Promoting Excellence : Cost Accounting Recommendations to the Field : Issues in Designing the Cost Evaluation

M.C. Escher's The following list of cost measurement issues reflects the unique challenges that arise when evaluating characteristics of end-of-life care and the innovative programs developed to provide that care. The list first addresses issues in designing the cost evaluation and then issues in computing costs. Workgroup recommendations follow each topic discussed.

In developing this set of concerns, Workgroup members focused primarily on resources and economic costs within different health care delivery systems that have some initial data available for developing cost accounting practice standards (direct health system and program costs). This approach does not take into account social cost trade-offs or indirect costs, including economic losses to family caregivers and the societal costs of lost productivity of family caregivers.

Plan Early and Fit Accounting Methods to Programmatic Interventions.

Most palliative and end-of-life care programs are tightly budgeted and many are demonstrations that may be funded for a limited period of time. Consequently, cost evaluation may not be feasible if it is added as an afterthought when the program is well underway or a demonstration project is nearly concluded. Even when evaluations are planned early in a project, the analysis proposed may be too ambitious to complete with the time and resources available.

Often clinical researchers do not involve economists in designing the program evaluation, or may interact too infrequently with clinical leaders and staff to understand the population served or the services delivered. Resulting inadequacies in the data gathered and its analysis may fail to produce meaningful findings about an intervention.


Plan the sample size.
It is important to plan the sample size in each arm of the study. Many palliative and end-of-life care demonstration programs do not serve enough patients to permit statistical comparisons of outcomes with patients in the control or comparison group. The number of "evaluable" patients in programs with larger enrollments can shrink quickly, sharply reducing statistical power to detect program effects on key outcome variables.

The sample size may be affected by: 1) the number of patients enrolled in the program or assigned to the control or comparison group; 2) the number of these patients who die; 3) time of death relative to study entry; 4) the period of time during which patients are accrued; 5) the anticipated cost difference associated with the program; and 6) the distribution of costs in the study and control or comparison groups.


Identify the perspective for cost analysis.

The costs of care during the last phase of life can be assessed from different perspectives. The fact that a cost evaluation is being conducted from one particular perspective does not make it "wrong." However, the language and interests of the stakeholders being served will affect the questions addressed, the data collected, the analyses completed and the conclusions reached. If the perspective for a cost analysis is not clearly defined, evaluators may include cost elements that are not relevant to the affected party, and other stakeholders may misinterpret results or apply them inappropriately.

Examples of different perspectives include: perspectives of patients and their families (first party); physicians, hospitals, community health programs and other service providers of a managed care plan (second party); private insurers, Medicare, Medicaid or other payers (third party); employers or other plan sponsors (fourth party); or society as a whole. The perspective used for defining costs affects how costs are measured in practice.

Patients and their family members might define costs in terms of personal, out-of-pocket expenses for health insurance coverage and co-payments, the more intangible costs of purchasing or not purchasing uncovered services or the costs of unpaid caregiving. Agencies providing care may define costs in terms of billed services or a negotiated per diem rate, while third party public and private insurers may consider only those costs that they reimburse or the costs of capitated per member, per month payments. Policy-makers may focus on the costs of funding a program versus alternative uses of these resources. Health plan administrators, policy-makers and advocates need data on the costs of delivering a program to beneficiaries in various markets, ideally from a cost perspective of providing services on an individual basis and for all individuals within a group. If a specialized palliative or end-of-life care program is included in a bundle of services provided as an employee benefit, employers are concerned about the costs of coverage as well as potential savings from reductions in employee absenteeism and turnover related to family caregiving needs.

Evaluating program costs from a social perspective involves comprehensively assessing the costs of all health care resources utilized regardless of source, as well as opportunity costs and uncompensated costs of family caregiving. Costs of unpaid caregiving, both to employers and to family members and friends, are becoming increasingly important in the policy arena. Caregiver costs include not only unpaid time and effort devoted to a loved one's care, but also commonly include "opportunity costs" of foregone earnings, loss of health insurance, life insurance and retirement benefits. Associated caregiver stress can result in increases in morbidity (and mortality) and costs associated with health care utilization by caregivers. The data required for such analysis are neither readily available nor easily collected, but such a full analysis is important in determining whether a program that appears to reduce costs for one stakeholder has merely shifted them to others.

Predictably, the accounting perspective of a cost analysis derives from the intended use or purpose of the study. An agency such as a hospice or palliative care organization that provides specialized care to dying patients likely will be interested in the costs of delivering its services compared to projected revenues. In assessing the overall financial impact of a new program (to determine long-term sustainability of the program and decide whether or not to maintain, expand or discontinue specific component services and staffing levels) the program's impact on total health services utilized and on system efficiency may be considered. When two or more agencies have collaborated in the development and delivery of a palliative or end-of-life care program, partners are likely to be more interested in costs to their particular organization than in an analysis of total costs. Results may indicate the need for an adjustment in resource contributions or renegotiation of how capitated health plan payments are apportioned in managed care risk-sharing arrangements.


Clearly describe the palliative or end-of-life care program being studied.

Cost analyses often devote minimal attention to describing the health service delivery programs whose costs and financial outcomes are being assessed. However, because palliative and end-of-life care programs differ in structure, staffing, component services, admission criteria, and the populations they serve, the costs and outcomes of a specific programmatic model of care may not accurately represent those of another. For example, palliative and end-of-life care programs based in hospitals, skilled nursing facilities, outpatient settings and prisons are likely to incur different costs because they serve different populations, provide different packages of services and deliver them in different ways. Similarly, the design, requisite costs and financial outcomes of a palliative and end-of-life care program in a fee-for-service environment are likely to vary from one based in an environment of prepaid (and capitated) health care.

Hospice is a defined programmatic model for delivering palliative care at the end of life. Yet even hospices differ in the populations they serve, as well as in the range of services and quality of care they provide. An analysis of costs incurred in delivering the Medicare Hospice Benefit may not apply to estimating the costs of caring for adults under age 65 or of dying children and their families. Hospice costs and financial outcomes may not apply to programs that provide concurrent palliative care and life-prolonging treatments. Different interpretations and meanings ascribed to the terms "palliative care" and "hospice" further complicate comparisons of programs and their outcomes.

Typical Features of Palliative Care in Promoting Excellence Projects

  1. Ongoing communication among patients, families and providers
  2. Advanced care planning and patient-centered decision making that is iterative and reflective of patients' values and preferences
  3. Formal assessment and treatment of physical and psychosocial symptoms
  4. Care coordination (also known as case management) to streamline access to services and monitor quality care
  5. Spiritual care
  6. Anticipatory guidance in coping with illness and issues of life completion and life closure
  7. Crisis prevention and early crisis management
  8. Bereavement support
  9. An interdisciplinary team approach to care
  10. 24/7 availability of a clinician knowledgeable about the case

Unless a cost analysis clearly describes salient features of the clinical and health service delivery program being evaluated, results may be inappropriately generalized or used in misleading comparisons. Nevertheless, experts have different opinions about how programs should be described. Some researchers advocate identifying each component of care to facilitate the creation of different service "packages" tailored to the needs of sub-groups of patients and the realities of care delivery in different settings. Other authors fear that if program services are unbundled, Medicare and other payers will analyze the itemized costs of each component of care and decline to pay for some components. These may include "low-tech," psychosocial services that are central to quality palliative and end-of-life care programs, but more difficult to directly attribute to measurable impacts on health services, pharmaceuticals or durable medical use.


Describe at least the following features of the palliative and end-of-life care program being evaluated:

Identify appropriate comparison groups.

Although double-blinded randomized controlled trials (RCT) are the gold standard for clinical and health service delivery research, the RCT design may not be feasible or ethically appropriate for some palliative and end-of-life care programmatic settings.

Health plan and agency administrators, physicians and others may not agree to randomization because they are committed to providing equal access for all patients eligible for a program or service. Patients and their family members also may refuse randomization either because they want - or do not want - to be served by a palliative and end-of-life care program. Even if these obstacles can be overcome to accomplish randomization, participants in the evaluation are likely to know whether they are assigned to the treatment or control condition.

When randomization is not possible, it is important, although challenging, to define an appropriate, comparable group of patients who are not enrolled or otherwise receiving the palliative and end-of-life services. Groups of enrolled and not enrolled patients may differ in diagnosis, illness severity, acceptance of approaching death, caregiver resources and many other characteristics that affect utilization and costs of health services. A common practice is to use historical controls, e.g., "We're spending 10 percent less this year than last year." However, such analyses can be misleading because both the mix of patients served and the costs of care may change over time. Further, a Hawthorne Effect, the phenomenon of something changing simply because it is being observed, may result if prior year data collection occurs without knowledge that a program will be evaluated.


Define the period of care to be studied.

In clinical research, estimates of the time required for patients being studied to respond to their treatment usually define the period for assessing related costs. However, differences in the trajectory of disease and the uncertainty of death complicate definition of the time period for cost analyses of palliative and end-of-life care programs. No consensus has been reached on when "the end of life" begins. From a purely scientific perspective there are advantages to retrospectively comparing all costs during a crisply defined period, such as the last year of life. Yet health service delivery must function in real time and respond to the clinical needs of patients and their families. Patients' functional status, disease trajectory, experienced needs and health care choices are highly individual. It is, therefore, not possible to prospectively determine when the last year of life begins.

Within the category of palliative and end-of-life programs there is significant variability of populations being served and programmatic settings, services and interventions. Some programs are designed and best suited to hospital and ICU settings and focus on the last weeks, days or even hours of life. Others are designed for assisted-living, nursing home or private residential settings and focus on the last months or years of life.

Within a single program, patients do not enter it at the same interval preceding death. The reasons are many. Prognosis is inherently difficult and uncertain, particularly for conditions other than solid tumors. Physicians may be reluctant to refer patients to palliative care or hospice programs when admission criteria are based on prognosis. Denial of the approach of death by physicians, patients and family members also may deter both the timeliness of referrals and their acceptance.

While in the abstract, measuring program costs for an arbitrary time period might appear to be the logical solution, in actuality the length of stay in palliative and end-of-life care programs tends to be highly variable. Confounding factors include the fact that some patients may die too soon after admission for the intervention to have had a significant effect, and others may die before a given study's end-point is reached. Alternately, some patients will live beyond the designated evaluation end-point so that the full impact of the program and its interventions of the costs of their care at the end of life cannot be captured. (Yet it is possible that their longevity may indicate that these patients have been helped the most by the program's services.) Patients also may be discharged because they become medically stable, move or change their health insurance. Simple comparisons of costs incurred by patients in the program and control or comparison group during an arbitrary time period are likely to be flawed by uneven exposure to the intervention in the program condition, differences in patient survival and missing data in both the treatment and control or comparison arms of the study.

Additional accounting challenges arise from the knowledge that resource consumption is not spread evenly over whatever time frame is selected for the cost analysis. Palliative care is often more intense during the inception of specialized care and during the final days of life. Evaluation should include both of these periods. Although costs of admission to a program tend to be relatively fixed, they rise to represent a disproportionate amount of total costs when length of service is very brief. Hospice experience demonstrates that this situation can severely challenge providers who are reimbursed at fixed per diem rates. From a methodological perspective, if many patients are enrolled in a program only briefly the calculated average costs of service per patient or per unit of time may be higher than actually experienced for patients served for a longer period. Moreover, because intake costs are not incurred for patients in the control or comparison condition, a specialized palliative and end-of-life program that serves patients for only brief periods may appear to cost more than usual care, not because of faulty program design, but because of insufficient time to realize potential cost advantages.

A crescendo of pain and other distress during the last days of life also requires an intensification of services. Costs therefore are greater for patients who die in the program than for those who do not. Transfers may occur to improve quality of care, but they also represent intentional or unintentional cost shifting. For example, nursing homes and oncology clinics may send dying patients to the hospital or to hospice, and some "upstream" palliative and end-of-life care programs identify transfer to hospice as their desired outcome.


The M.C. Escher piece at the top of this page is titled "Path of Life."

<<< Previous Next >>> [ Go Up ]

Promoting Excellence in End-of-Life Care was a national program of the Robert Wood Johnson Foundation dedicated to long-term changes in health care institutions to substantially improve care for dying people and their families. Visit for more resources.

Promoting Excellence logo and link to home page