Resource Use in the Last 6 Months of Life Among Medicare Beneficiaries With Heart Failure, 2000-2007

Provision of high-quality health care at the end of life poses challenges for both health care providers and policy makers. End-of-life care has many dimensions, including patient preferences and values, health care provider practices, and concerns about the appropriate use of resources. Although most patients prefer to die at home, many die in hospitals or nursing homes.^1-3 The cost of health care at the end of life is also substantial. More than one-quarter of Medicare spending occurs in the last year of life, a figure that has remained stable for several decades.^4-6

Heart failure is listed on 1 in 8 death certificates in the United States.⁷ Although some people live with heart failure for years, more than one-quarter of Medicare beneficiaries die within 1 year of the incident diagnosis, and 36% die within 1 year of a heart failure–related hospitalization.^8,9 Palliative care, sometimes in concert with evidence-based, disease-modifying therapy, is increasingly advocated for the treatment of advanced heart failure when patients have substantial cardiac dysfunction and symptoms such as dyspnea or fatigue at rest or with minimal exertion.^10-13 Palliative care aims to improve quality of life and relieve suffering for patients with serious disease, whose prognosis may be uncertain, using a multidisciplinary and holistic approach.¹⁴Hospice care, like palliative care, involves aggressive symptom control for patients with advanced disease, but the hospice benefit under Medicare requires a prognosis of life expectancy of 6 months or less and does not cover curative treatments. The hospice movement initially focused on the provision of home-based care for patients with cancer, and cancer remains the most prevalent hospice diagnosis.¹⁴ However, patients with advanced heart failure experience many distressing symptoms¹⁵ and, depending on prognosis and patient preferences, might benefit from palliative care consultation and hospice referral.¹¹

Understanding the health care experiences of people who die with heart failure, and how those experiences have changed over time, provides necessary context to discussions of new models of care for this population. Therefore, we examined health care resource use in the last 6 months of life among elderly Medicare beneficiaries with heart failure from 2000 through 2007.

We used the 5% national sample from the Medicare inpatient, outpatient, and carrier standard analytic files and the corresponding denominator files. Specifically, we obtained research-identifiable files for January 1, 1991, through December 31, 2007, from the Centers for Medicare and Medicaid Services (CMS). These files include institutional claims for facility costs covered under Medicare Parts A and B and noninstitutional claims for physician services covered under Medicare Part B. The denominator files include beneficiary demographic characteristics, dates of death, and program eligibility and enrollment information. We also obtained the Medicare 5% analytic files for skilled nursing facilities, hospice care, home health care, and durable medical equipment for 1999 through 2007. We limited the analysis to beneficiaries 65 years or older living in the United States. Only claims filed during periods of fee-for-service coverage were included. The institutional review board of the Duke University Health System, Durham, North Carolina, approved the study.

The study population included Medicare beneficiaries with heart failure who died between January 1, 2000, and December 31, 2007, and within 8 years of an incident diagnosis of heart failure. We identified beneficiaries with heart failure on the basis of a heart failure diagnosis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM ] codes 428.xx, 402.x1, 404.x1, or 404.x3) on a single inpatient claim or on 3 or more carrier or outpatient claims within 20 months.⁸ We defined the date of the incident diagnosis as the earlier of (1) the date of the earliest inpatient diagnosis or (2) the date of the third outpatient or carrier diagnosis. Beneficiaries with an incident diagnosis were required to have at least 12 months of continuous fee-for-service Medicare eligibility before the incident date. Only beneficiaries with continuous fee-for-service eligibility from the incident date until death were included in the analysis.

Demographic characteristics available for the analysis included age at death, sex, and race. Medicare beneficiaries report race at the time of enrollment, and we used the category “black” and grouped all others as “other” for this analysis.¹⁶ Comorbid conditions were defined using previously described coding algorithms^17,18 based on inpatient, outpatient, and carrier claims for the 365 days preceding the date of death. Comorbid conditions included cancer, cerebrovascular disease, chronic obstructive pulmonary disease, dementia, diabetes mellitus, hypertension, ischemic heart disease, peripheral vascular disease, and renal disease (eTable 1). We recorded the number of comorbid conditions per patient and created a composite variable indicating the presence of 4 or more comorbid conditions.

For data on resource use, we searched the inpatient files for all-cause hospitalizations in the 180 days before the date of death. Among patients with a hospitalization in the 180 days before death, we calculated the number of days in the hospital and the number of days in an intensive care unit (ICU). We calculated days in the ICU by summing the number of units billed for revenue center codes 200 through 209. We calculated the cost to Medicare of inpatient care by summing the payment amounts and per diem adjustments for all inpatient claims in the 180 days before death. We identified heart failure hospitalizations on the basis of diagnosis-related group (DRG) 127 before October 1, 2007, and DRG 291, 292, or 293 on or after that date.

We also examined the use of skilled nursing facilities, hospice care, home health care, and durable medical equipment. For skilled nursing facilities, we calculated the number of days in a skilled nursing facility for distinct stays and summed payment amounts for the time period. Medicare provides limited coverage for skilled nursing facility care up to 100 days after a hospitalization of 3 or more days.¹⁹ For hospice care and home health care, we calculated the total number of days of enrollment and summed the payment amounts for claims submitted in the 180 days before death. Medicare covers home health care for homebound patients with short-term needs for skilled care, such as nursing care or physical therapy. The Medicare hospice benefit covers comprehensive supportive and medical services if a physician has certified that the patient is likely to have 6 or fewer months to live and has elected to receive hospice benefits in place of routine Medicare benefits.²⁰ For durable medical equipment, we summed payment amounts for the time period.

For other outpatient resource use, we used Healthcare Common Procedure Coding System (HCPCS) codes 992.xx-994.xx to identify outpatient evaluation and management visits with a physician. To avoid double counting, we considered only 1 physician visit per patient, per place of service, and per day. We examined carrier claims to identify the use of echocardiograms, stress tests, cardiac catheterization, coronary artery bypass graft surgery (CABG), implantable cardioverter-defibrillator placement, pacemakers, ventricular assist devices, and dialysis (eTable 2).

Finally, we calculated mean costs to Medicare for inpatient services, outpatient services, physician services, skilled nursing facility care, hospice care, home health care, and durable medical equipment by summing the payment amounts by claim type in the 180 days before death across all patients. Total costs to Medicare included Medicare Part A inpatient services, skilled nursing facility care, hospice care, and home health care, and Medicare Part B outpatient facility, physician, and durable medical equipment claims. We adjusted all costs to 2007 US dollars.

We constructed mutually exclusive categories based on the level of care reflected in the claims submitted at the time of death. We first identified patients who died during an inpatient stay. For patients not hospitalized at the time of death, we searched for a skilled nursing facility claim or a hospice claim at the time of death or an indication of death at the time of service. For patients not receiving inpatient, skilled nursing, or hospice care at the time of death, we looked for a home health claim at the time of death. For the remaining patients, we categorized the level of care at the time of death as “other/unknown.” This category included patients who died at home and were not receiving home health or hospice services provided through Medicare, as well as patients who died in nursing facilities not paid by Medicare.

We present categorical variables as frequencies and continuous variables as means (SDs). We applied a Cochran-Mantel-Haenszel test for nonzero correlation to test for yearly trends. We used generalized linear models with a log link and Poisson distribution to examine the unadjusted and adjusted relationships between covariates and total Medicare costs in the last 6 months of life. Covariates included age at death, sex, race, comorbid conditions, duration of heart failure, US geographic region, and year of death. We used SAS statistical software (version 9.2; SAS Institute Inc, Cary, North Carolina) for all analyses.

The study population included 229 543 Medicare beneficiaries with heart failure who died between January 1, 2000, and December 31, 2007. The mean age at death was 83 years, and nearly a quarter of patients were older than 90 years (Table 1). Days from incident diagnosis to death declined slightly during the study period (P < .001). The presence of most comorbid conditions increased over time, most notably renal disease and hypertension. In 2000, 61% of patients had 4 or more comorbid conditions, and the percentage increased to 73% in 2007 (P < .001).

As shown in Table 2, approximately 80% of patients with heart failure were hospitalized in the last 6 months of life, and this finding was consistent throughout the study period. Although the mean number of inpatient days remained steady, days in an ICU increased from 3.5 to 4.6 (P < .001). A higher percentage of patients spent time in skilled nursing facilities in 2007 compared with 2000 (38.8% vs 33.4%; P < .001), and length of stay increased during this period. Hospice use in the last 6 months of life increased dramatically, from 19% of patients to almost 40% (P < .001). Total days in hospice also increased from 36.5 days to 44.0 days (P < .001). In 2000, approximately 5% of patients who enrolled in hospice did so after a hospitalization. Over time, an increasing percentage of patients in hospice care entered hospice directly from a hospital discharge (34% in 2007). Nearly one-third of patients received home health services. The mean number of outpatient physician visits increased slightly (from 7.0 to 8.3; P < .001). The rate of echocardiography use increased from 41.5% to 51.3%, but the use of other tests and procedures was low and remained steady. The use of ventricular assist devices was too low to be reported.

The mean cost to Medicare per patient rose 26% during the study period (from $28 766 to $36 216; P < .001) (Table 3). From 2000 to 2007, the cost to Medicare of physician services increased by 24% (from $4319 in 2000 to $5334 in 2007). Consistent with the increased use of hospice, the mean cost of hospice care per patient more than doubled (from $964 to $2594).

The proportion of patients who died in the hospital declined slightly during the study period (from 40.2% to 35.2%; P < .001), although the use of the ICU during terminal hospitalizations increased (from 42.4% to 50.2%; P < .001) (Table 4). The proportion of patients who died while receiving Medicare-covered services in a skilled nursing facility remained stable over time at approximately 10%. The use of home health services at the time of death was low and did not change over time. More than one-third of patients died while receiving hospice in 2007, up from 17% in 2000. The proportion of patients who died at home (without home health or hospice) or during a skilled nursing facility stay not covered by Medicare declined during the study period.

After adjustment for age, sex, race, comorbid conditions, duration of heart failure, and geographic region, the cost of care in the last 6 months of life increased by 11% from 2000 to 2007 (cost ratio, 1.11; 95% confidence interval, 1.10-1.13) (Table 5). Increasing age was strongly and independently associated with lower costs, whereas renal disease and chronic obstructive pulmonary disease were strong independent predictors of higher costs. Costs were highest among patients who died within 1 year of an incident diagnosis. Black race was independently associated with higher costs. Regional differences persisted after adjustment, with higher costs of care in the Northeast and West compared with the South.

To our knowledge, ours is the first study in a nationally representative sample to describe temporal trends in care at the end of life in a well-defined cohort of patients with heart failure. We found that the care of patients with heart failure in the last months of life has changed over time, including a large increase in the use of hospice services. Despite this shift to greater use of hospice care, there continues to be a high level of use of inpatient care in the last 6 months of life, with approximately 80% of patients spending some time in the hospital. After adjustment for differences in patient demographic characteristics and disease burden, total costs to Medicare from 2000 through 2007 increased by 11%.

The use of hospice care increased dramatically during the study period, a trend described in other studies^21-23 and consistent with the prominence of heart disease as the leading hospice diagnosis (12.2% in 2006) apart from cancer.²⁴ Some studies have found hospice care to be more cost-effective than nonhospice care,^25-27 but we did not observe lower use of other services as the use of hospice increased. Rates of inpatient hospitalization remained high, suggesting that the potential for hospice to prevent costly hospitalizations has yet to be fully realized. It remains unclear whether hospice serves as a complement to or a substitution for usual acute care.⁶

In the 1990s, researchers described decreasing mean lengths of stay in hospice, driven largely by increasing numbers of patients with short hospice stays.^22,28 In contrast to these studies, we found that the mean length of hospice stay increased from 2000 through 2007. The percentage of patients with short stays—less than 3 days (approximately 19%) or less than 7 days (approximately 37%)—remained steady. Short durations of hospice care provide inadequate opportunity for patients and their families to receive the full benefit of hospice services and limit the ability of hospice care to affect resource use. Our findings suggest that, even with greater awareness and use of hospice services by patients and physicians, the proportion of patients who receive services only during the last few days before death remains substantial.

One barrier to timely hospice referral for patients with heart failure is the difficulty of predicting life expectancy, even for patients with advanced disease.^10,29 A recent study of patients in hospice care found that patients with heart failure were more likely than patients with cancer to use hospice services longer than 6 months or to be discharged alive.³⁰ These findings call into question whether the hospice model is designed to meet the needs of patients with advanced heart failure, who may have a variable disease course.^10,14

With findings consistent with trends reported among both Medicare beneficiaries who died and those who did not,⁶ we determined that risk-adjusted costs increased by 11% during the study period. One potential driver of rising costs is the greater use of medical tests and procedures among elderly patients. We found low rates and no change over time in the use of costly invasive cardiac procedures in the last 6 months of life, including cardiac catheterization, pacemaker placement, implantable cardioverter-defibrillator implantation, and CABG. However, we found increasing use of echocardiography. The use of ventricular assist devices in this cohort was too low to be reported. Ventricular assist devices are costly,³¹ and recent research showing improved survival with newer devices may lead to an increase in their use among patients with advanced heart failure.³² We observed a significant increase in ICU days. Also, despite no increase in hospitalizations, there was a substantial increase in the use of skilled nursing facilities, indicating that a greater proportion of patients are being discharged to post–acute care settings instead of home.

This analysis of resource use and costs associated with heart failure at the end of life has both clinical and policy implications. First, a hospitalization of a patient with heart failure presents an important opportunity to discuss the goals of care. Previous research has found a high risk of death in the year after hospitalization for heart failure,⁹ and 80% of the patients in our cohort had at least 1 hospitalization. In patients who have a poor prognosis despite receiving the best medical care, palliative care consultation may be appropriate. Some studies of palliative care have demonstrated improvements in patient satisfaction and cost savings.^33,34 Although an increasing proportion of patients in hospice care in our analysis were enrolled directly from an inpatient hospitalization, only about 15% of hospitalized patients were discharged to hospice in 2007.

Discussions of the goals of care could be considered as an inpatient measure of quality for patients with heart failure. Current measures endorsed by the CMS include provision of instructions at discharge, assessment of left ventricular function, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker at discharge, and smoking cessation counseling. It will be critical to show that discussions of goals of care have the desired effects on patient-centered outcomes.

Studies that “look back” at health care resource use in the last months of life among patients who have died can provide useful descriptions of care in multiple settings. Although not without limitations, such studies are most appropriate for serious, eventually fatal chronic illnesses like heart failure because the timing of death is unpredictable.³⁵ However, findings from such studies cannot be interpreted as reflecting the care of patients who are expected to die and thus provide no insight about the relationship between resource use and outcomes.³⁵ A recent study³⁶ used administrative data to identify a cohort of patients with heart failure at the time of hospitalization and compared “looking forward” from that event with “looking back” in the subset of patients who died. The investigators found a correlation between hospitals with higher resource use and lower mortality. However, another study³⁷ found that resource use does not consistently correlate with outcomes at the end of life and that “looking forward” and “looking back” yield similar results.

Our study has some limitations. Although we describe costs to the Medicare program, we could not assess out-of-pocket costs borne by patients. The SUPPORT investigators²⁷ found a significant financial impact of caring for patients with advanced heart failure; nearly one-quarter of patients' families reported the loss of most or all of their savings at the time of the patient's death. A recent analysis³⁸ found that Medicare beneficiaries are spending more of their income on health care costs over time—12.8% in 2000 vs 16.1% in 2005. Moreover, our estimates do not include costs to Medicaid or private insurance. Patients who remained in skilled nursing facilities after exhausting the Medicare post–acute care benefit paid privately for those services or became Medicaid beneficiaries. If they died in a skilled nursing facility, we did not capture that information. Third, we relied on ICD-9-CM codes and a validated algorithm to identify the study population, but we did not have access to medical charts. Previous studies^18,39,40 have suggested that a single inpatient heart failure diagnosis code has 95% specificity for the diagnosis of heart failure. Fourth, although all of the patients had a diagnosis of heart failure, the cohort was clinically heterogeneous, the burden of comorbid disease was high, and the costs and resources were not necessarily specific to heart failure. Fifth, we did not have access to important clinical data, such as functional status, which would inform discussions on appropriateness of care. Finally, claims data are available for fee-for-service Medicare beneficiaries only, so the results may not generalize to patients enrolled in Medicare managed care.

In conclusion, in this longitudinal analysis of resource use in the last 6 months of life among Medicare beneficiaries with heart failure, we found that most patients frequently accessed the health care system and spent some time in the hospital. The use of health services at the end of life in this population has increased over time with higher rates of intensive care and higher costs. However, the use of hospice services has also increased markedly, representing a substantial shift in patterns of care at the end of life.

Correspondence: Lesley H. Curtis, PhD, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27715 (lesley.curtis@duke.edu)

Accepted for Publication: April 26, 2010.

Published Online: October 11, 2010. doi:10.1001/archinternmed.2010.371

Author Contributions: Dr Curtis had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Unroe, Greiner, Kaul, and Peterson. Acquisition of data: Unroe, Greiner, and Curtis. Analysis and interpretation of data: Unroe, Greiner, Hernandez, Whellan, Schulman, and Curtis. Drafting of the manuscript: Unroe and Greiner. Critical revision of the manuscript for important intellectual content: Unroe, Greiner, Hernandez, Whellan, Kaul, Schulman, Peterson, and Curtis. Statistical analysis: Greiner. Administrative, technical, and material support: Unroe and Schulman. Study supervision: Unroe, Schulman, and Curtis.

Financial Disclosure: Dr Hernandez has been a consultant for AstraZeneca, Corthera Inc, Eli Lilly & Co, Medtronic Inc, Novartis Pharmaceutical Co, and Thoratec Corp and has received research grants or contracts from GlaxoSmithKline, Johnson & Johnson Pharmaceutical Research and Development, Medtronic Inc, Merck Group, Proventys, and Scios Inc. Dr Schulman has been a consultant for Advanced Health Media LLC, Alnylam Pharmaceuticals Inc, the American College of Medical Genetics, the American Society of Clinical Oncology, Amgen Inc, and Cancer Consultants Inc, Certified Medical Representatives Institute Inc, East Carolina University, Faculty Connection LLC, GlaxoSmithKline, IBM Corp, Integrated Therapeutics Group Inc, Johnson & Johnson Inc, Lifemasters Supported SelfCare Inc, McKinsey & Co Inc, Medical Decisions Network, Medtronic Inc, National Pharmaceutical Council, Novartis, Novo Nordisk, Pfizer Inc, Quintiles Transnational Corp, Rutgers State University of New Jersey, Social & Scientific Systems Inc, Wallen and Associates, and William Beaumont Hospital (Royal Oak, Michigan) and has received research grants or contracts from Actelion Pharmaceutical Ltd, Allergan Inc, Amgen Inc, Arthritis Foundation, Astellas Pharma Inc, Bristol-Myers-Squibb, the Duke Endowment, Genentech Inc, Inspire Pharmaceuticals, Johnson & Johnson Inc, Kureha Corp, Medtronic Inc, Merck & Co Inc, Nabi Pharmaceuticals, National Patient Advocate Foundation, National Patient Advocate Foundation, NovaCardia Inc, Novartis, OSI Eyetech, Sanofi-Aventis US LLC, Scios Inc, Tengion Inc, Theravance Inc, Thomson Healthcare, and Vertex Pharmaceuticals Inc. Dr Peterson has been a consultant for Tethys, and has received research grants or contracts from the American College of Cardiologists, the American Heart Association, Bristol-Myers-Squibb/Sanofi-Aventis, Merck/Schering Plough, Eli Lilly & Co, Johnson & Johnson Inc, and the Society of Thoracic Surgeons. Dr Curtis has been a consultant for Pfizer Inc and has received research grants or contracts from Allergan Inc, Eli Lilly & Co, GlaxoSmithKline PLC, Johnson & Johnson Inc, Medtronic Inc, Merck Co Inc, Novartis, OSI Eyetech, Pfizer Inc, and Sanofi-Aventis US LLC.

Additional Contributions: Damon M. Seils, MA, Duke University, provided editorial assistance and prepared the manuscript. Mr Seils did not receive compensation for his assistance apart from his employment at the institution where the study was conducted.