Cost-effectiveness analysis (CEA) combines information about the costs (Chapter 4) and
outcomes (Chapter 3) of a community prevention program to produce information that can
be used to answer questions about whether a program is cost-effective, whether program
expansions would be cost-effective, and/or whether a program is more or less cost-effective
than alternative prevention strategies.
CEA evaluates the costs required to yield a specific nonmonetary outcome, such as the cost
per life-year gained, the cost per asthma-related emergency room (ER) visit averted, or the
cost per inch lost from the waist. In this chapter, we begin by providing some advice for
selecting program outcomes for use in CEA. We then describe how to calculate the
necessary cost-effectiveness ratio(s) (CER) to answer your study questions and how those
CERs can be used to allocate resources across prevention strategies.
5.1 Selecting Outcome Measures for Cost-Effectiveness Analysis
Although the selection of outcome measures for evaluating program success was discussed
in Chapter 3, in this subsection we discuss some issues to consider when selecting outcome
measures specifically for use in CEA. The effectiveness of a community prevention program
can be assessed in terms of both immediate and longer-term outcomes (see Chapter 3 for
additional discussion on measuring program effectiveness). For example, in the short-term,
a program that encourages physical activity through financial rewards may lead to increases
in the percentage of participants who achieve the recommended level of physical activity
per week. In the long-term, if improvements in physical activity levels are sustained, the
program may lead to reductions in chronic disease incidence, such as diabetes, stroke, heart
disease, and cancer. Such improvements in health may eventually yield longer and better
lives for program participants—outcomes that can be quantified as life-years gained or
quality-adjusted life years (QALYs) gained. All of these outcomes—increases in physical
activity, chronic disease cases averted, and improvements in life-years gained or QALYs
gained—are legitimate outcomes for use in CEA. The decision about which of these program
outcomes to use in CEA should be made by considering which outcome measure best
answers the study question, what outcomes are most easily comprehended by the target
audience for the cost-effectiveness (CE) study, and whether data are available to link shortterm
program outcomes to longer-term changes in health and mortality.
Although final health outcomes, such as the number of strokes averted or life years gained,
are generally recommended for use in CEA (see Haddix, Teutsch, and Corso, 2003, and
Drummond et al., 1997), intermediate outcomes may be appropriate when evaluating many
types of community prevention efforts, especially if these outcomes are more easily
understood by the target audience for CE studies. For example, because data are not readily
available to link a physical activity intervention to health outcomes that would be realized
Guide to Analyzing the Cost-Effectiveness of Community Public Health Prevention Approaches
5-2
decades in the future, it may be appropriate to use an intermediate program outcome, such
as changes in physical activity levels among participants. Drummond et al. (1997)
recommend that when an intermediate outcome is selected, it should either be viewed as
valuable in its own right (e.g., increased physical activity or vaccination against disease) or
have a well-established link to a desirable health outcome (e.g., link between increased
physical activity and reductions in chronic disease or link between vaccination and reduced
incidence of vaccine-preventable disease).
To help you decide on the most appropriate outcome measure(s) in CEA of your prevention
program, we provide hypothetical program examples and their corresponding outcome
measures in Table 5-1. Information about the program or study factors that influenced the
outcome measure decision is also provided.
Table 5-1. Hypothetical Prevention Programs and Outcomes for CEA
Program Description Outcome(s) for CEA Issues in Outcome Selection
Worksite program to
increase physical activity
Increased number of
employees meeting physical
activity recommendations
Program comparison is to other
worksite approaches to increase
physical activity (vs. to prevention
strategies with noncomparable
outcomes)
School-based asthma
management program
Asthma-related acute care
visits averted and QALYs
gained
Program impact on health and wellbeing
can be estimated in the shortterm
(vs. many years in the future)
City policy to create
additional sidewalk space
Increased number of walking
trips per month
Program comparison is to other
approaches to increase walking/
biking trips
Difficulty of linking increased
walking trips to future health
improvements
Screening and
intervention program to
identify and reduce
cardiovascular disease
risks
Number of participants with
reduced blood pressure
Change in systolic or diastolic
blood pressure
Reduction in the average 10-
year probability of
cardiovascular disease
Multiple program outcomes needed
to be summarized in a single CE
measure for policy makers
Program comparison is to other
approaches to reduce cardiovascular
disease risks
No single right answer exists to the question of which outcome measure to use in CEAs of
prevention programs. In fact, you may choose to use multiple outcome measures to address
the interests of diverse stakeholders. For example, although a company’s wellness
committee chairperson may be interested in the cost per additional employee achieving
weekly physical activity recommendations, the human resources vice president may be
more interested in the cost per missed work day averted (or related measure of change in
Chapter 5 — Conducting a Cost-Effectiveness Study
5-3
employee productivity). To satisfy the interests of both stakeholders, CERs should be
calculated using both types of program outcome measures, if possible. Results would then
be reported as the program cost per additional employee achieving physical activity
recommendations (= total program costs/change in number of employees meeting physical
activity recommendations among targeted employees) and as the program cost per missed
work day averted (= total program costs/change in number of work days missed among
targeted employees).
5.1.1 Quality-Adjusted Life Years and Their Measurement
If program stakeholders are interested in being able to make cost-effectiveness
comparisons across prevention programs with disparate outcomes, then a common outcome
measure, such as life-years saved or quality adjusted life-years (QALYs) gained, will need to
be used for each program under consideration. For example, when comparing a program
that seeks to increase physical activity to a program that seeks to reduce asthma-related
health care visits and costs, the measured program outcomes—increased minutes of weekly
physical activity and number of asthma-related ER visits averted—will need to be converted
to a common measure so that CERs can easily be understood and compared across the two
programs. Both programs may be viewed as worth the investment if, for example, the first
has a cost of $500 per additional 30 minutes of weekly physical activity and the second has
a cost of $350 per asthma-related ER visit averted. However, without estimating a common
outcome measure for both programs, questions about which program is more cost-effective
cannot be answered.
QALYs have been recommended for use in CEA to improve the comparability of results
across CE studies (Gold et al., 1996). QALY measures are generally preferred to measures
of life-years gained because the QALY measure captures gains both from increased life and
from reduced morbidity. To use QALYs as the outcome measure of interest, it is necessary
to first collect data on short-term or intermediate program outcomes, such as changes in
physical activity or ER use among asthma patients. These outcomes are then linked to
information from the literature or other data sources to estimate the future health outcomes
likely to result from participation in the prevention program, such as a reduced likelihood of
diabetes, stroke, or death. Finally, health outcomes are then converted to QALYs using
estimates of people’s preferences for being in various health states ranging from excellent
health (valued at 1) to death (valued at zero). Dasbach and Teutsch (2003) discuss the
estimation of QALYs and possible sources in the literature for valuing the quality of life in
various health states, such as having diabetes, asthma, hypertension, and/or other specific
health conditions. Drummond et al. (1997) provide a detailed treatment.
Guide to Analyzing the Cost-Effectiveness of Community Public Health Prevention Approaches
5-4
5.2 Analyzing the Cost-Effectiveness of a Prevention Program
After program costs have been analyzed and decisions have been made about which
outcomes to use as measures of program effectiveness, CERs can be calculated to answer
the key study questions.
5.2.1 Average Cost-Effectiveness
Average CERs are useful for considering the cost per additional outcome achieved by a
program as compared to a baseline of doing “nothing.” An average CER is calculated by
dividing program costs by the change in outcomes generated by the program:
Outcome
Cost
Ave CER
Δ
_ = (5.1)
In a recent study on the cost-effectiveness of different vaccination strategies in hospitals,
average CERs were calculated for each hospital in the study (Honeycutt et al., 2006). CERs
were calculated by dividing total vaccination program costs (estimated by the study
authors) in each hospital by the number of patients vaccinated (data collected from each
hospital). The cost-effectiveness of the study’s hospital vaccination programs ranged from
$22 to $362 for each additional patient vaccinated.
If the outcomes under consideration are health outcomes (e.g., cases of diabetes averted),
as opposed to short-term or intermediate program outcomes (e.g., increased minutes of
physical activity), then the measure of cost used in the numerator of the CER should be net
of disease costs and productivity losses averted by the program. Estimating the societal cost
of each program in this manner helps to ensure the comparability of CERs across alternative
programs. The societal cost of a program that produces measurable health benefits is the
cost of the program less any cost savings that can be attributed to the program. Gift,
Haddix, and Corso (2003) provide details on calculating CERs when health outcomes are
used as measures of program effectiveness. For simplicity, our treatment simply uses the
term “costs” to represent program costs. Those costs should be net of disease costs and
productivity losses averted if health outcomes are used in the analysis.
A program is often considered to be cost-effective if its CER is below the commonly used
threshold of $50,000 per life-year gained (Hlatky, 2002). However, because CERs are
measures of how a program’s costs compare to its outcomes, judgments about whether the
outcomes achieved are worth the costs are for policymakers to decide, not researchers. If
study results indicate that a program has a cost per stroke averted of $2,500, policymakers
must decide whether it is worthwhile to invest in the program.
5.2.2 Incremental or Marginal Cost-Effectiveness
In making decisions about whether to expand a prevention program or whether to fund one
prevention program versus another, it is important to calculate CERs that compare a
Chapter 5 — Conducting a Cost-Effectiveness Study
5-5
program expansion to the existing program (marginal CERs) or that compare all alternatives
(incremental CERs).
Incremental, or marginal, CERs calculate the extra cost required to get an additional unit of
the outcome. In the vaccination program example, incremental CERs were calculated to
compare three different vaccination program types. The incremental CER was the additional
cost per additional patient vaccinated under Program B as compared to Program A.
Incremental CERs are used to compare a program with its next best alternative (i.e., an
intervention that provides the next highest effectiveness).
The incremental, or marginal, CER is calculated as follows:
Outcome B Outcome A
Cost B Cost A
Inc CER
Δ _ Δ _
_ _
_
−
−
= (5.2)
The numerator represents the difference in program (or net program) costs between
Programs A and B, and the denominator captures the difference in impact on outcomes
between Programs A and B. Equation 5.2 assumes that Program B has higher effectiveness
than Program A.
When several programs or program options are available, some evaluators will calculate the
average CER according to Equation 5.1 and then choose to implement the program or
program option with the lowest CER. However, this approach may not always be
appropriate. In some cases, the program to implement will not be the one with the lowest
CER (Bala and Zarkin, 2002). Selection among several alternative programs or program
options depends on several factors, including overall program effectiveness and budget
constraints.
The examples in each of the subsections that follow demonstrate how comparisons across
programs should calculate and use measures of incremental cost-effectiveness when
selecting the program or program option to fund or implement. Chapter 6 contains further
discussion about how policy makers could use information about the incremental costeffectiveness
of alternative programs to select programs that best meet policy criteria for
funding (e.g., achieve greatest total effectiveness or achieve greatest effectiveness per
program participant).
Decision Making Using CERs
Policy makers are often interested in CE study results to help inform decisions about which
prevention programs to fund from among several alternatives. In some cases, interest is in
selecting from among programs that target different behaviors or outcomes. In others, it is
in selecting from among programs that target the same behavior or outcome. The next two
subsections provide examples that use incremental CERs to inform decision making about
which programs should be funded, given that the budget for prevention spending is limited.
Guide to Analyzing the Cost-Effectiveness of Community Public Health Prevention Approaches
5-6
A. Programs Targeting Different Health Conditions. An example of two programs
targeting different health conditions are the National Breast and Cervical Cancer Early
Detection Program (NBCCEDP) targeting cancer and the WISEWOMAN program aimed at
reducing cardiovascular disease risk factors. Although these programs serve similar
populations (middle-aged, low-income, uninsured women), they are aimed at improving
different health conditions; as a result, they can be delivered to the same (or different)
participant groups independent of each other.
To compare programs targeting different health conditions, the average CER presented in
Equation 5.1 should be calculated for each program. To facilitate comparison across
programs with disparate outcomes, a common measure of effectiveness must be used for
each of the programs. For that reason, use of broad health outcome measures, such as life
years gained or number of deaths averted, is generally recommended.
Once the average CER is calculated, programs should be rank ordered based on their CER.
Programs with lower CERs represent more cost-effective strategies. For example, Table 5-2
presents cost-effectiveness results for three programs targeting different conditions. In this
example we use the number of deaths averted by the program as our measure of
effectiveness. This is a broad measure that is equally appropriate and relevant for programs
targeting cancer, cardiovascular disease, and other health conditions.
Table 5-2. Cost-Effectiveness of Three Programs Targeting Different Health
Conditions
Net Program Costs Number of Deaths Averted Average CER
Program A B C = A/B
1 $500,000 100 $5,000 (lowest)
2 $100,000 10 $10,000 (middle)
3 $1,500,000 125 $12,000 (highest)
When comparing programs that address different health outcomes, based on costeffectiveness
alone, programs with lower average CERs should be given priority over
programs with higher cost-effectiveness ratios. In our example, Program 1 has the lowest
CER ($5,000/death averted) and is therefore the most cost-effective program.
However, in order to decide which program(s) should be implemented, the available
prevention budget must be taken into consideration. Following the example in Table 5-2, if
the available prevention budget is less than $500,000, then you should implement as much
of Program 1 as your budget allows. By doing so, you would have the greatest impact
possible on reducing deaths for the given budget. For example, if $250,000 was spent on
Chapter 5 — Conducting a Cost-Effectiveness Study
5-7
Program 1, the expected number of deaths averted would be 50. By comparison, if
$250,000 was spent on Program 2, the expected number of deaths averted would be 25.
If the available prevention budget is exactly $500,000, then you should implement all of
Program 1. If the budget is $500,000 to $600,000, then you should implement all of
Program 1 and as much of Program 2 as you can. The logic here is the same as described
above. If $550,000 were available for prevention efforts and $500,000 were spent on
Program 1, with the remaining $50,000 spent on Program 2, then the expected number of
deaths averted is 105. If, instead, the $550,000 were spent on Program 3, with a lower
average CER, then the expected number of deaths averted would be almost 46
([$550,000/$1,500,000] × 125). If the prevention budget is more than $600,000, then you
should implement all of Programs 1 and 2 and as much of Program 3 as possible to achieve
the greatest possible impact on number of deaths averted through funded prevention
programs.
B. Programs Targeting the Same Health Conditions. An example of programs targeting
the same health condition are different smoking cessation strategies (e.g., self-help
materials vs. physician counseling vs. pharmacotherapy vs. quit lines). When considering
only programs that target the same behavior or condition, the effectiveness measure can be
more narrowly focused (such as the number of quitters for smoking cessation programs)
because the programs being compared have a common aim.
Perhaps counter to intuition, selection among programs targeting the same risk factor or
health condition is often a more complicated process than selection among programs
targeting disparate outcomes. Based on CE study results alone, in many cases, you will find
that the optimal solution would be to offer a combination of two programs; that is, some
participants from your target population should get one program and some participants
should get another. However, for equity reasons, it might be unethical to provide
interventions with different levels of effectiveness to individuals within the same target
population (Ubel et al., 1996; Cantor, 1994). Thus, you should offer the same program to
everyone even if it results in lower effectiveness than a mix of two programs.
Equity Concerns: Offering the Same Program to Everyone in the Target Population.
If you are limited to offering the same program to everyone in your target population, then
you should pick the program that provides the highest effectiveness at a cost below your
budget. To do so, you should calculate the average CER using Equation 5.1, rank order
programs based on their effectiveness, and pick the program that results in the highest
effectiveness but costs no more than the budget.
Table 5-3 presents cost-effectiveness results for four programs targeting the same health
condition rank ordered based on their effectiveness. If the available budget is $150,000 and
equity concerns dictate that every participant in the target population must receive the
same intervention, then Program B should be selected because it provides the greatest
Guide to Analyzing the Cost-Effectiveness of Community Public Health Prevention Approaches
5-8
Table 5-3. Cost-Effectiveness of Four Programs Targeting the Same Risk Factor
or Health Condition
Program Total Program Costs Number of ER Visits Averted Average CER
A $125,000 10 $12,500
B $100,000 15 $6,667
C $750,000 50 $15,000
D $1,000,000 100 $10,000
effectiveness (15 averted ER visits) without going over the available budget. Program A
should not be selected because even though it costs less than the available budget, it is less
effective than Program B (10 averted ER visits). Programs C or D cannot be implemented
because they are not affordable and even if they could be partially implemented with the
$150,000, the expected number of ER visits averted would be lower than under Program B
(10 for Program C and 15 for Program D versus approximately 22 if the full $150,000 were
spent on Program B).
Efficiency Concerns: Offering a Mix of Programs Within the Target Population. If it
is acceptable to provide individuals within the target population with a combination of
different programs, then your decision-making process will be different from the one
outlined above and you may be able to achieve more of the desired program outcome (e.g.,
ER visits averted) for a given investment of prevention dollars. As in the example where
equity concerns required that the same program be given to everyone in the target
population, the first step of this process requires rank ordering interventions based on their
effectiveness. Next, incremental CERs should be calculated for each intervention using
Equation 5.2.
Identifying a program to implement based on CE results requires comparing the incremental
CERs of each program. An example is shown in Table 5-4, where both average and
incremental CERs are given for four programs that target the same health condition
(asthma). The programs have already been rank ordered by effectiveness (measured by the
number of ER visits averted). The least effective intervention (Program A) has the same
incremental CER as its average CER because it is compared to the alternative of doing
nothing (with zero costs and zero effectiveness). For Program B, however, the incremental
CER is calculated by dividing the difference in program costs between Program B and
Program A ($100,000 – $125,000 = –$25,000) by the difference in effectiveness between
the two programs (15 – 10 = 5). Note that because Program B is both less costly and more
effective than Program A, its incremental CER is negative. The literature on costeffectiveness
will often state that such programs are “cost saving” because moving from the
alternative program to a cost saving program can save society money while achieving the
Chapter 5 — Conducting a Cost-Effectiveness Study
5-9
Table 5-4. Incremental Cost-Effectiveness of Four Programs Targeting the Same
Risk Factor or Health Condition
Program
Total Program
Costs
Number of ER
Visits Averted Average CER Incremental CER
A $125,000 10 $12,500 $12,500
B $100,000 15 $6,667 ($5,000)
C $750,000 50 $15,000 $18,571
D $1,000,000 100 $10,000 $5,000
Note: Values shown in parentheses represent negative numbers.
same level of effectiveness. In the language of CEA, Program B is said to “strongly
dominate” Program A. In general, a program is strongly or weakly dominated if its
incremental CER is higher than the incremental CER of the alternative with the next highest
effectiveness (as Program A’s incremental CER is higher than that of Program B). Greater
effectiveness could be achieved by implementing a mix of two alternative programs (Cantor,
1994).
Strongly dominated programs should be removed from the analysis. Once a strongly
dominated program (e.g., Program A in the example in Table 5-4) is excluded from the
analysis, the incremental CERs for the remaining programs must be recalculated to make
comparisons across the remaining programs (Table 5-5). Program B’s incremental CER is
now equal to its average CER because it is being compared to the “do nothing” alternative.
Table 5-5 shows that the incremental CER of Program B ($6,667) is less than the
incremental CER of Program C ($18,571), indicating that Program B should not be excluded
from consideration. Now consider Programs C and D. The incremental CER of Program C
($18,571) is greater than the incremental CER of Program D ($5,000), which has the next
highest effectiveness levels. Program C should therefore be excluded from further
consideration because, for the same level of investment in asthma management, a mix of
Programs B and D will be more effective than Program C alone. Program C is said to be
weakly dominated by Programs B and D. For example, if $750,000 were available to support
asthma management efforts, investing all the money in Program C would be expected to
avert 50 ER visits. By investing $100,000 in Program B and the remaining $650,000 in
Program D, the expected number of ER visits averted is 15 from Program B and 65 from
Program D ([$650,000/$1,000,000] × 100), for a total of 80 ER visits averted. Clearly, this
strategy is preferred.
Guide to Analyzing the Cost-Effectiveness of Community Public Health Prevention Approaches
5-10
Table 5-5. Incremental Cost-Effectiveness of Four Programs Targeting the Same
Risk Factor or Health Condition after Eliminating Program A
Program
Total Program
Costs
Number of ER
Visits Averted Average CER Incremental CER
B $100,000 15 $6,667 6,667
C $750,000 50 $15,000 18,571
D $1,000,000 100 $10,000 5,000
Weakly dominated strategies (Program C in this example) should again be removed from
the analysis and incremental CERs for the remaining programs should be recalculated
(Table 5-6). Because the incremental CER of Program B ($6,667) is less than the
incremental CER of its next best alternative, Program D ($10,588), neither should be
eliminated. Further selection of the most appropriate program or combination of programs
to be implemented will be determined by budget constraints and program characteristics. If
$1,000,000 were available to spend on asthma management and Program B could easily be
expanded by tenfold, devoting the entire budget to Program B would yield the greatest
benefits (= 150 ER visits averted). If Program B cannot easily be expanded, the best
solution is to give all of Program B to some individuals in the target population and avert 15
ER visits, then give $900,000 worth of Program D to the rest of the target population to
avert 90 ER visits, for a total of 105 ER visits averted. Spending the full $1,000,000 on
Program D would yield 100 ER visits averted—less than the 105 ER visits averted in the
solution that involves a mix of Programs B and D.
Table 5-6. Incremental Cost-Effectiveness of Four Programs Targeting the Same
Risk Factor or Health Condition after Eliminating Programs A and C
Program
Total Program
Costs
Number of ER
Visits Averted Average CER Incremental CER
B $100,000 15 $6,667 6,667
D $1,000,000 100 $10,000 10,588
5.3 Sensitivity Analysis
Sensitivity analysis is an important additional step in conducting CE studies to examine the
extent to which changes in the cost or effectiveness values used to calculate CERs affect
conclusions. For example, if the cost of a particular service is uncertain, sensitivity analysis
should recalculate CERs using high and low values for the service to examine the extent to
which differences in costs affect study conclusions.
Chapter 5 — Conducting a Cost-Effectiveness Study
5-11
In the example shown in Table 5-6, if the cost of Program D were reestimated as $500,000
in a sensitivity analysis, the average CER for Program D would be $5,000 and the
incremental CER, as compared to Program B, would be $400,000/85 = $4,706, which is
lower than the incremental CER for Program B of $6,667. In this case, Program D would
dominate Program B, making it the best investment of prevention dollars, based solely on
CE results.
5.4 Checklist for Cost-Effectiveness Analysis
The checklist in Table 5-7 summarizes the key steps in performing CEA. This checklist may
help guide your efforts to assess the cost-effectiveness of your program and make
comparisons with alternative programs.
Table 5-7. Checklist for Cost-Effectiveness Analysis
Decision
Check When
Complete
- Select outcome(s) for CEA.
- Calculate average CER for each program evaluated.
- Rank possible programs by effectiveness (lowest to highest) and
calculate incremental CERs. - Eliminate all dominated (strongly or weakly) programs, recalculating
CERs each time a program is eliminated. - Perform sensitivity analyses.
The post CONDUCTING A COST-EFFECTIVENESS STUDY appeared first on My Assignment Online.
