Measuring Costs for Cost-Effectiveness Analysis
Guidelines for cost-effectiveness analysis specify that health services be assigned the opportunity cost based on a long-term, societal perspective. Cost-effectiveness analysts usually use reimbursements (the amount that the sponsor paid the provider) as a proxy for this opportunity cost. When the billed charge for a hospital stay is available, the cost-adjusted charge may be a better source, as it reflects variation in resource use that do not affect reimbursement. The cost adjusted charge is the amount billed for the hospital stay multiplied by the hospital wide ratio of cost-to-charges.
Using reimbursements as a proxy for costs is not possible for pharmaceuticals or newly developed interventions. There is no published schedule of reimbursements for individual drugs, and the analyst must use a drug price database, such as the Federal Supply Schedule or a private data source, and adjust to the price to reflect the best estimate of the cost to the sponsor, net of reimbursements.
Newly developed interventions are not found on reimbursement schedules, and the analyst must directly estimate their cost.
To find the cost of an intervention the analyst should include the cost of all activities needed to replicate the intervention in a typical healthcare setting. Costs incurred only to study the intervention should be excluded. When an activity involves both delivery of the intervention and research on its effect, the cost of any activity needed to deliver the intervention is included.
For example, consider the cost of a follow-up telephone call. The study participant is asked to return to a clinic to receive more intervention and to fill out a research assessment. The call is a cost of intervention. In order to replicate the intervention in the real world, the follow-up call will still be needed so that the patient will return to clinic to receive more intervention. A strict accounting of intervention cost would exclude any extra cost that was exclusively attributable to research--for example, any extra minutes spent describing the research assessment. This extra cost would not be needed to replicate the intervention in the real world. Another example is a laboratory test conducted to identify patients who are eligible for the intervention being studied. The test is a cost of the intervention because it would be needed to replicate the intervention elsewhere with the same level of effectiveness.
Staff cost should be fully burdened with the cost of benefits, employer contributions to taxes, and non-productive time such as vacation and sick leave. This can be done in the calculation of the hourly cost of staff time. Total staff cost is divided by the number of applied (productive) hours, the time spent on activities that involve patient care. Hours on overhead activities such as vacations, sick leave, and professional training are excluded from the count of applied hours. For more details on measuring the cost of staff time, view Measuring Staff Activities.
When the cost is found in this manner, it represents the hourly cost of a worker engaged in productive activities. This method distributes the cost of vacation and sick leave across the productive hours of the employee. The analyst should consider if some of the cost of administrative duties, phone calls, and other activities should also accrue to the intervention; if so, they may be excluded from applied hours, and the cost distributed using this same method.
There may be costs arising from subcontracts with outside firms. Contract costs should be included if they relate to the intervention. Beyond the stated value of a contract, there will also be direct non-medical costs relating to the bidding process and contractor oversight. These direct non-medical costs should also be included. Clinical studies may involve a more intensive level of patient assessment than would occur under usual circumstances. For example, physicians may order more tests in a clinical study in order to detail patient outcomes as fully as possible in the final report. By contrast, in general practice there is typically pressure to minimize costs by performing only those tests that are medically indicated. A knowledgeable clinician can determine whether the intervention is being carried out differently from how it would occur in typical practice settings. If so, a discussion of cost-effectiveness could present additional figures for the cost of the intervention under typical circumstances.
Economic analyses often use data from two or more calendar years. Price inflation causes the value of a dollar to fall over time, and so the same dollar amount in two different years will usually represent different amounts of purchasing power. To counteract this problem, analysts typically adjust dollar figures to account for inflation. Figures that have not been adjusted for inflation are said to be in 'nominal dollars,' while those that have been adjusted are in 'real dollars.' Here we describe how to adjust for inflation so that dollar values are expressed in terms of a single year's currency.
Inflation adjustments are made using price indices. Each index consists of numbers representing the price level in each year relative to a base year. Some indices have values that correspond to shorter periods as well, such as months or quarters. What distinguishes the indices is how the price levels are established.
To correct for inflation the analyst selects a base year. The goal is to adjust all dollar figures so that they are expressed in terms of dollars in that year. Often the base year is chosen to be the current year or the final year of study data. For example, suppose that you adjust for inflation using the Consumer Price Index (CPI) for all urban consumers. To express a cost from calendar year (CY) 2000 into CY 2005 dollars, simply multiply the cost by the CY 2005 index and divide by the CY2000 index. The relevant index values are 113.4 for CY 2005 and 100.0 for CY 2000. If the cost were $20 in CY 2000, this would be the calculation:
$20.00 x (113.4/100.0) = $20.00 x 1.1342 = $22.68
Converting costs to 'real dollars' allows us to compare costs incurred in different years. For example, which is more expensive in CY 2005 dollars: A, which costs $20 in CY 1995, or item B, which costs $23 in CY 2000? Using the CPI for all urban consumers, item B is more expensive in real terms: it costs $26.09 in CY 2005 dollars, whereas item A costs only $25.63 in CY 2005 dollars.
One can convert costs to any year for which a price index exists. The base year does not affect which good is more expensive. If A is more costly than B in one base year, it will be more costly in terms of any other base year.
At HERC we often use the U.S. Consumer Price Index (CPI) for all urban consumers. Values for this index tell what a market basket of consumer goods that cost $100 in CY 1983 would cost in the year in question. It quantifies the erosion of purchasing power by inflation. If most of the costs you are considering derive from staff, as is often the case for health care, then the general CPI is appropriate. The chain-weighted CPI is likely to be more accurate than the standard CPI, but it has only been figured since CY 2000.
We do not recommend using the medical care components of the Consumer Price Index (CPI). It analyzes changes in the cost of providing a day of stay and an outpatient visit. In recent years fewer but more expensive days of stay and visits are needed to treat an illness. This change is captured by the index without considering the change in productivity, overstating the increase in cost. Other faults of the medical care CPI are its reliance on list prices and the weighting of component medical care goods and services based on consumers out-of-pocket costs rather than overall health care expenditures. Berndt et al. (2000) provide a more thorough discussion of shortcomings in the medical care CPI and propose a fundamental reform of it. Additional information about CPI and its medical components can be found at the Bureau of Labor Statistics website.
The Second Panel on Cost-Effectiveness in Health and Medicine proposed that the Personal Health Care (PHC) Expenditure deflator (developed by the Centers for Medicare and Medicaid Services) could be used to adjust for disease-specific costs. The PHC Expenditure deflator uses chain-weighted price index calculations that take into consideration the various goods and services associated with Personal Health Care services. A comparison of the different methods used by federal agencies to adjust for inflation was performed by Dunn et al (2015). Additional comparisons can be found at the Medical Expenditure Panel Survey website.
The US Department of Labor Bureau of Labor Statistics and US Department of Commerce Bureau of Economic Analysis posts tables of index values for three common inflation indices used for health care research: the Consumer Price Index (CPI) for all urban consumers; the chain-weighted CPI for all urban consumers; and the Gross Domestic Product (GDP) implicit price deflator. Additionally, the Agency for Healthcare Research and Quality lists tables of index values for the Gross Domestic Product (GDP), Personal Consumption Expenditures (PCE), and Personal Health Care (PHC) Expenditures on their website.
The Chain-Weighted All-Item CPI differs from the standard All-Item CPI in that the weights assigned to goods in the index change over time. The percentage figures reveal that the chain-weighted index yields slightly lower inflation rates than the standard CPI in the period 2000-2009. The GDP Implicit Deflator, also a common inflation measure, yields an inflation rate similar to those of the two CPI indices but not consistently higher or lower than either of them.
What About Discounting?
If you are comparing two interventions, each involving a series of expenditures over time, you need to consider the time value of money (the fact that a dollar spent today is a bigger expense than a dollar spent a year from now). This requires application of a discount rate. The Public Health Service Panel on Cost-Effectiveness in Medicine recommends a discount rate of 3% (Lipscomb et al., 1996), which the Second Panel on Cost-Effectiveness in Health and Medicine continues to support (Basu and Ganiats, 2017). To find the incremental cost-effectiveness ratio, the discount rate should be applied to both real costs and to outcomes measured as quality-adjusted life years.
Discounting should not be confused with adjusting for inflation. Both are needed. The inflation adjustment reflects the change in purchasing power of currency (e.g., adjusting past costs to curren values). Discounting reflects the loss in value when there is a delay in obtaining an item of value (e.g., adjusting for future value). We discount expenses and health outcomes if there is a delay in realizing them.
Discounting reflects the loss in economic value that occurs when there is a delay in realizing a benefit or incurring a cost. Cost-effectiveness analysis incorporates the economic fact that costs and benefits that are deferred have lower value than those that are realized immediately.
Discounting should not be confused with adjustment for inflation. All costs should be expressed in real terms (adjusted for inflation) before discounting is done.
Both cost and outcomes should be discounted. Failure to discount outcomes as well as costs can result in a paradox described by Keeler and Cretin (1983). If costs are discounted, and outcomes are not, the cost-effectiveness of an intervention can always be improved by delaying its implementation indefinitely.
Costs and outcomes enter into the cost-effectiveness analysis expressed in their present discounted value. Most analysts discount on an annual basis. Expenses incurred in the first year are not discounted. If a discount rate of 3% is chosen, then expenses incurred in the second year are discounted by 3%, that is, they are divided 1.03. Third year expenses are divided by (1.03)2. Each successive year is discounted by an additional 3%.
The present value (PV) of costs incurred by a subject from the first year of the study (t=1) until the study ends (t=n) is thus:
Present Value of Cost
Note that Ct represents the cost incurred in year t, and that i is the discount rate, e.g., .03. The present value of outcomes is calculated in the same manner.
What is the Appropriate Rate of Discounting?
A discount rate of 3% was recommended by the Public Health Service Panel on Cost-Effectiveness in Medicine (Lipscomb, et al, 1996). The Panel recommended that an alternative analysis be done with a 5% discount rate, so that results are comparable to those studies that use this higher rate.
Simplification for Costs and Benefits Realized at a Constant Rate.
When either cost or outcomes are realized at a constant rate, a formula may be used to find their discounted present value.* It can be used for models that assume that annual healthcare cost are incurred at a constant rate. It can also be used to find discounted life-years of remaining survival. The formula incorporates a discount factor (r):
which includes the discount rate (i). Other variables in the formula are the annual rate at which costs are incurred (a) and the number of years over which they will be incurred (n).
Present Value of Cost
The formula can also be applied to outcomes. If it is assumed the quality of life of the remaining years of survival is constant, then (a) represents the quality of life adjustment. If the outcomes are measured in life years, this factor takes a value of one, and drops out of the equation.
* This is the formula for a finite geometric series. A finite geometric series consists of a series of terms. The ratio of each term to its predecessor is a fixed constant (r):
The value of a geometric series (Sn) may be determined by subtracting r(Sn) from both sides of the equation, and applying simple algebra to yield:
If you are comparing two interventions, each involving a series of expenditures over time, you need to consider the time value of money (the fact that a dollar spent today is a bigger expense than a dollar spent a year from now). Benefits and cost should be discounted at the same rate. The Public Health Service Panel on Cost-Effectiveness in Medicine recommends a discount rate of 3%. View the section, "Discounting Healthcare Costs".
Benefits must be discounted for the same reason as costs are discounted. A benefit realized today has greater value than a benefit realized a year from today. Failure to discount benefits, or to discount costs and benefits at a lower rate, results in the Keeler-Cretin paradox.
Here is a simplified version of the Keeler-Cretin argument. Consider an intervention that costs $100,000 and saves 10 lives. If you chose to wait, and fund the intervention 10 years from now, the cost would be discounted, that is, the net present value would be less than $100,000, but it would still save 10 lives. The "waiting" strategy would be more cost-effective. If you waited 20 years, the intervention would even be less costly, and would still save 10 lives. You would always prefer to defer to the future, and you'd never do the intervention. The paradox is described more fully in the Gold, 1996.
There are controversial questions in discounting health consequences (e.g., QALYs). Research analysts interested in this topic should consult the articles listed in the References section.
Basu A, Ganiats TG. Discounting in Cost-Effectiveness Analysis. In: Neumann PJ, Sanders GD, Russel LB, Siegel JE, Ganiats TG, editors. Cost-Effectiveness in Health and Medicine, Second Edition. Oxford University Press; New York: 2017. pp. 277-288.
Berndt ER., Cutler DM., Frank RG., Newhouse JE, Triplett JE. Medical Care Prices and Output. In: Culyer A, Newhouse JE, editors. Handbook of Health Economics. Elsevier; Amsterdam: 2000. pp. 119–180.
Cairns J. Discounting and health benefits: another perspective. Health Econ.1992;1:76-9.
Dunn A, Grosse SD, Zuvekas SH. Adjusting Health Expenditures for Inflation: A Review of Measures for Health Services Research in the United States. Health Serv Res. 2018 Feb;53(1):175-196.
Gyrd-Hansen D, Sogaard J. Discounting life-years: whither time preference? Health Econ.1998;7:121-7.
Keeler EB, Cretin S. Discounting of life-saving and other non-monetary effects. Management Science.1983;29:300-306.
Lipscomb J, Weinstein MC, Torrance GW. Time preference. In Cost-Effectiveness in Health and Medicine, M Gold, J Siegel, LB Russell, and MC Weinstein, eds. New York: Oxford, 1996.
Owens, D. K. Interpretation of cost-effectiveness analyses [Editorial]. J Gen Intern Med.1988;13:716-717.
Parsonage M, Neuburger H. Discounting and health benefits. Health Econ. 1992;1:71-6.
Smith DH, Gravelle H. The practice of discounting in economic evaluations of healthcare interventions. Int J Technol Assess Healthcare. 2001;17:236-43.
van Hout BA. Discounting costs and effects: a reconsideration. Health Econ. 1998;7(7):581-594
Last updated: April 20, 2021