| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Commentary |
1 Department of Radiology, SUNY Upstate Medical University, 750 E Adams St.,
Syracuse, NY 13210.
2 Department of Diagnostic Radiology, Yale University School of Medicine, New
Haven, CT.
Received July 10, 2008; accepted after revision July 13, 2008.
Address correspondence to C. S. Poon
(poonc{at}upstate.edu).
OBJECTIVE. This commentary provides a brief overview of cost-effectiveness analysis, which is increasingly applied in radiologic research. The purpose is to familiarize readers with the basic concepts in this topic and to provide help in appraising original articles in this area of research, as featured in this issue of the AJR.
CONCLUSION.Despite some limitations, decision-analytic modeling provides a useful tool for cost-effectiveness analysis in emerging technologies and helps to direct future research and the practice of radiology.
Keywords: breast MRI • cost-effectiveness • decision analysis • MRI • MR lymphangiography • technology assessment
Over the past several decades, we have seen tremendous technologic advances in radiology that have transformed the practice of medicine. These new technologies now enable us to image human anatomy, physiology, and diseases with unprecedented detail. Patients have an increasingly wide variety of imaging options. The rapidly increasing use of radiologic procedures is a testament to the success of these new imaging technologies. Yet, the rosy future of our technologic success is hampered by constraining economic resources. The limited health care budget, an aging population, increasing demands from patients and referring physicians, and overuse of some new technologies are now posing a new challenge to these technologies. Faced with this dilemma, many health care providers, health policy makers, and insurers are questioning the true value of these new technologies. Traditionally, radiologic research emphasizes development of new technology and assessment of clinical efficacy. Increasingly, these new technologies are also being scrutinized for their cost-effectiveness in patient care [1]. In response to this demand, there has been an increase in the literature on cost-effectiveness of radiologic procedures. This issue of the AJR provides another excellent addition to the literature on this topic [2].
In choosing the best radiologic test for a particular clinical problem, three questions often arise: First, what is the diagnostic efficacy of the test? Second, does the test provide a good value given its cost? Third, how does the test compare with other alternatives in terms of providing the biggest "bang for the buck"? Cost-effectiveness analysis is an attempt to formally address the latter two questions. Although a radiologic test may be ranked highly in a technologic sense (e.g., high diagnostic accuracy), it may not be an ideal test because it is prohibitively expensive or provides only a small marginal benefit compared with other less expensive alternatives.
Cost-Effectiveness Assessment
Ideally, assessment of the cost-effectiveness of a new technology should be conducted from a well-designed empiric study [1, 3]. Patients can be randomized into experimental arms including a group receiving the test and a separate group receiving no test or an alternative test. The patients are followed for a defined period of time and patient outcomes (test accuracy, morbidity, mortality, patient preference, costs, and so on) are measured directly. Such an empiric study, performed in an ideal situation, will result in the most accurate measurement of the cost-effectiveness. In reality, this ideal approach is virtually impractical. The problem of technological assessment is highly complex and involves many variables that are difficult to control or measure accurately. Such a study may also be prohibitively long in duration, inflexible, and costly. An alternative approach is to perform the cost-effectiveness analysis using a statistical model [4–7].
Decision-Analytic Modeling
Many studies in cost-effectiveness analysis are performed using
decision-analytic modeling
[4–7].
Cost-effectiveness analysis is only meaningful when alternative diag nostic or
therapeutic options are compared
[7,
8]. These alternatives may
include no test or no treatment (watchful waiting). Therefore, a properly
conducted cost-effectiveness analysis will first require an explicit statement
of the diagnostic or therapeutic options to be compared
[8]. The next step is an
appropriate estimation of the costs associated with a test or intervention.
The costs can include the direct monetary costs (such as the professional fees
and technical fees associated with a test), indirect costs (such as the costs
of further unnecessary investigation resulting from a false-positive test),
and intangible costs (such as pain and suffering resulting from a
false-positive or false-negative test result).
Measurement of Effectiveness
The measurement of effectiveness will need to be explicitly defined. A
number of measures of effectiveness can be used. These include number of
diseases detected, morbidity and mortality, life expectancy, and
health-related quality of life. However, because the technologic success of a
diag nostic test (measured by its ability to detect a high number of diseases)
may not translate into better patient outcomes (such as improved patient
survival or quality of life), the use of universal patient health outcomes
(e.g., quality-adjusted life expectancy) is advocated
[1,
7]. This will also help to
compare a new technology with other health care services that compete for the
same health care resources. For example, given a constrained health care
budget, should the money be used to fund mammographic screening of older women
versus some other health care services, such as promotion of smoking
cessation?
Decision Analysis
The costs and effectiveness of alternative diagnostic or treatment options
can then be calculated based on a decision-analytic model
[4,
7,
9]. Decision analysis is a
technique that formalizes the problem of choosing alternative strategies. A
decision tree is constructed to model the alternative choices in the workup
and management of patients and the chains of events or health outcomes that
may result from these alternative decisions. An example can be seen in the
Figure 1 of the article by Pandharipande et al.
[2]. In a graphic
representation, the decision and its outcomes are represented by nodes. The
nodes are connected by lines that show the relationship of the nodes. Each
node represents a decision (such as to receive or not to receive a diagnostic
test) or a probabilistic event (such as the result of a test or outcome of a
treatment). The branching of a decision tree is constructed to model the
chains of events involved in a decision process. The final nodes to the far
right of the decision tree represent the possible health states of a patient
at a given time (such as living well, living with disease, or death).
Markov Model
Because the final outcome of a decision may take years to be realized and,
over time, many variables can change, drawing a complete decision tree to map
out all the possible chains of outcomes is cumbersome and impractical. To
overcome this, a Markov model can be used
[9,
10]. The Markov model assumes
a hypothetical cohort of patients who may be assigned one of the many possible
but exclusive health states at the end of a specified follow-up period. An
iterative analysis is performed in which a patient is allowed to transit from
one health state to another during each hypothetical follow-up period. For
example, in a model to evaluate mammographic screening, the follow-up period
may be annually, and the probability to transit from a health state of living
with disease to death is determined by annual mortality for patients with
disease. The measures of costs and effectiveness are allowed to accumulate and
accounted for as the hypothetical patient cohort is followed through an
extended duration of time. Cost-effectiveness analysis based on
decision-analytic modeling, in essence, is a computer simulation of an empiric
study to evaluate the cost-effectiveness of various options. Such an analysis
can be constructed using com mer cially available software. The data required
for the modeling (such as sensitivity and specificity of a diagnostic test,
morbidity and mortality related to treatment and diseases, probability for a
patient to transit from one health state to another) can be obtained from a
review of the literature and published databases.
Rationale for Decision-Analytic Modeling
Several factors make decision-analytic modeling appealing for
cost-effectiveness analysis. First, decision-analytic modeling facilitates an
easy integration of data obtained from various sources and required for a
complex sequence of decisions and events
[3]. It also offers more
flexibility than an empiric study. A model can be updated relatively easily as
we gain new insights in the problem of interest or as new data from the
literature become available.
Another unique pro perty of decision-analytic modeling is that it facilitates decision making under uncertain ties [3]. Because of the complexity of health care issues, the data that one relies on for decision making are often sketchy, incomplete, or questionable. Decision analysis allows the alternative options to be analyzed even in the presence of incomplete or questionable data inherent in a complex decision process. In the presence of incomplete data, a best estimate of the data can be made, such as by expert opinion, to provide a base-case analysis. The dependence of the results on such incomplete data can be assessed through sensitivity analysis, in which the input parameters are varied systematically.
Sensitivity Analysis
A sensitivity analysis is analogous to the estimation of a confidence
interval in an empiric study. The simplest sensitivity analysis is performed
by changing only one variable at a time (one-way sensitivity analysis). The
limitation of a one-way sensitivity analysis is that it may not be adequate to
reflect the full range of uncertainty of the results because errors on more
than one variable may have compounding effects. A more complete assess ment of
the stability and validity of decision analysis can be performed through a
multiway sensitivity analysis, in which multiple variables are varied
simultaneously. This may be performed through more sophisticated variations of
the modeling [11].
Unfortunately, sensitivity analyses involving more than three variables are
difficult to perform and interpret. A two-way sensitivity analysis is
therefore often a reasonable compromise.
Interpretation of Cost-Effectiveness Analysis
Interpretation of a cost-effectiveness analy sis requires a critical appraisal of the validity of a decision tree in its representation of the problem and the quality of the input parameters. In situations in which accurate input parameters are not available, adequate sensitivity analysis will need to be performed to test the stability of the results with respect to uncertainties of these input parameters. In ranking the cost-effectiveness of the various decision options, an understanding of incremental cost-effectiveness is important [1, 6, 7] because cost-effectiveness analysis is always performed in the context of comparing one option to another (often a better established alternative).
Incremental cost-effectiveness ratio is defined as the ratio of the change in costs and the change in effects because we adopt a new test or intervention compared with its alternative. A test that is both less costly and more effective is obviously more cost-effective. A test that is more costly and less effective is obviously less cost-effective. In reality, most new technologies are more effective but also more costly. To decide which one of these more costly options we should adopt, we will need to consider their incremental cost-effectiveness. If a test provides a greater increase of effectiveness on the basis of per dollar increase in cost (i.e., a smaller incremental cost-effectiveness ratio), it is considered more cost-effective compared with its alternatives because, even though it is more costly, it provides a better value for its additional cost. The options that are considered less cost-effective are said to be "dominated" by the other options in the sense of cost-effectiveness. We can also consider cost-effectiveness of a test with respect to other health care services that have already been well accepted by our society [12]. A new technology that has a similar cost-effectiveness ratio compared with other well-adopted services can also be considered cost-effective and is adoptable from a societal perspective.
Cost-Effectiveness Analysis for Evaluation of Emerging technology
Cost-effectiveness analysis based on decision-analytic modeling is well suited for evaluation of an emerging technology, as featured in the article of Pandharipande et al. [2]. In that article, the authors compared the cost-effectiveness of various strategies of axillary staging for early breast cancer using MR lymphangiography, sentinel lymph node biopsy, combined MR lymphangiography and sentinel lymph node biopsy, or no axillary staging. MR lymphangiography is a new axillary staging technique with limited published data to date. Although a few studies, each with a small number of patients, have been published and are referenced in the article, the diagnostic accuracy and long-term benefits to patients of MR lymphangiography remain to be established.
Before more resources are invested to evaluate the clinical efficacy of MR lymphangiography, a preliminary cost-effectiveness analysis will be advantageous to decide whether further studies are even warranted. If MR lymphangiography is found to be dominated in cost-effectiveness by better established alternatives, meaning it is clearly less cost-effective even when accounting for a reasonable range of uncertainties, further research on MR lymphangiography may not even be necessary because the technique probably will not be adopted.
Pandharipande et al. [2] concluded that MR lymphangiography, either used alone or in combination with sentinel lymph node biopsy, is favorable in cost-effectiveness compared with using sentinel lymph node biopsy alone. To compensate for the uncertainties in the input parameters of their models, the authors performed an extensive sensitivity analysis, including a broad range of analysis that was designed to capture the "worst-case scenarios," and a two-way sensitivity analysis on input parameters that potentially could have the greatest effect on the analysis. These parameters included the uncertainties in the sensitivity and specificity of MR lymphangiography and sentinel nodal biopsy, estimation of the costs of MR lymphangiography and its alternative options, and an estimation of the quality of life reduction in patients treated by axillary lymph node biopsy or resection. The analysis was sensitive to estimation of the sensitivity of MR lymphangiography and sentinel lymph node biopsy and the quality of life reduction after nodal biopsy or resection. It was insensitive to other variables. The results of the analysis highlight the needs for better empiric study to assess the clinical efficacy of MR lymphangiography and quality of life issues related to sentinel lymph node biopsy and axillary nodal dissection.
Cost-effectiveness analysis using decision-analytic modeling is not a complete substitute for well-conducted empiric studies. The results of cost-effectiveness analysis can only be as good as its assumptions and input parameters. High-quality input data can only be derived from well-conducted empiric studies. In addition, cost-effectiveness is only one consideration in the evaluation of a new technology. Many other factors such as pref erences and expectations of patients and health care providers, medical–legal considerations, and availability can all affect whether a new technology will be adopted. Although the study of Pandharipande et al. [2] is far from conclusive, it does serve the important purpose of highlighting the potential and providing directions for future research on MR lymphangiography. Similar principles can also be applied to the many exciting new developments occurring in radiologic practice.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
