| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Respiratory Medicine, Royal Devon and Exeter Hospital, Exeter, United Kingdom
Correspondence and requests for reprints should be addressed to David M.G. Halpin, F.R.C.P., Department of Respiratory Medicine, Royal Devon and Exeter Hospital, Barrack Road, Exeter EX2 5DW, UK. E-mail: david.halpin{at}rdehc-tr.swest.nhs.uk
ABSTRACT
Studies describing the economic impact of chronic obstructive pulmonary disease (COPD) are used for several purposes. There can, however, be limitations as costs based on results of a clinical trial are likely to be significantly different from real world practice. Sometimes, it may be more useful to capture the costs of the important components accurately rather than the often unachievable aim of capturing every cost however small. Burden of illness studies can help identify clinical targets or patterns of care—for example, hospitalization—that are major health care cost drivers. In the United Kingdom, burden of COPD studies suggest an annual cost of £781–£1,154 per patient. Cost analyses can be divided into four types: cost minimization, cost-effectiveness, cost benefit, and cost utility. Utilities such as quality-adjusted life year (QALY) measure the effectiveness of different therapies, and can be obtained in various ways and in different populations, potentially leading to significant differences in the results. Payers often apply cost per QALY thresholds when assessing whether a new therapy should be used or not. In the United Kingdom, it is accepted that there is a sigmoid relationship between the cost per QALY and the likelihood of a therapy being recommended, with a lower inflection between £5,000 and £15,000, below which rejection is unlikely and an upper inflection between £25,000 and 35,000, above which acceptance is unlikely, but not impossible. On this basis, pulmonary rehabilitation and inhaled steroids are unlikely to be rejected but lung volume reduction surgery may be.
Key Words: burden of illness • cost-effectiveness • therapy
Societies have a right to expect that money spent on health care is spent wisely, and health economic analyses aim to inform the debate about whether such spending is prudent. They can also aid the assessment of new drugs and technologies (1). Health economic analyses are, however, only part of the overall evaluation process and must be considered alongside efficacy and effectiveness analyses.
In all Western countries health care costs have increased significantly over the last 30 years, rising most steeply in the United States (Figure 1A). In Europe, current health care spending averages about 8% of gross domestic product, whereas in the United States it is about 14% (2). As well as overall health care expenditure rising as a percentage of gross domestic product, the proportion spent on pharmaceuticals has also risen in most countries, particularly in the last 15 years (2) (Figure 1B).
|
Two broad types of health economic evaluation have been undertaken in chronic obstructive pulmonary disease (COPD): burden of illness studies and cost evaluations. This article reviews the principles underlying these approaches and will address uncertainties in the analyses, an area that generally receives little attention. The U.S. Panel on Cost-Effectiveness Analysis concluded that uncertainty in such analyses can arise principally from either uncertainties in the costs or parameters included in the model (parameter uncertainty) or uncertainty about the correct way of combining these parameters in the model (model uncertainty) (5).
BURDEN OF ILLNESS STUDIES
Burden of illness studies aim to describe the economic impact of COPD in a particular setting, usually a country. They are often used for campaigning or political purposes to highlight the importance of a particular condition (6). These studies aim to quantify the direct costs that result from managing a condition, such as the costs of drugs, physician time, and hospitalization. They also attempt to quantify the indirect costs to society, which for COPD are significant.
Indirect costs represent a principal area of uncertainty in burden of illness studies and account for much of the variation in estimates of economic costs between different studies. Indirect cost estimates include the costs of lost productivity but do not usually include the costs of any social security payments or other monetary benefits paid to a patient or their carers as these are not considered as costs to society as a whole, but rather as transfers from taxpayers to the recipients (7). Indirect costs may be substantial and, in many countries, are similar to the direct cost of managing COPD. Recent data from the United Kingdom have shown that 44% of patients with COPD were under retirement age (8); 24% of patients reported being totally unable to work, and a further 9% were limited in their ability to work. On average, a mean of 12 d were lost per patient per year. COPD also resulted in 5% of carers missing time from work (8).
For calculation of the costs of COPD, a number of issues arise. Analyses based on randomized control trials remain the gold standard in economic evaluations because of their high internal validity, but results should be interpreted with caution because of their low external validity. That is, it may not be possible to generalize because of aspects of the trial protocol or patient inclusion and exclusion criteria (9). There are a number of options to enhance external validity; of these, additional modeling and observational data based on real-world disease management are the most promising (10). There have also been concerns about the analysis and interpretation of cost data from published trials (11–13). Potentially misleading conclusions about the relative costs of different or new therapies have often been reported in the absence of supporting statistical evidence.
Costs can be calculated by collecting data at a top level (e.g., national spending on oxygen therapy) or by collecting detailed data on individual item costs incurred in the care of specific patients (14). Deciding which approach to adopt depends largely on the intervention being assessed. Bottom-up or microcosting is often more appropriate when there is a large component of staff input or overheads, significant sharing of staff or facilities between interventions or patient groups, or if the health care costing system does not routinely allocate costs to the intervention level (15). In the consideration of costs, the infrastructure or capital costs of buildings (such as hospitals) should not be overlooked (16).
The costs of lost productivity can be estimated either by the human capital approach, which is based on market wage rates (17), or the friction cost approach, which assumes that in a society with less than full employment another worker can replace the absent worker (18). The choice of approach to this question can have a profound influence on the calculation of indirect costs.
If a detailed bottom-up costing method is used, there may be practical and logistical problems in collecting accurate cost data for every item. In these circumstances it may be more useful to concentrate on capturing the costs of the important components rather than the often unachievable aim of capturing every cost however small.
Four burden of illness studies for COPD in the United Kingdom have been published. Three used a top-down approach (19–21) and one used a bottom-up approach (8). With the year in which they were undertaken and their design taken into account, they resulted in similar estimates of the economic burden of COPD—between £781 and £1,154 per patient per year with an overall annual cost (both direct and indirect) of between £800M and £1,500M. The breakdown of direct costs of COPD care in the United Kingdom from the most recent of these studies is shown in Figure 2. It is clear that just over half the average costs are due to hospitalization. This is also reflected in the breakdown of societal costs according to disease severity, where it is evident that patients with severe COPD account for most of the direct and indirect costs (8) (Table 1).
|
|
Burden of illness studies are also often used to make comparisons of the impacts of different diseases within a society. For these to be meaningful it is important that the diseases have a similar epidemiology, for example, prevalence and age group affected. To avoid "comparing apples with pears," the diseases should have similar impacts on morbidity or mortality and similar impacts on society.
Burden of illness studies are also sometimes presented as comparisons of the impact in different countries. Again for these to be meaningful it is important that the disease has a similar prevalence and that there are not major differences in the health care systems. These estimates are also often subject to reporting or classification biases that reflect local reimbursement arrangements. In addition, there are technical issues about comparing costs in different countries and whether this should be done purely on exchange rate conversions or on the basis of purchasing power within the country (known purchasing power parity) (23).
COST EVALUATIONS
Cost-effectiveness evaluations are increasingly important in health care settings where budgets are limited and the introduction of a new therapy may divert resources from other existing therapies (7). The principal audiences for these assessments vary in different countries. They include national pricing and reimbursement decision makers, regional health authorities, and local budget-holding prescribers. Payers may reach their own conclusions about the data or be guided by organizations such as the National Institute for Clinical Excellence (NICE) in the United Kingdom, formulary committees, health maintenance organizations, employers, physicians, and Health Technology Assessment (HTA) organizations.
When reviewing health economic evaluations, payers have to balance cost-effectiveness against affordability. Many therapies can be shown to be cost-effective to a greater or lesser extent, but it is unlikely that a health care community can afford all of them (24). Cost evaluations can help establish priorities for what should be funded and what may not be affordable.
As well as formal evaluations by bodies such as NICE, individual clinicians are often placed in a position of comparing the cost-effectiveness of therapies. Although there is generally little formal training in health economic evaluation as part of medical school or postgraduate training, as consumers, physicians are used to making decisions about the cost-effectiveness or cost benefits of particular products on a regular basis.
Cost evaluations can be divided into four basic approaches: cost-minimization analyses, cost-effectiveness analyses, cost-benefit analyses, and cost-utility analyses (7) (Table 2). Each has a well-established methodology, but there are still a number of important technical issues that must be considered when evaluating the conclusions of these studies; particularly how uncertainty is handled and the time frame of the evaluation.
|
In response to these deficiencies, more sophisticated qualitative and statistical methods have recently been proposed to deal with parameter and model uncertainty (5, 25–27). Multivariate sensitivity analyses, where two or more parameters are varied simultaneously are a more accurate way of assessing uncertainty than univariate sensitivity analysis. An alternative form of sensitivity analysis involves using a set of extreme values for each parameter to give the highest and lowest cost-effectiveness ratios; however, this approach has been criticized as it is unlikely in practice that all extreme values would occur at the same time. Neither of these techniques allows the calculation of an estimate of the uncertainty of the cost-effectiveness, such as a confidence interval.
Probabilistic sensitivity analysis permits the examination of joint uncertainty in the variables without resorting to the extreme conditions specified in extreme scenario analysis. Statistical methods of handling uncertainty include the delta method, the calculation of joint confidence intervals, bootstrapped estimates, and Markov Monte Carlo simulations.
The delta method applies a mathematical technique known as second-order Taylor series expansion to the estimation of the variance of a function such as the cost-effectiveness ratio. This uses estimates of the variance of the variables determining the cost-effectiveness ratio to determine the variance of the ratio itself (28).
In response to legitimate concerns about the use of parametric approaches to confidence interval estimation given the unknown nature of the sampling distribution of the cost-effectiveness ratio, a number of nonparametric approaches have been developed. Bootstrapping is a computational technique that allows the distribution of the cost-effectiveness ratio to be constructed empirically (29), whereas Markov Monte Carlo simulations are based on a model's disease progression (30). In these models, the disease is defined in terms of different states that are chosen to represent clinically and economically important events in the disease process. Transition probabilities are assigned for movement between these states over a discrete time period known as a "Markov cycle" (30). Markov Monte Carlo simulations follow a large number of patients through the model, thereby allowing an overall profile of costs and outcomes to be generated for each patient according to the path that they follow through the model. As well as giving an overall estimate of the average costs and effects in each arm, averaging these costs and effects over a large number of patients also gives an estimate of the likely variance associated with the parameters estimated by the model. This representation of uncertainty in the estimated costs and effects relates simply to the inherent uncertainty of the probabilistic structure of the model.
The time frame over which costs and benefits are to be calculated is critical for cost-effectiveness evaluations in long-term conditions. To be clinically meaningful, the costs and benefits must be calculated over a time period that reflects the longevity of the effects of the intervention. For example, in COPD, the benefits derived from a course of pulmonary rehabilitation last longer than the duration of the program, whereas the costs of running the program only occur during the time that it is running. There may also be delayed benefits from an intervention that need to be taken into account.
When appraising economic evidence it is essential to consider both the quality of the clinical studies that have been used to assess the effectiveness of a therapy as well as the quality of the economic analysis and costing methods that have been used. Not infrequently, the clinical studies that underpin the economic analysis are of poor methodologic quality whereas the economic analysis itself is of sound methodology. It is also important to ensure that the analysis was based on costings and a model of health care that is relevant to the country of interest. When assessing published economic studies, there is now a tool for analyzing the quality of the economic methods that have been used, but this is relatively new (31).
All four approaches to cost evaluation have been used in COPD. A recent review of economic studies in COPD (32) found 34 economic evaluations of interventions for COPD in the literature. These assessed pharmacotherapy, oxygen therapy, home care, surgery, exercise and rehabilitation, and health education. Of the 15 articles that addressed the cost attributes of various pharmacotherapy, 6 were cost-minimization studies, 8 were cost-effectiveness studies, and 1 contained both types of analysis. Of the remainder 6 were cost-minimization studies, 5 were cost-benefit analyses, 4 were cost-effectiveness analyses, and 3 were cost-utility analyses. The principles underlying each of these approaches are discussed below.
Cost-Minimization Analysis
Cost-minimization analysis is the simplest cost-evaluation method (33). It takes therapies with equal efficacy and compares their costs. Ideally, this should include all aspects of the costs of therapy, including acquisition, storage, administration, and monitoring costs.
In practice very few interventions in COPD can be considered to have equal efficacy and thus the use of cost-minimization analysis is inappropriate: in these circumstances, cost-effectiveness analyses are needed. In practice, these limitations apply more generally to cost-minimization analyses to the point where their use under any circumstances has been questioned (34).
Cost-Effectiveness Analysis
Cost-effectiveness analyses involve calculating the costs of different therapies that achieve different clinical outcomes and comparing these costs on the basis of the cost to achieve a particular outcome (35). The difference between therapies is often expressed as the incremental cost-effectiveness ratio—that is, the additional cost to achieve a better outcome.
An example of such an analysis is the study by Oostenbrink and colleagues that reported the cost-effectiveness of tiotropium versus ipratropium (36). They found that the mean number of exacerbations per year was 1.01 in patients treated with ipratropium and 0.74 per year in those treated with tiotropium. The mean annual health care costs were 
541 and 1,721, respectively. Thus, the incremental cost-effectiveness ratio for tiotropium compared to ipratropium was 667 per exacerbation avoided.
Cost-Benefit Analysis
Cost-benefit analyses calculate the cost of particular interventions and relate these to the cost savings that result from that approach (37). The time over which costs are spread is an important factor in such analyses.
When viewed from the perspective of the whole health care budget, the cost savings can sometimes be virtual rather than real. For example, saving hospital bed days may be very important to allow other patients to be admitted, but rarely reduces overall costs to the health care system as another patient will be admitted to that bed. Nevertheless, this does allow more patients to be treated for the same expenditure and so is a more efficient use of resources.
An example of cost-benefit analysis is a study of the number of patients with COPD needed to treat with a combination of budesonide and formoterol compared with formoterol alone to prevent an exacerbation, and relating this to the costs saved by preventing that exacerbation (38). The number needed to treat for avoiding one exacerbation is 2.1 to 2.4 based on the clinical trials. The mean unit cost of medication per patient per year was estimated at 912 for budesonide/formoterol combination therapy and 718 for formoterol based on Swedish costs. This represents a cost of between 428 and 490 per exacerbation prevented. The costs of treating an exacerbation have been estimated as being between 1,278 for patients with an FEV1 of less than 40% predicted and 382 for patients with an FEV1 between 40 and 59% predicted (39).
Cost-Utility Analysis
Cost-utility analysis is becoming the most widely used form of cost evaluation for new therapies (40). It is essentially a subtype of cost-effectiveness analysis but uses utility to measure the effectiveness as a way of comparing therapies in different disease areas or therapies that achieve different types of outcomes within a disease (e.g., increased exercise capacity vs. reduced exacerbation rates).
The commonest utility used to assess the outcome in these studies is the quality-adjusted life year (QALY) (41–43). Like cost-effectiveness analyses, cost-utility analysis can also be used to derive a cost-utility ratio, which is the incremental cost of an intervention to achieve one QALY compared with an alternate intervention.
The QALY gain or loss is the equivalent number of fully healthy years of life lost to a disease or gained as a result of a therapy. It is based on the actual number of years gained or lost corrected for the quality of life during those years on the basis of a preference weight ranging from 0 (dead) to 1 (perfect health). Preference weights are based on utility values that reflect a person's preference for a particular health state. Utility values for a given health state have been measured by using different populations, including samples of the general public, patients with the disease, and clinicians.
All cost-utility analyses depend on the robustness of the derivation of the utility values, and the methods used to derive these values are presented below. As with economic modeling, there has been little work looking at uncertainty in the estimates of utility values. This is clearly essential when considering the results of cost-utility analyses. The most commonly used methods to derive utilities are the visual analog scale, the standard gamble, the time trade off (42), and the person trade off (44). Each has its advocates (44–50) but the methods are not always interchangeable as the techniques used can generate different values (51).
In the standard gamble, respondents are asked to choose between alternative outcomes, one of which involves uncertainty. They are asked what risk of death they are prepared to accept to avoid the certainty of the health state being valued (52). The time trade-off asks respondents to state which proportion of their remaining life they would be willing to sacrifice in return for being relieved of a given health problem. The more time they are prepared to sacrifice the greater the burden and the lower the utility of their current health state (52).
These choice-based methods for assessing utility values are generally thought to be the most robust and to have the greatest theoretical validity and acceptable levels of reliability (52), but there are limited data on the precision of these estimates. In one study of utility values in COPD, a time trade-off approach in 59 patients yielded median utility values of 0.91 with an interquartile range of 0.80–0.98 (53). Such uncertainty of the estimate is often not reflected in the resultant cost-utility analysis.
Utility values can also be derived from quality-of-life data. It is well known that health related quality of life in COPD is poor and is affected by poor physical functioning, distressing symptoms, inability to work, social isolation, frequent hospitalization, and depression. Health-related quality of life is usually measured with a disease-specific questionnaire, such as the St. George's Respiratory Questionnaire (54), but if a generic measure such as the Short Form-36 (SF-36) is used, the results can be converted into utility values (55, 56). The SF-36 has been shown to be reliable (57) and responsive (58) in COPD, and converting its results to a utility value may give values that are more directly comparable with other diseases. However, such conversions are probably the least reliable way of calculating utility values.
In addition, utility values can be assessed in a population using questionnaires such as the Euroqol (EQ-5D) (59) or Health Utilities Index (60). Both the EQ-5D and the Health Utilities Index are easy to use but they may be insensitive to changes in health status in specific conditions. For example the EQ-5D was unable to detect a change in patients with COPD who said that their health had changed between assessments and who showed a significant change in the SF-36 (61). There is also limited information on the reproducibility of utility values derived from these preference-based measures (52).
In all cases of utility measurement, there is an assumption that utility is being measured on an equal interval scale. In practice this may not be the case (62). The results of utility assessments also depend on whether individual patients are asked to assess their own condition or whether healthy individuals or clinicians are asked to score the impact of theoretic conditions. For making assessments about the relative merits of, for example, whether or not to treat COPD compared with another chronic condition, utility values derived from the general population may be more appropriate than those derived from the choices of patients, but for assessing the most cost-effective treatment for COPD, values derived from patients with COPD are likely to be preferable (63).
Once the utility value has been estimated, QALYs are calculated by multiplying the utility value by the time for which it applies: 2 yr with a utility of 0.4 equates to 0.8 QALYs. A time of 4 yr with a utility of 0.5 gives the same number of QALYs as 8 yr with a utility of 0.25.
Although QALYs are widely used in cost-utility analyses, as has been seen, they are not entirely objective and their use remains controversial. QALYs have been claimed to discriminate against the old as they have fewer life years to gain and also against the sick as they have less quality of life to gain. The derivation of QALYs is seen as rather opaque and thus generates suspicion. Conversely, it is claimed that QALYs ensure that the outcome of a program can be assessed without prejudice and value judgments.
The NICE believes that QALYs are robust and extensively validated (64), and give a good all-around assessment of the impact of disease by taking account of physical mobility, the ability to care for self and undertake activities of daily living, the absence of pain, and the absence of anxiety and depression. QALYs are unaffected by differential productivity, and NICE does not believe that they disadvantage children or the elderly. It also believes that QALYs are directly comparable irrespective of the patient's age or disease: thus, a 4 QALY gain in a 20 yr old is the same as a 4 QALY gain in a 70 yr old.
It has been suggested that a rational way to allocate health care resources would be to use a league table of programs ranked by cost per QALY, starting with the lowest. Resources would be allocated to the most cost-effective program first followed by the next best, and so on until all resources have been allocated. This may be the most logical approach but could result in the closure of high cost but desirable programs such as transplantation. Furthermore, given the uncertainties in estimating utility values the ranking may be incorrect.
QALY gains can be less controversially used to assess the effectiveness of newly introduced therapies. For assessing the cost utility of a therapy, the QALY changes can be plotted against costs and an envelope of certainty can be drawn around the point estimate. The ideal new therapy would be significantly more effective (i.e., result in more QALYs) and cost less. This is called a "dominant" effect. Cost-utility analysis comes into its own when the treatment is significantly more effective but also costs more. The benefit can be assessed as the cost per QALY. This is the incremental cost-utility ratio.
Payers are often thought to apply cost per QALY thresholds when assessing whether a new therapy should be used or not. In the Netherlands a threshold of 20,000 per QALY has been suggested, in Canada a threshold between $25,000 and $75,000 (65), and in the United States, a threshold of $50,000 has been suggested (66). In the United Kingdom, NICE denies having a threshold because it states that there is no empirical basis for deciding at what value a threshold should be set (64). The Institute also states that there may be circumstances when it would want to ignore a threshold. NICE believes that to set a threshold would imply that efficiency has absolute priority over other objectives and, also, as many of the technology supply industries are monopolies, a threshold would discourage price competition. NICE does, however, accept that there is a sigmoid relationship between the cost per QALY with a lower inflection between £5,000 and £15,000 below which rejection is unlikely and an upper inflection between £25,000 and £35,000 above which acceptance is unlikely, but not impossible (64).
There are a growing number of cost-utility analyses of COPD therapies in the literature. Pulmonary rehabilitation has been estimated to cost between £2,000 and £6,000 per QALY (67) and thus in England and Wales would fall in the unlikely-to-be-rejected zone of the NICE evaluation. Inhaled steroids for GOLD stage 2–3 disease have been estimated to cost about £8,000 per QALY (68) and thus again in England and Wales would fall in the unlikely-to-be-rejected zone of the NICE evaluation. Conversely, lung volume reduction surgery has been estimated to cost around £100,000 per QALY when all patients are operated on and around £50,000 per QALY even if surgery is limited to those shown to derive the greatest benefit (69). This estimate lies in the likely-to-be-rejected zone of the NICE evaluation.
CONCLUSIONS
Despite the impact of the disease, there is still a paucity of health economic evaluations in COPD. For development of the NICE COPD guideline (70), an extensive literature search was undertaken. Sixty-eight papers were thought to be of potential relevance but when appraised critically many were found not to be good quality, formal, economic evaluations.
Bronchodilator therapy and pulmonary rehabilitation had the largest number of economic evaluations followed by hospital-at-home schemes and noninvasive ventilation. It is surprising that there are not more good quality evaluations of pharmacologic therapies. When viewed from the perspective of a single health economy, most health economic literature is problematic as it relates to other health economies and often the evaluation is sponsored by a pharmaceutical company, which may lead to publication bias.
In conclusion, it is clear that health economic data about COPD are still patchy, of variable quality, and frequently fail to report the uncertainty of the estimates. However, health economic data are likely to be of increasing importance for payers when approving new therapies. Burden of disease data are important for campaigning and for helping to identify clinical and research priorities.
FOOTNOTES
Conflict of Interest Statement: D.M.G.H. received less than $10,000 from AstraZeneca over the last 3 yr for speaking at international symposia, including the Lund Symposium.
(Received in original form July 29, 2005; accepted in final form November 25, 2005)
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  G. S. Bauldoff, N. A. Ryan-Wenger, and P. T. Diaz Wrist Actigraphy Validation of Exercise Movement in COPD West J Nurs Res, November 1, 2007; 29(7): 789 - 802. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
