 years, left ventricular ejection fraction ≤40%) were longitudinally evaluated for psychological stress (Perceived Stress Scale) and AEs (CV hospitalisations/death) at 2-week intervals for 3 months and at 9-month follow-up.

Results 42 patients (29.2%) had at least one CV hospitalisation and nine (6.3%) died. Patients reporting high average perceived stress across study measurements had a higher likelihood of AEs during the study period compared with those with lower stress (odds ratio=1.10, 95% confidence interval=1.04 to 1.17). In contrast to average levels, increases in stress did not predict AEs (p=0.96). Perceived stress was elevated after a CV hospitalisation (B=2.70, standard error (SE)=0.93, p=0.004) suggesting that CV hospitalisations increase stress. Subsequent analysis indicated that 24 of 38 (63%) patients showed a stress increase following hospitalisation. However, a prospective association between stress and AEs was present when accounting for prior hospitalisations (B=2.43, SE=1.23, p=0.05).

Conclusions Sustained levels of perceived stress are associated with increased risk of AEs, and increased distress following hospitalisation occurs in many, but not all, patients with HF. Patients with chronically high stress may be an important target group for HF interventions aimed at reducing hospitalisations.

View Full Text

Statistics from


Heart failure (HF) is characterised by high mortality, frequent hospitalisations and impaired quality of life. Rehospitalisation rates in the US are estimated to be 20%–35% within 30 days after discharge.1 ,2 Despite significant advances in treatment, hospitalisations cause substantial physical and psychological burden for patients with HF.3 ,4 Interventions for reducing HF exacerbations, adverse events (AEs) and hospitalisations have focused on preventing and treating known biomedical and behavioural precipitating factors,5 ,6 including haemodynamic, renal and pulmonary dysfunction, myocardial ischaemia and infarction and poor medication adherence.7 ,8 Chronic psychosocial risk factors such as depression and low social support also predict HF clinical outcomes in prospective studies.9 ,10 However, the effects of increases in psychological precipitating factors, such as perceived stress11 ,12 on short-term hospitalisations and AEs in HF are not well understood.

Psychological stress can adversely affect important predictors of HF-related AEs, including haemodynamic and neuroendocrine function,10 medication and diet adherence and self-care.13 ,14 Most research on stress and HF is limited to retrospective studies,15 or to laboratory studies that have shown that mental stress can induce myocardial ischaemia in patients with HF and coronary artery disease.16 ,17

Studies examining whether short-term increases in psychological stress are associated with adverse outcomes such as hospitalisation in patients with HF are rare.4 ,15 ,18 Therefore, it is not known whether chronic stress levels and/or shorter-term increases in psychological stress can precipitate AEs in patients with HF. These questions would have implications for choosing appropriate preventive targets.

In addition, a posthospitalisation syndrome may result from the stressful experience of a hospitalisation,4 suggesting that the physiological and psychological consequences of being hospitalised may heighten risk of subsequent hospitalisations and AEs.14 Evidence for the prevalence of stress as part of a posthospital syndrome is limited,18 and it is not known whether short-term increases in psychological stress predict subsequent hospitalisations or whether chronically stressed patients with HF are more susceptible to AEs.

The purpose of the present study is to determine whether psychological stress is a predictor of cardiovascular (CV) hospitalisations or death in patients with HF, and to assess the frequency and magnitude of posthospitalisation effects on patient distress. To test these relationships, we used a prospective study design involving multiple assessments of stress and AEs (CV hospitalisation or death) over time.


In a prospective clinical cohort study (the BETRHEART study), 144 patients (mean age 57.5±11.5, range 23–87 years, 77% male) with a primary diagnosis of systolic HF were recruited at the University of Maryland Medical Center (UMMC). Inclusion criteria were: left ventricular ejection fraction (LVEF) ≤40% assessed by echocardiography, age ≥18 years and symptomatic HF (New York Heart Association (NYHA) class II–IV) for ≥3 months. Exclusion criteria were: significant valve disorder as primary diagnosis or thyroid dysfunction as primary aetiologies, myocarditis or alcohol abuse in last 6 months, current left ventricular assist device, active cancer treatment, nursing home residence and significant cognitive impairment. The study was approved by Institutional Review Boards of the UMMC and the Uniformed Services University. All participants provided written informed consent.


The study design is illustrated in figure 1. All patients attended a baseline clinic assessment at the HF clinic at UMMC, where detailed clinical assessments were carried out by the clinical care team. In addition, sociodemographic, psychosocial and behavioural variables were measured via clinical interviews and questionnaires.

Figure 1

Mean (SD) and sample size of patients with PSS data at each time point, and number (%) of patients experiencing at least one adverse event before the subsequent assessment. Values are means (SD) or N (%). Adverse events=number of patients experiencing one or more cardiovascular hospitalisations or death from any cause before the subsequent assessment. A total of 144 patients had complete data on the PSS at baseline, and 140 patients had complete data on adverse events. PSS, Perceived Stress Scale; SD, standard deviation.

Following baseline assessment, all patients were scheduled for biweekly telephone interviews between baseline and a 3-month follow-up clinic visit (intervening weeks 2, 4, 6, 8 and 10). During interviews, patients were administered the Perceived Stress Scale (PSS)19 to quantify psychological stress in the previous 2 weeks. Hospitalisation events occurring in the previous 2 weeks were determined from patient reports and subsequently verified through systematic review of medical records. Deaths were recorded from household member reports, and verified in medical records.

At 3 months following baseline (week 12), all patients attended a clinic assessment, during which patients repeated the same clinical assessments as at baseline, and hospitalisation events recorded and subsequently verified.

Six months after the 3-month assessment (9-month follow-up), patients were again interviewed via telephone for AEs that occurred since last contact, and hospital records verified. If the patient had died, spouse or next of kin was interviewed regarding date and cause of death.

Study measures

Perceived stress

The 10-item PSS19 measures generalised perception of stress over the previous 2 weeks. This standardised instrument was developed to assess the extent to which an individual perceives events or situations as stressful or out of control (eg, during the last 2 weeks: ‘how often did you feel that you were unable to control the important things in your life’? and ‘how often have you felt confident about your ability to handle your personal problems’?). Reliability and validity of the PSS is well established20 and it has been used widely to quantify stress in epidemiological and clinical research.12 ,21 PSS scores range from 0 to 40 with higher scores indicating greater perceived stress. Scale internal reliability (Cronbach's α) ranges from 0.85 to 0.92.

Adverse events

AEs were defined as hospitalisations for CV causes or death. CV causes included: HF-specific (diagnosis of fluid overload or pump failure), angina, myocardial infarction, ischaemia or new-onset arrhythmia. Hospitalisation causes were expert-adjudicated by a cardiologist on the research team and confirmed via medical records. Death from any cause was used as an endpoint. An ‘adverse event’ variable was computed by determining, at each assessment, the presence of CV hospitalisations (or death) events occurring between that assessment and the subsequent assessment. For each assessment, each patient was assigned a ‘0’ (for no AE before next assessments) or ‘1’ (for ≥1 AEs before the next assessment).

Data analysis

Data are presented as means (M) and standard deviations, or N and percentages, as appropriate. Linear mixed models (LMMs) analyses using SAS PROC MIXED (continuous outcomes) or PROC GLIMMIX (binary outcomes) were conducted. LMMs account for dependence between observations and allow for different numbers of participant observations.22 A random (subject-specific) intercept was employed. Parameter estimates are reported as B (standard error, SE) or as odds ratio (OR) (95% confidence interval (CI)); t values with associated p values test the null hypotheses that the B value is equal to zero, or that the OR is equal to one, in the population. Time was entered as a continuous variable in all models.

In LMMs, stress was coded as ‘Mean PSS’ (ie, PSS scores aggregated over all available assessments for each subject), and ‘Deviation PSS’ (the difference between the PSS score at each assessment and each subject's Mean PSS score).23 A significant coefficient for Mean PSS reflects a patient characteristic or ‘between subjects’ effect that examines ‘who’ is generally at risk of AEs. A significant coefficient for Deviation PSS reflects an effect of short-term changes in stress, that is, a ‘within-subject’ effect that captures ‘when’ an individual is at risk of an AE.

To examine whether hospitalisation results in increased posthospitalisation stress, LMM tested whether PSS scores were elevated at assessments occurring immediately following, compared with observations not immediately following, a CV hospitalisation. We further examined whether stress was associated with AEs independent of the influence of prior recent hospitalisations. In this analysis, PSS scores of the patient subgroup without any AEs in the study (‘No event patients’) were compared with PSS scores of patients who had AEs, using only assessments before their initial study event (‘Event patients—before first event’). Associations between baseline only PSS scores (Baseline PSS) and subsequent AEs are also presented to determine whether serial stress assessments improve AE prediction compared with a baseline only assessment. Cox proportional hazards survival analysis was also used to examine the relationship between Baseline PSS and time to first AE; parameter estimates are presented as hazard ratios (HR) and 95% CIs.


Due to the relatively small sample size, the covariates were limited to demographic variables, indices of disease severity, and a limited number of risk factors chosen because of their possible association with stress or the study outcome. The following covariates, measured at baseline, and each with complete data, were used: age (years), sex, race, Body Mass Index (in kg/m2), smoking, NYHA classification (II; III and IV), LVEF (%) and hypertension history. Two additional variables had missing data and were not related to cardiac events in this sample: creatinine levels (mg/dL) (four missing values) and household income (in US$: <15 k, ≥15 to <30 k, ≥30 to <70 k and ≥70 k) (one missing value). In supplementary analyses not presented, we used multiple imputation of missing values and also included these variables as covariates. Results from these analyses did not differ from the analyses presented.


Sample demographic and clinical characteristics are presented in table 1, as are baseline comparisons of patients who did and did not have AEs during the study. The study design and descriptive statistics for predictor and outcome variables between baseline and 9 months are presented in figure 1.

Table 1

Baseline sociodemographic and clinical characteristics of patients who did and did not experience adverse events during the study

During the study, nine patients died and 42 (29.2%) experienced at least one CV hospitalisation; four of these subsequently died and five (3.5%) died without a previous hospitalisation. Four of these subsequently died, and five (3.5%) died without a previous hospitalisation. There were a total of 62 events in the 51 patients with at least one event.

Sustained high stress, short-term changes in stress and AEs

The odds of experiencing an AE were greater in patients with higher (vs lower) Mean PSS scores (unadjusted OR=1.11, 95% CI 1.04 to 1.18 per unit PSS; adjusted OR=1.10, 95% CI 1.04 to 1.17). This result means that individuals with higher Mean PSS scores have higher odds of AEs during the study period. In contrast to Mean PSS relationships, there was no association between Deviation PSS and AEs (unadjusted OR=1.00, 95% CI 0.94 to 1.07, p=0.95; adjusted OR=1.00, 95% CI 0.94 to 1.06, p=0.96), indicating that short-term increases in psychological stress were not associated with increased risk of subsequent AEs. In addition, the Baseline only PSS score was not associated with AEs (unadjusted B=0.04, SE=0.03, p=0.11; adjusted B=0.04, SE=0.03, p=0.14). Using Cox proportional hazards regression, Baseline PSS was similarly not associated with time to AE in either an unadjusted (HR=1.029, 95% CI 0.993 to 1.065, p=0.11) or adjusted analysis (HR=1.034, 95% CI 0.996 to 1.073, p=0.10).

The effect of Mean PSS was not attributable to depression9 ,13 since the effect remained significant when controlling for baseline depression scores on the Beck Depression Inventory-II (BDI-II)24 (adjusted OR=1.14, 95% CI 1.04 to 1.25). Baseline BDI-II was also not associated with AEs (p=0.29).

Psychological stress following CV hospitalisations

To determine whether experiencing a CV hospitalisation increased stress (posthospitalisation stress effect), we conducted an analysis using only those patients who experienced at least one CV hospitalisation during the study. PSS scores measured at assessments occurring immediately after one (or more) CV hospitalisations (‘After adverse event’) were compared with PSS scores at assessments not preceded by a hospitalisation (‘Control assessments’). PSS scores after AEs were significantly higher compared with control assessments (M=15.47±8.14 vs M=12.34±8.12; unadjusted B=2.72, SE=0.94, p=0.004; adjusted B=2.70, SE=0.93, p=0.004) (figure 2). This indicates that stress was significantly higher following a CV hospitalisation (when controlling for time and other covariates). A closer examination of these results indicated that this effect was not universal with 24 of 38 patients (63.2%) who were hospitalised reporting a ‘positive’ posthospitalisation effect (higher stress at assessments following a CV hospitalisation vs control assessments).

Figure 2

Association of prior hospitalisations for heart failure-specific or cardiovascular (CV) causes with subsequent stress levels. Box and whisker plot of Perceived Stress Scale (PSS) scores from patients who experienced at least one hospitalisation. Scores are aggregated over all assessments that followed baseline. ‘After adverse event’=PSS scores measured at assessments that occurred after a hospitalisation for CV causes (n=53). ‘Control assessments’=PSS scores measured at assessments that were not preceded by a hospitalisation for CV causes (n=231). Boxes and internal lines represent 25th and 75th percentiles and median, and whiskers show range and outliers.

Prospective association of psychological stress with AEs in patients without prior hospitalisations

To determine whether the association between higher mean perceived stress (Mean PSS) and AEs was prospective and not a result of recent prior hospitalisation, analyses were conducted using only PSS scores measured prior to a patient having any AE in the study. Among patients with at least one AE, we selected all their PSS scores before their first AE in the study (‘Event patients—before first event’). These scores were then compared with PSS scores of those who had no AEs in the study (No event patients) (figure 3). PSS scores of ‘Event patients—before first event’ were significantly higher than scores of no event patients (M=12.95±8.42 vs M=10.65±7.73: unadjusted B=2.51, SE=1.21, p=0.04; adjusted B=2.43, SE=1.23, p=0.05). Figure 3 also indicates that PSS scores declined over time (B=−0.47, SE=0.11, p=0.0001).

Figure 3

Prospective association of perceived stress with cardiovascular hospitalisation or death in the study. Data are Perceived Stress Scale (PSS) scores aggregated over all assessment time points that occurred prior to a patient's first adverse event in the study. Each assessment point between baseline and 3 months only includes patients who have not yet had an event at that assessment point. PSS scores following a hospitalisation at any point in the study are excluded from this analysis. ‘Event patients—before first event’=PSS scores of 47 patients who experienced subsequent adverse events during the study (n=238 assessments). ‘No event patients’=PSS scores of 93 patients who had no adverse events during the study (n=615 assessments).

Additional evidence indicated that prospective results for higher stress were unlikely to be attributable to hospitalisations that occurred prior to study enrolment. Table 1 shows that ‘Adverse event patients’ were not more likely to have been hospitalised within the 30 days before study enrolment compared with ‘No adverse event patients’, and PSS scores remained higher in AE patients after adjusting for the presence of a recent (30 days) CV hospitalisation (adjusted B=2.52, SE=1.21, p=0.04).


This study demonstrates that sustained high levels of perceived stress are associated with increased risk of cardiac hospitalisation or death in patients with HF. In contrast, increases in PSS measured on a biweekly basis were not associated with exacerbations in the weeks immediately following each assessment. Therefore, measurement of psychological stress predicts which patients are at risk of AEs (who), but not the timing or when clinical events will occur. Identification of individuals with sustained high stress levels may help identify those patients with HF at risk of subsequent AEs. Other acute biomedical or behavioural factors (eg, increases in blood pressure and fluid overload, comorbidities and poor medication adherence) may be more important for predicting the timing of acute exacerbations.7 ,8 ,25

We also observed a posthospitalisation stress effect, with stress levels significantly elevated at assessments in the 2 weeks following a CV hospitalisation. This is consistent with a posthospitalisation syndrome,4 indicating that the experience of hospitalisation adversely affects risk factors and heightens likelihood for subsequent hospitalisation.18 History of HF hospitalisation is a known predictor of subsequent hospitalisations and adverse outcomes,14 and the present findings suggest that this may be partially explained by the increase in psychological stress following hospitalisation. Patients with HF with elevated perceived stress during emergency room admission are at great risk of 30 days rehospitalisation,18 and individuals with high stress after hospitalisation may therefore be at increased risk of subsequent AEs. However, prior studies have not been able to determine whether increased stress reflects a chronic patient characteristic, or is associated with a specific hospitalisation experience.4 In the present study, not all individuals showed an increase in stress levels posthospitalisation. More than 1/3 (37%) of patients did not show this effect, and this may reflect chronic patient characteristics or differences in treatment during hospitalisation. This issue has implications for whether preventive interventions should be targeted at modifying the hospitalisation experience or reducing patient distress that may be more chronic in nature.

Despite the increase in stress following hospitalisation, the predictive value of stress for future events in this study was not simply attributable to recent prior hospitalisations. A prospective relationship between perceived stress and subsequent AEs was present in patients without a recent prior hospitalisation, and associations between stress and subsequent hospitalisations remained significant after controlling for recent hospitalisations prior to study enrolment. Perceived stress may be both a risk factor for future AEs as well as a consequence of hospitalisation. HF is characterised by multiple exacerbations and hospitalisations over time, and a study design involving multiple longitudinal measurements allowed us to clarify acute and chronic associations between stress and AEs. This study design has been employed in research involving serial assessments of B-type natriuretic peptide,26 and may also prove useful for investigating other risk factors for HF exacerbations, such as acute blood pressure elevations or medication non-compliance.

Several factors may account for the observation that average stress scores decreased between baseline and 3 months. Repeated contact and telephone conversation with study personnel may have reduced patients' perceived stress, or attentive treatment because of study participation may have resulted in these patients' stress level to decrease over time. Alternatively, a selection bias might have caused high stress patients to miss follow-up assessments.

Pathways underlying observed associations of psychological stress with AEs may involve stress-induced increases in sympathetic, haemodynamic, proinflammatory and/or endocrine function that can contribute to worsening HF.9 ,10 ,18 ,27 Acute stress can also induce myocardial perfusion abnormalities and trigger malignant arrhythmias in patients with HF.17 ,24 ,27 Psychological stress has also important indirect effects on factors that increase risk of hospitalisation, such as reduced ability to follow complex medication and lifestyle regimens,11 ,25 impaired cognitive functioning and reduced physical activity.28

Study limitations

The present sample was majority African-American, male and had impaired LV function (systolic HF), which may limit the ability to generalise these findings to other populations. However, these data provide important information relevant to HF in minority populations and lower socioeconomic groups, who are known to develop markedly higher incident HF at early ages.29

Caution should also be exercised to not infer causality from observed associations of perceived stress and subsequent events. It is possible that factors such as an increase in symptoms or other physiological changes indicative of worsened clinical status may lead to heightened distress, or that those patients who are more distressed were generally sicker. This latter possibility is unlikely in this study since observed findings were present in multivariate analyses where HF risk factors were controlled. It is also possible that acute distress in the hours before the event may contribute to HF exacerbations, but this was not captured by our study design. If hospitalisations are triggered by a sudden worsening of the risk factor in the day or hours before the event, study designs that assess patients more frequently during daily life would be valuable.

Another limitation is the use of eight covariates in relation to the 62 events in the study. This may have led to a lack of sufficient power in some of the analyses (eg, the relationship of the baseline PSS measurement to subsequent events) to detect effects should they exist.

Clinical implications

Most interventions to reduce rehospitalisations include patient education and self-management strategies designed to increase adherence to complex medication regimens, making dietary changes and assuring medical follow-up.30 Stress is known to interfere with these behavioural changes. Behavioural interventions to reduce stress and enhance coping techniques may therefore enhance the efficacy of other multicomponent interventions in reducing AEs. These data also suggest that evaluation of patient distress may help identify those individuals who are at increased risk of HF hospitalisations and AEs who may benefit most from preventive interventions. Evidence for a posthospital syndrome4 ,18 in a subgroup of patients also points to the need to identify individuals who may be vulnerable to these effects and also address aspects of hospitalisation that may cause distress.4 Taken together, these findings suggest that psychological stress may be an important factor to address when evaluating patients' risk of short-term hospitalisations or death in HF.

Key messages

What is already known on this subject?

  • Psychosocial and behavioural variables (eg, compliance with medications or diet and family support) may be important in heart failure exacerbations. However, little is known about the influence of psychological stress on heart failure exacerbations, including rehospitalisations or death.

What might this study add?

  • This study determined that perceived stress in patients with heart failure is prospectively associated with short-term cardiac events in patients with systolic heart failure, and that there is a posthospitalisation increase in psychological distress. These findings suggest that stress may precipitate heart failure exacerbations, and that posthospitalisation stress may have further negative effects.

How might this impact on clinical practice?

  • Assessment of distress in patients with heart failure may help identify those at higher risk of short-term hospitalisation. Reducing distress and addressing stressful aspects of hospitalisation may help decrease future rehospitalisations and/or prevent heart failure exacerbations.


View Abstract


  • Contributors All authors listed have read and approved the submitted manuscript. In addition, each author has made a significant contribution to the manuscript and meet the authorship criteria of ICMJE: having made substantial contributions to the conception, design, acquisition, analysis or interpretation of data; drafting for important content; final approval of the version published and agreement to be accountable for the work. In addition, RE, AJW and DSK had full access to all data in the study and take responsibility for the integrity of the data and accuracy of data analysis. Detailed contributions of each of the authors are unchanged from what was provided with the initial submission of the manuscript.

  • Funding This work was supported by National Heart Lung and Blood Institute grant RO1-HL085730. The opinions and assertions expressed herein are those of the authors and are not to be construed as representing the views of Uniformed Services University or the US government.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval University of Maryland School of Medicine and Uniformed Services University of the Health Sciences.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.