Introduction

A close interrelationship exists between depression and cardiovascular diseases.¹ With a prevalence of 20 to >40%, comorbid depression is 2 to 5 times more common in heart failure than in the general population² and associated with adverse outcomes as reduced quality of life,^2,3 increased all-cause mortality,^2,4 frequency of rehospitalization,^2,4,5 and augmented healthcare costs.^2,4 Furthermore, there is evidence to suggest that depression adversely affects adherence to pharmacological and nonpharmacological treatments and healthy lifestyle.⁶

Clinical Perspective on p 472

Although the definite diagnosis of depression requires a structured interview, several validated self-assessment questionnaires may ease detection. The 9-item patient health questionnaire (PHQ-9) is a commonly used instrument facilitating not only diagnosis but also estimation of severity of depressive symptoms^7,8 (Data Supplement). Depression as assessed by the PHQ-9 was recently shown to independently predict healthcare use and mortality in patients with heart failure,⁴ and predicted rehospitalization in subjects with various cardiovascular disorders.⁹

In routine care, depression remains often undiagnosed,¹⁰ and guidelines for specific therapy are lacking in cardiovascular diseases. Reports on depression management in heart failure are scarce. The Sertraline Against Depression and Heart Disease in Chronic Heart Failure (SADHART-CHF) study tested for 12 weeks the selective serotonin antagonist sertraline against placebo.¹¹ As in previous trials in coronary disease (overview in study by Angermann et al¹²), no specific antidepressant effect was found. The Heart Failure-A Controlled Trial Investigating Outcomes of Exercise Training (HF-ACTION) trial showed that aerobic exercise improved depressive symptoms significantly, but only modestly.¹³ Results of the Effects of Selective Serotonin Re-uptake Inhibition on Morbidity, Mortality, and Mood in Depressed Heart Failure Patients (MOOD-HF) study,¹² which evaluates long-term effects of escitalopram on mortality and morbidity, are still pending. Management models, in which clinical care is tailored to individual needs seem promising because they improved mental health-related quality of life¹⁴ and survival.¹⁵ Broader appreciation of the relevance of this comorbidity might enhance development of more targeted treatment strategies, and routine screening for depression in heart failure might be a first step to approach this goal.

The PHQ-2 consists of only the first 2 PHQ-9 items.¹⁶ Its completion took less than a minute in an earlier study, whereas longer questionnaires proved more time-consuming and cumbersome to score.¹⁷ Thus, this instrument can easily be applied in routine care. It was previously recommended for depression screening after myocardial infarction¹⁸ and identified in the Heart and Soul Study subjects at increased risk for adverse cardiovascular outcomes with good specificity.¹⁹

Rollman et al²⁰ recently demonstrated an increased mortality risk in patients with heart failure, in whom depression was suspected on clinical grounds, and who endorsed one or both PHQ-2 items compared with those who responded negatively to both. To date, this instrument has not been explored in unselected patients with heart failure, however, and it is unknown how the prognostic performance of the shorter PHQ-2 compares with that of the PHQ-9 in such populations.

We studied a large consecutive cohort of patients hospitalized for acute cardiac decompensation, who underwent psychometric assessment before discharge. Our primary aim was to evaluate the predictive potential of PHQ-2 scores about both all-cause mortality and rehospitalization risk versus that of the PHQ-9 scores. To aid interpretation of the PHQ screening results, we further aimed to describe in this population the association of depressive symptoms according to PHQ screening with various patient characteristics and the relation of these variables to outcomes.

Methods

Study Flow and Patients

The Interdisciplinary Network for Heart Failure (INH) program evaluated effects and mode of action of a nurse-coordinated telephone-based disease management program (Heart-NetCare-HF) in patients previously hospitalized for cardiac decompensation (unique trial identifier: ISRCTN 23325295) at 9 study sites. Details of study design, end points, and patient characteristics were previously published.²¹ The INH study protocol conformed with the Declaration of Helsinki, and was approved by all responsible ethics committees. All participants of the present post hoc analysis were recruited according to INH inclusion and exclusion criteria,²¹ and provided written informed consent.

Patients were eligible, if aged >18 years and hospitalized with signs and symptoms of decompensated heart failure, and if their left ventricular ejection fraction was ≤40% before discharge from hospital. The only exclusion criteria were new-onset structural heart disease as the primary cause of cardiac decompensation and inability or unwillingness of the patient or logistic reasons precluding participation in a telephone-based healthcare intervention. We report on consecutive patients undergoing depression screening with the PHQ-9 during the INH baseline assessment. Follow-up about all-cause death and rehospitalization was 18 months (100% complete).

Baseline Examination and Depression Assessment

Patients underwent a standardized clinical evaluation, including medical history, physical status, assessment of comorbidities, laboratory testing, ECG, and quantitative echocardiography. Depressive symptoms were assessed using the German version of the PHQ-9, which evaluates the intensity of 9 symptoms reflecting the diagnostic criteria for a depressive episode according to the Diagnostic and Statistical Manual of Mental Disorders IV, and represents an established tool for depression screening in cardiovascular disease.²² The abbreviated PHQ-2 contains only the first 2 questions of the PHQ-9 and has also been validated in different languages, including German.²³ It is regarded as a simple, rapidly to administer proxy of the more time consuming diagnostic interview for depression diagnosis and assessment of severity.^16,23 Both tools refer to symptoms in the last 2 weeks. The answer “not at all” scores 0, “on several days” scores 1, “on more than half of the days” scores 2, and “nearly every day” scores 3, respectively. Overall scores may thus range from 0 to 27 (PHQ-9) and from 0 to 6 score points (PHQ-2), respectively. An American Heart Association Science Advisory recommended further evaluation of PHQ-2 positive cases using the PHQ-9.¹⁸ We extracted the PHQ-2 scores ex post from the PHQ-9 results.

End Points

End points of this post hoc analysis were time to all-cause death or rehospitalization. Eighteen months after the baseline assessment vital status was ascertained in all patients either by a follow-up examination in the INH outpatient clinics or by structured telephone follow-up, which was performed by a trained nurse or a physician. Records of general practitioners or cardiologists, hospital discharge letters, reports from patients and relatives, and death certificates were used as source documents to assess hospital readmissions and to determine the date of death in deceased patients.

Data Analysis and Statistics

Baseline demographics, physical status, symptoms, and diagnostic variables were compared according to PHQ-2 scores, using Kruskal–Wallis testing for continuous and Pearson χ² testing for categorical variables. To evaluate the association of continuous PHQ-2 and PHQ-9 scores with outcomes Cox proportional hazard regression analyses were used, and hazard ratios (HRs) with 95% confidence intervals (CIs) calculated.

According to literature^16,22,24 patients were dichotomized in a nondepressed group (PHQ-2=0–2, and PHQ-9=0–9 score points, respectively) and a group suspected to currently have depression (depressed, PHQ-2>2, and PHQ-9>9 score points, respectively) to allow calculation of univariable HRs for all-cause mortality and rehospitalization. To visualize the analyses Kaplan–Meier curves were created. To demonstrate their correspondence we used cross-tabulation of categorical PHQ-2 versus categorical PHQ-9 results.

To describe the association between different degrees of depression severity and the frequency of various patient characteristics patients were allocated by PHQ-2 scores to 4 subgroups with 0, 1 to 2, 3 to 4, and 5 to 6 score points. To compare all-cause mortality and rehospitalization in these subgroups Kaplan–Meier curves were created.

For prediction models of all-cause mortality and rehospitalization risk Harrell c-statistics²⁵ were calculated from the PHQ-2 and PHQ-9 scores, respectively, both unadjusted and adjusted for age, sex, and treatment allocation. The 95% CI and P values for the test against 0.5 were computed using the Jackknife method.²⁶ Using the same approach, differences between c-statistics of the models for PHQ-2 and PHQ-9 were analyzed. To visualize the comparative predictive value of PHQ-2 versus PHQ-9, receiver-operating characteristic curves were computed.

IBM SPSS version 21.0 software (Armonk, NY) was used for the statistical analyses. A 2-tailed P<0.05 was considered significant.

Results

Baseline Characteristics

Of 1022 consecutive patients recruited according to INH inclusion criteria 852 participated in the psychometric assessment and were included in this analysis. Baseline characteristics of this study population and of the 4 subgroups composed according to the baseline PHQ-2 scores are displayed in Table 1. As a whole, patients were elderly and predominantly men, and the most common pathogenesis of heart failure was coronary artery disease. Subjects with higher PHQ-2 scores were older, had a lower ejection fraction, and were more often on treatment with diuretics and less frequently on β-blockers. Furthermore, chronic obstructive pulmonary disease and more severe symptoms (ie, New York Heart Association [NYHA] functional classes III–IV) were more common in the subgroups with higher PHQ-2 scores, who also had higher amino-terminal pro-B-type natriuretic peptide levels and tended to have more often diabetes mellitus. Elevated depressive symptoms were observed in 33.5% of the patients based on PHQ-2 scores >2 or PHQ-9 scores >9, respectively. Table 2 shows frequencies of categorical PHQ-2 and PHQ-9 values to demonstrate their correspondence. The Spearman correlation coefficient between PHQ-2 and PHQ-9 scores was 0.85 (P<0.001).

Mortality Risk According to PHQ Results

During the 18-month follow-up period, 152 (17.8%) of the study participants died. Both, PHQ-2 and PHQ-9 scores were significant predictors of all-cause mortality risk (HR, 1.18; 95% CI, 1.09–1.29; P<0.001 and HR, 1.07; 95% CI, 1.04–1.09; P<0.001, respectively). According to the dichotomized PHQ scores, mortality risk was significantly higher in patients with elevated depressive symptoms (PHQ-2 score >2 and PHQ-9 score >9) than in patients scoring only ≤2 points in the PHQ-2 and ≤9 points in the PHQ-9, respectively (PHQ-2: HR, 1.70; 95% CI, 1.23–2.33; P=0.001 and PHQ-9: HR, 2.09; 95% CI, 1.52–2.87; P<0.001).

Kaplan–Meier survival curves of dichotomized PHQ-2 and PHQ-9 scores of nondepressed patients versus those, whose scores indicated current depressive episode, are superimposed on each other in Figure 1A. The close vicinity of the PHQ-2 and PHQ-9 curve pairs in each of the 2 subpopulations indicates comparable potential of both questionnaires to predict the risk of all-cause death.

• Download figure
• Download PowerPoint

Rehospitalization Risk According to PHQ Results

Four hundred eighty-two patients (56.6%) were rehospitalized at least once during the 18-month follow-up period. Both PHQ-2 and PHQ-9 scores proved significant predictors of rehospitalization risk in univariable analysis (HR, 1.07; 95% CI, 1.01–1.21; P=0.02, and HR, 1.03; 95% CI, 1.01–1.04; P=0.001, respectively). All-cause rehospitalization risk proved significantly increased for the PHQ-9 dichotomized at >9 score points (HR, 1.30; 95% CI, 1.04–1.51; P=0.02), but not for the PHQ-2 dichotomized at >2 score points (HR, 1.14; 95% CI, 0.95–1.38; P=0.17). Figure 1B depicts the Kaplan–Meier curves for dichotomized PHQ-2 and PHQ-9 scores in the nondepressed versus the depressed subjects to demonstrate their comparative predictive potential also about this end point.

Patient Baseline Characteristics and Prognosis

Table 3 displays associations of the baseline characteristics (as displayed in Table 1) with all-cause mortality and rehospitalization; whereas ischemic pathogenesis, parameters of heart failure severity, various comorbidities, and treatment with diuretics and antidepressants were associated with an increased HR, treatment with angiotensin-converting enzyme inhibitors or angiotensin-2 receptor blockers and β-blockers reduced the HR for both outcomes. Some of the correlates of elevated PHQ-2 scores presented in Table 1 were also associated with increased HRs for all-cause death and rehospitalization.

Comparability of the 2-Item Versus the 9-Item PHQ

The c-statistics for the prediction of all-cause mortality and rehospitalization risk, both unadjusted and adjusted for age, sex, and treatment allocation, are given in Table 4. No significant differences between the predictive values of PHQ-2 and PHQ-9 scores were evident. To visualize their relationship, receiver-operating characteristic curves for PHQ-2 and PHQ-9 scores were created (Figure 2A and 2B), which demonstrate the comparable performance of both questionnaires about prediction of all-cause death, but also a somewhat inferior predictive potential of the PHQ-2 about rehospitalization.

• Download figure
• Download PowerPoint

Severity of Depressive Symptoms and Prognosis

Figure 3A and 3B depict Kaplan–Meier curves for survival and freedom of rehospitalization in the 4 subgroups composed according to the PHQ-2 scores. Compared with the referent subgroup, these curves suggest a dose–response relationship between the severity of the depressive symptoms and the risk of all-cause death, whereas only patients scoring 5 to 6 points exhibited also a significantly increased risk of rehospitalization.

• Download figure
• Download PowerPoint

Discussion

In this analysis of a large, consecutive sample of patients hospitalized for an episode of acute decompensation because of systolic heart failure, we demonstrated that not only PHQ-9 scores but also the PHQ-2 scores obtained at baseline were associated with an increased risk of all-cause death and rehospitalization in the subsequent 18 months. The capacity of the shorter PHQ-2 to identify patients at increased risk was similar to that of the longer and more time-consuming PHQ-9.

Comparability of the 2-Item and the 9-Item PHQ

When dichotomizing our patient population in a nondepressed group and a group with suspected clinically relevant depressive symptoms using cut-off scores as previously proposed^16,22,24 the prevalence of depression was 33.5% (n=285) according to the PHQ-2 as well as PHQ-9 screening results. As indicated by previous meta-analysis,²⁷ prevalence rates of comorbid depression among hospitalized patients vary considerably. Part of this variability is probably because of the use of different instruments and cutoffs for depression assessment. With our definitions and cutoffs depression prevalence proved comparable with both instruments reconfirming also that depression is one of the most frequent comorbidities in heart failure.²

Kaplan–Meier survival curves for subpopulations with and without a suspected episode of current depression showed that both, PHQ-2 and PHQ-9 results predicted mortality risk in a comparable fashion. They also indicated a significantly higher mortality risk in the depressed group. High rehospitalization rates in the first 3 months post discharge in both subpopulations indicated that factors other than depression were in this period primary determinants of that outcome. From month 4 onwards rehospitalization declined more in the nondepressed than the depressed subjects. Because this difference remained small in absolute terms, however, the clinical relevance of the lower predictive power observed with PHQ-2 screening seems negligible.

C-statistics comparing PHQ-2 and PHQ-9 confirmed that the loss of prognostic power was small and nonsignificant when PHQ-2 instead of PHQ-9 scores were used for depression assessment. Furthermore, direct comparison of the corresponding receiver-operating characteristic curves showed comparable predictive values.

Like previous studies in patients with cardiovascular diseases,^28,29 which used the Beck Depression Inventory as a psychometric tool, PHQ screening identified patients at risk for all-cause death and rehospitalization. Our results are in line with recent findings by Moraska et al⁴ who using the PHQ-9 with a cut-off score of >10 as depression criterion reported a 4-fold increase in all-cause mortality and nearly 2-fold increased hospitalization rate in the highest quartile of a large heart failure cohort during a mean follow-up of 1.6 years. Our study extends these findings to the PHQ-2. Rollman et al²⁰ failed to demonstrate a dose–response relationship when applying the PHQ-2 in patients with heart failure, in whom depression was already suspected on clinical grounds. The authors considered this as a possible consequence of clinical prescreening, which rendered only high-risk patients eligible for enrolment. More congruent with our results, Beach et al⁹ found in a secondary analysis from a depression management trial³⁰ in subjects with various cardiac disorders that higher initial depression scores were also associated with higher cardiac readmission rates at 6 months.

Correlates of Depression and Predictive Relevance of Patient Baseline Characteristics

Various features of the heart failure syndrome were previously identified as depression correlates, such as comorbidities, low left ventricular ejection fraction, and severity of heart failure symptoms,³¹ and our results strengthen these observations. Meta-analysis demonstrated a rise of aggregate depression rates from 11% in NYHA class I to 42% in NYHA class IV.² In accordance with this observation, prevalence and severity of depression according to PHQ-2 scores were associated with the severity of heart failure symptoms in our analysis because >50% of patients scoring 3 or 4 and almost 70% of patients scoring 5 or 6 points were in NYHA class III or IV, whereas the majority of subjects with lower scores were in NYHA class I or II. Among the postulated mediators involved in the reciprocal adverse effects of cardiovascular disorders and their comorbidities and depression the degree of dysregulation of autonomic nervous control, the level of circulating catecholamines, the extent of activation of inflammatory, oxidative, and nitrosative pathways and increased hemodynamic stress are recognized markers of heart failure severity.³² Vice versa, NYHA class is a strong predictor of incident depression.⁶ Furthermore, increasing depressive symptoms were associated with worsening heart failure,³³ and depression remission was associated with improved cardiac outcomes.³⁴ Last but not least, outcomes may be modulated by behavioral risk factors as unhealthy lifestyle habits,⁶ nonadherence to treatment and lack of social support.³¹

Although our present observations are in accordance with the concept of a pathophysiological link between heart failure and depression, not all patient characteristics with adverse prognostic significance correlated also with higher depression scores. Coronary artery disease was, for example, associated with an increased HR for both, death and rehospitalization, but not with higher PHQ scores. Although treatment with angiotensin-converting enzyme inhibitors or angiotensin-2 receptor blockers and β-blockers improved outcomes significantly, heart failure medication was not consistently associated with lower depression scores. To what extent the correlation of depression history and use of antidepressants or diuretics with higher PHQ-2 scores and concomitant higher HRs for adverse outcomes relates to more severe depression or heart failure or to unwanted drug effects cannot be clarified from our data. Altogether our results indicate, however, a partial overlap between the effects of depression and those of clinical prognosticators.

Depression—Rarely Diagnosed in Heart Failure

Before the index hospitalization, depression was unknown in the majority of patients because only 17.1% of participants scoring 3 to 4 and 22.1% of those scoring 5 to 6 at PHQ-2 screening reported a history of depression, and even less, 10.6% and 15.1%, were on antidepressants, which may, however, also reflect the lack of evidence-based pharmacotherapy. Compared with a contemporary American cohort described by Moraska et al⁴ these proportions are even smaller, thus re-emphasizing the need for better understanding of the pathophysiological links between cardiovascular disease and comorbid depression and better appreciation of this comorbidity in clinical practice. Patients with heart failure may often misinterpret emotional distress as integral part of their heart disease, or are reluctant to communicate such feelings for fear of being labeled as mentally ill. In turn, cardiologists and general practitioners may lack adequate training and time to diagnose comorbid depression. It is essential that both, patients and physicians recognize that depressive symptoms are not an integral component of the heart failure syndrome. Routine PHQ-2 screening bears the promise of raising this awareness.

Limitations and Strengths

Several limitations of this study must be acknowledged. First, self-reports on depression were not confirmed by structured clinical interview, which must remain the gold standard for depression diagnosis. However, this study aimed primarily to assess the capability of the PHQ-2 compared with the PHQ-9 to identify patients at increased risk of mortality and rehospitalization. To this end, recognition of depression using these tools seemed more important than precise quantification of the depressive symptoms. Moreover, in a previous validation study, PHQ screening results compared favorably also with the structured clinical interview.²² Another limitation is that only patients with impaired left ventricular ejection fraction were enrolled. Our findings can, therefore, not be generalized to the entire heart failure population. We further applied a limited number of exclusion criteria, and not all patients enrolled in the INH study followed the invitation for psychometric evaluation. We can thus not exclude that patients eligible and willing to complete the questionnaires differed from nonparticipants. Strengths of our study include the large sample size and the prospective nature of the protocol, precise characterization of the study cohort and careful end point collection and assessment. Eighteen months follow-up was 100% complete, all hospital admissions were assessed from discharge letters, and information on death was collected from hospital records, death certificates, and reports from relatives and physicians.

Conclusions

In this large consecutive cohort evaluated before hospital discharge after an episode of acute cardiac decompensation, the potential of the PHQ-2 to identify patients at increased risk for death and rehospitalization proved comparable with that of the longer PHQ-9. Depression severity as assessed by the PHQ-2 correlated with various patient characteristics, and a dose–response relationship between PHQ-2 scores and the risk of mortality and rehospitalization was demonstrated. Among the patient characteristics, which correlated with elevated PHQ-2 scores, some were also associated with an increased HR for all-cause death and rehospitalization indicating a partial overlap of depression effects with those of clinical prognosticators. Because better knowledge about patients’ mood might thus aid risk stratification and potentially help to improve patient management, depression screening should become part of the routine clinical assessment of patients with heart failure. Our results demonstrate that the PHQ-2 represents a suitable tool for this purpose.

Footnotes