Mental Disorders, Quality of Care, and Outcomes Among Older Patients Hospitalized With Heart Failure:  An Analysis of the National Heart Failure Project FREE

Previous studies have reported that patients with mental illness are at increased risk for developing cardiovascular diseases, including heart failure.^1- 4 This literature suggests not only an increased risk of incident heart failure but also poorer outcomes after diagnosis. Limited evidence suggests that patients with mental illness and heart failure require more expensive care⁵ and experience higher rates of hospitalization and mortality than those without psychiatric comorbidities.^6- 15 Such studies are based on patients drawn from single centers, registries, regions, or other selected patient populations.^{2,6- 7,9- 15} Furthermore, although several studies have assessed the effect of depression on heart failure outcomes,^6- 7,9- 15 comparatively few data are available regarding schizophrenia,² substance abuse, and other types of mental illness.¹⁶ Notably absent is an assessment of the specific treatment provided to patients with mental illness and heart failure. Although mental illness is associated with reduced rates of procedure use¹⁷ and quality of care¹⁸ in patients with myocardial infarction, few comparable data exist for patients with heart failure.

To address this issue, we undertook an assessment of a national cohort of Medicare patients hospitalized with heart failure to assess the effect of mental illness on treatment and outcomes. We specifically sought to assess whether mental illness was associated with differences in quality of care during hospitalization or readmission and mortality in the year following hospitalization.

The National Heart Failure (NHF) Project was a Centers for Medicare and Medicaid Services initiative to improve quality of care for Medicare beneficiaries hospitalized with heart failure. The NHF sampling and data collection strategy has been described in greater detail previously.¹⁹ Briefly, Medicare fee-for-service beneficiaries hospitalized from April 1, 1998, through March 31, 1999, and July 1, 2000, through June 30, 2001, with a principal discharge diagnosis of heart failure (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 402.01, 402.11, 402.91, 404.01, 404.91, or 428)²⁰ were identified and sampled by state. Records were forwarded to 1 of 2 data abstraction centers and underwent detailed abstraction for clinical data, including medical history, comorbidity, findings at admission, in-hospital course, status at discharge, medication and procedure use, and laboratory and other diagnostic evaluations. Trained reviewers, medical record abstraction software, and random record reabstraction were used to optimize data quality.¹⁹

The initial NHF Project sample included 78 882 medical records. Medical records of patients who had been readmitted (n = 3732), those undergoing chronic hemodialysis (n = 549), those who left the hospital against medical advice (n = 59), and those with no or an invalid Social Security number (n = 48) were excluded, in accordance with NHF Project exclusion criteria. Patients younger than 66 years (n = 7910), those who arrived by interhospital transfer (n = 2419), and those without heart failure at admission (n = 5003) were also excluded to ensure examination of a representative cohort of older patients directly admitted with heart failure. In total, 17 707 patients met 1 or more of these exclusion criteria, which left 61 175 patients eligible for analysis.

Data from the NHF Project were linked with the American Medical Association Physician Masterfile using the unique physician identification number of the attending physician for each hospitalization.²¹ Physician characteristics included self-reported specialty and board certification. Hospital characteristics, including bed size, US Census region, rural location, level of cardiac care facilities, ownership, and teaching status, were obtained by linking data from the 1998 American Hospital Association Survey of Hospitals.²²

Building on previous studies with similar methods,^17- 18 a mental illness diagnosis was identified by the presence of an administrative secondary diagnosis code of “any mental disorder” (ICD-9-CM codes 295.00-319.99) recorded at admission during the index hospitalization or at any encounter in the previous year. Patients for whom Medicare Part A administrative data were not available (n = 2516) were excluded from the analysis. Patients with active dementia, delirium, and other organic psychotic conditions (ICD-9-CM codes 290.00-294.99) were also excluded (n = 5345) because their diagnoses suggested a possible medical pathogenesis associated with a substantially higher risk of mortality rather than a mental illness. Exclusion of these patients produced the final study cohort of 53 314 patients.

The NHF Project's evaluation of patient treatment focused on the evaluation of the left ventricular ejection fraction (LVEF) and the prescription of angiotensin-converting enzyme (ACE) inhibitors, both widely accepted measures of quality of care for patients with heart failure.^19,23 Prescription of ACE inhibitor therapy at discharge was assessed among ideal candidates, defined as patients with left ventricular dysfunction who had no contraindications to ACE inhibitor therapy. To account for the possible use of angiotensin receptor blockers (ARBs) in lieu of ACE inhibitors, we also examined the use of ACE inhibitors or ARBs among ideal candidates for ACE inhibitors. Patients were considered to have undergone LVEF evaluation if their medical record contained an LVEF value measured before admission, their LVEF was measured during hospitalization, or the medical record contained a note of a planned LVEF assessment following discharge. Quality of care indicators are described in the Box.

Mortality within 1 year of admission, all-cause readmission, readmission for heart failure, and readmission for mental illness within 1 year of discharge were available for all patients and were assessed using the Medicare Enrollment database and billing records.²⁴

Differences in medical history, comorbidities, admission characteristics, and quality indicator eligibility between patients with and without mental illness diagnoses were assessed using χ² and Wilcoxon rank sum tests. Crude differences in quality of care and outcomes between patients with and without mental illness diagnoses were evaluated using χ² tests. Multivariate analyses were performed to determine whether differences in quality of care or outcomes among patients with mental illness diagnoses were independent of other factors. Logistic regression models adjusted for covariates used in previous studies,^25- 26 including patient age, sex, race, medical history (coronary artery disease, diabetes mellitus, hypertension, chronic obstructive pulmonary disease, LVEF, and admission source), admission characteristics (heart rate and serum sodium and creatinine levels), and severity of illness, as measured by a modified version of the Medicare Mortality Prediction System (including age, cancer, mobility, prior congestive heart failure, mean arterial pressure, heart rate, serum urea nitrogen level, white blood cell count, and the Acute Physiology and Chronic Health Evaluation II score).²⁷ Analyses also adjusted for physician specialty, physician board certification, and hospital characteristics, including number of beds, rural location, US Census region, ownership, teaching status, and level of cardiac care facilities.

Dummy variables were used to denote patients with missing physician or hospital data. Missing continuous variables (heart rate, n = 56; serum sodium level, n = 982; and serum creatinine level, n = 1037) were replaced with the study cohort's median value and a dummy variable indicating the replacement of missing data.

Analyses incorporated probability weights based on the inverse sampling fraction for each state. Statistical analyses accounted for clustering of patients by hospital and were conducted using Stata statistical software, version 8.0 (Stata Corp, College Station, Texas). This study was approved by the Yale University School of Medicine Human Investigation Committee.

Patients with mental illness diagnoses accounted for 17.0% of the study cohort. The most commonly documented mental illness diagnoses were depressive disorder (ICD-9-CM 311), neurotic disorder (ICD-9-CM 300), and nondependent abuse of drugs (ICD-9-CM 305); most other mental illness diagnoses were documented infrequently (Table 1). Compared with patients without mental illness diagnoses, patients with mental illness diagnoses were, on average, younger than patients without mental illness diagnoses and a greater proportion were female; white; had a history of heart failure, chronic obstructive pulmonary disease, or mobility limitations; and were admitted from residential facilities (Table 2).

Overall, 98.2% of patients were eligible for LVEF evaluation (98.3% patients with mental illness diagnosis vs 98.2% patients without mental illness diagnosis; P = .12) and 16.4% for evaluation of ACE inhibitor prescription (15.3% vs 16.7%; P = .02) or ACE inhibitor or ARB prescription rates among patients classified as ideal for ACE inhibitor therapy (Table 3).

Among patients classified as ideal for ACE inhibitor prescription, those with and without mental illness diagnoses had comparable crude rates for the end points of ACE inhibitor prescription (69.7% vs 71.3%, respectively; P = .40) and ACE inhibitor or ARB prescription (78.5% vs 79.7%; P = .47). Patients with mental illness diagnoses had lower rates of LVEF evaluation than did patients with no mental illness diagnoses (47.3% vs 53.0%; P <.001). These patterns of care were unchanged after multivariate adjustment. Patients with mental illness diagnoses were less likely to undergo LVEF evaluation (adjusted odds ratio [OR], 0.81; 95% confidence interval [CI], 0.76-0.87) but had comparable odds of being prescribed an ACE inhibitor (adjusted OR, 0.96; 95% CI, 0.80-1.14) or an ACE inhibitor or ARB (adjusted OR, 0.95; 95% CI, 0.78-1.17) (Table 3).

Patients with mental illness diagnoses had higher crude rates of 1-year all-cause readmission (73.7% vs 68.5%; P <.001), 1-year heart failure readmission (65.2% vs 60.6%; P <.001), and 1-year mental illness readmission (31.1% vs 7.7%; P <.001) than did patients without mental illness diagnoses. After multivariate adjustment, patients with mental illness diagnoses had a persistently higher likelihood of 1-year all-cause readmission (adjusted OR, 1.30; 95% CI, 1.21-1.39), 1-year heart failure readmission (1.23; 1.15-1.32), and 1-year mental illness readmission (5.13; 4.73-5.56). Patients with mental illness diagnoses had a higher 1-year mortality rate than did patients without mental illness diagnoses (41.0% vs 36.2%; P <.001). Presence of a comorbid mental illness diagnosis continued to be associated with 1-year mortality after multivariate adjustment (adjusted OR, 1.20; 95% CI, 1.12-1.28) (Table 4).

In a large, nationally representative sample of elderly patients hospitalized with heart failure in the United States, nearly 1 in 6 had a secondary diagnosis of mental illness. Patients with a mental illness diagnosis received poorer to comparable quality of care, were more likely to be readmitted within the year following discharge, and had an increased mortality risk. Our findings identify areas of concern in the treatment and outcomes of patients with heart failure and mental illness.

The quality of medical care for patients with mental illness diagnoses presents a mixed picture, with no significant differences in ACE inhibitor prescription but less frequent left ventricular systolic function evaluation. All patients in the study cohort had Medicare insurance and, by being hospitalized, had demonstrated access to the health care system, thus diminishing the explanatory role of these factors. Physicians may be referring patients with mental illness for LVEF assessment at a lower rate than those without mental illness. Possible reasons for lower referral rates may range from less aggressive management of heart failure in patients with mental illness to a more insidious bias arising from the stigma associated with mental illness.²⁸ It is also possible that patients with mental illness may be unable or unwilling to undergo LVEF assessment.

Consistent with the results of previous studies,^9- 10,13,15 we identified a notably higher risk of all-cause readmission among patients with mental illness, as well as increased risks of readmission for heart failure, even after adjustment for a variety of comorbid conditions. There are several possible mechanisms underlying the increased risk of readmission for heart failure among patients with mental illness. Patients with mental illness may be less adherent to treatment recommendations, including medication and rehabilitation regimens, thereby increasing their likelihood of readmission.^29- 30 Differences in heart failure symptom recognition may result in patients with mental illness deferring care until initial problems are sufficiently advanced that hospitalization is required. Mental illness may also cause patients with heart failure to underestimate their functional status, particularly for patients with depressive symptoms,³¹ thereby precipitating readmission when it may not be clinically necessary. Patients with mental illness were disproportionately women and admitted from residential facilities, characteristics associated with lower levels of social support.⁸ Poor social support after hospital discharge may diminish mentally ill patients' abilities to cope with heart failure and thereby predispose them to readmission. Mental illness may also be a proxy for other unmeasured characteristics, including lower socioeconomic status and greater functional impairment, which are also associated with high rates of heart failure readmission.³² However, without detailed information concerning the quality of processes of care after discharge, it is unclear whether mental illness is independently associated with readmission or reflects poorer quality of care after discharge.

Heart failure readmission and mental illness readmissions may arise from a synergy of common processes underlying exacerbations of both conditions. For example, sympathetic nervous system activation has been observed in patients with depression.³³ Augmentation of the sympathetic nervous system is also associated with the pathogenesis of heart failure and decompensation among patients with heart failure.³⁴ Therefore, enhanced sympathetic nervous system activity in patients with heart failure may act to worsen depression and vice versa, thereby producing deterioration in both disease conditions.

Our findings of an increased risk of mortality associated with mental illness in patients with heart failure are consistent with previous reports of increased harm.^{2,6- 9,11- 15} Some of the same mechanisms, which may account for higher rates of readmission in patients with mental illness and heart failure, likely also contribute to higher mortality, including poorer adherence after discharge and poorer access to follow-up care. Neurohumoral alterations and autonomic dysfunction associated with mental illness may also serve as more proximal antecedents of excess mortality,³⁵ particularly for patients with heart failure.³⁶ However, the magnitude of the mental illness–associated adjusted 1-year mortality risk (OR, 1.20) is notably lower than previously published estimates of the mortality risk conferred by mental illness in patients with heart failure.^{6- 9,13- 15} The attenuated mortality hazard associated with mental illness observed in our cohort is similar to that reported in a previous study of the effect of a mental illness diagnosis on elderly patients with myocardial infarction.¹⁸ Studies finding higher mortality rates typically used diagnostic interviews rather than physician diagnoses to identify mental disorders,^{6- 9,13- 15} which may identify persons with more acute symptoms or those who are untreated. The advanced age common to both studies suggests that the decreased mortality risk may reflect survivorship bias whereby patients with mental illness who survive to older age represent a healthier cohort than patients without mental illness in ways that are unmeasured in this study. Concerns of such selection bias are heightened by the shorter life expectancies associated with most mental illnesses.³⁷

The NHF Project constitutes the largest national, contemporary evaluation of elderly patients hospitalized with heart failure. The use of quality-of-care indicators and detailed clinical data represents significant methodological improvements on prior assessments of heart failure treatment patterns. However, our study has certain limitations that merit consideration. Administrative data produced from a single hospital encounter may not adequately identify all patients with mental illness.³⁸ We sought to limit this effect by also evaluating administrative data for the full year preceding hospitalization. Nevertheless, mental illness remains markedly underdiagnosed by physicians in the community and the hospital setting.³⁹ As such, our findings should not be seen as an estimate of the true prevalence of mental illness in elderly patients with heart failure but rather as an assessment of physician-recognized mental illness and its effects on patient treatment.

We restricted our analysis to Medicare patients 66 years and older who were hospitalized with heart failure; therefore, our findings may not be applicable to patients younger than 66 years or those not enrolled in fee-for-service Medicare. However, more than 80% of heart failure patients in the United States are 65 years or older⁴⁰ (although this proportion may be smaller for patients with mental illness, who have an earlier median age of heart failure onset), and patients enrolled in Medicare managed care constitute less than 15% of Medicare beneficiaries,⁴¹ suggesting that both exclusions may result in only a minor loss in study generalizability. Our findings may not be applicable to patients who are not hospitalized or to practice patterns in the ambulatory setting. However, acute exacerbations of heart failure requiring hospitalization are common, particularly among elderly patients.⁴²

Patients with mental illness diagnoses are a notable subgroup of elderly persons hospitalized with heart failure in the United States. As a group, these patients receive poorer to comparable quality of care during hospitalization and are more likely to be rehospitalized or die in the year following discharge than patients without mental illness. Greater efforts are needed to improve quality of care and reduce rehospitalization for elderly patients with heart failure and mental illness.

Correspondence: Saif S. Rathore, MPH, Section of Cardiovascular Medicine, Department of Internal Medicine, Yale University School of Medicine, Rm IE-61 Sterling Hall of Medicine, 333 Cedar St, PO Box 208025, New Haven, CT 06520-8025.

Submitted for Publication: April 11, 2008; final revision received June 25, 2008; accepted June 25, 2008.

Author Contributions: Mr Rathore and Drs Druss, Masoudi, and Krumholz had full access to all the data in the study and take full responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant GM07025 from the National Institute for General Medical Sciences Medical Scientist Training (Mr Rathore), a Clinical and Translational Science Award (grant UL1 RR024139) from the National Center for Research Resources (Mr Rathore), and grants 5K24MH07586702 and 5RO1MH07043703 from the National Institute of Mental Health (Dr Druss). The analyses on which this publication is based were performed under contract number HHSM-500-2006-OK003C, entitled “Utilization and Quality Control Peer Review Organization for the State of Oklahoma,” sponsored by the Centers for Medicare and Medicaid Services, Department of Health and Human Services and under federal contract number 4-662-OK-0608.

Disclaimer: The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US Government.

Previous Presentations: The results of this study were presented in part at the American Heart Association Quality of Care and Outcomes Research in Cardiovascular Disease and Stroke Conference 2008; May 1, 2008; Baltimore, Maryland.