Discounting Future Costs
The last issue regarding costs involves timing. Costs occurring in future years (i.e., beyond the first year of analysis) should be discounted at a rate of 2% to 5% per year. This is particularly important when evaluating interventions that have differential timings related to costs. Furthermore, if it is assumed that costs will increase over time, then they should be adjusted for inflation. More detail on discounting and inflation can be found in Methods for the Economic Analysis of Health Care Programs (Drummond et al., 2005).
All economic evaluations require evaluating costs. However, inclusion of outcome measures depends on the type of study being conducted. Cost-effectiveness, cost-utility, and cost-benefit analyses employ a ratio examining cost per outcome achieved. A good economic evaluation will incorporate high-quality outcome (effectiveness) data. Effectiveness data can come from a variety of sources; however, not all sources are equal. Obtaining quality effectiveness data can be challenging. As with any critical analysis, effectiveness data should be carefully scrutinized on the basis of the study designs (Drummond et al., 2005). The U.S. Preventive Service
Task Force has listed randomized controlled trials as the gold standard of evidence. This is followed by observational data (e.g., cohort, case-control, and crosssectional studies), uncontrolled experiments, descriptive series, and expert opinion (Gold et al., 1996). Prior to using data, its quality should be considered.
Cost-effectiveness studies use physical measures of effectiveness. Ideally, longterm outcomes are used (e.g., life-years gained, or deaths avoided). However, cost-effectiveness studies also use intermediate outcomes as that is what is available in the literature (e.g., depression cases identified, or percentage reduction in blood pressure). When intermediate outcomes are used, they should be linked to longterm outcomes (e.g., survival or long-term care placement). Linking intermediate outcomes to long-term outcomes is challenging and can be accomplished through additional modeling (Briggs et al., 2006; Hunink & Glasziou, 2001). For a detailed discussion of linking intermediate to long-term outcomes, please see Methods for the Economic Evaluation of Health Care Programs (Drummond et al., 2005).
Cost-effectiveness studies sometimes employ outcomes from health status questionnaires (e.g., the Medical Outcomes Study Short Form, 36-item; SF-36). Health status questionnaires are appealing to use because they capture many dimensions of health and provide a single summary score. However, many health status questionnaires arbitrarily weigh responses to provide summary scores. These health status instruments are not preference based, and summary scores are often not interpretable.
Quality-of-life weights use preference-based methods to evaluate the importance of each response levels, questions, and dimensions in a health status questionnaire. Preference-based quality-of-life weights produce a single summary score that represents the quality of life associated with a given health state for an individual. Preference-based quality-of-life weights are sometimes referred to as “health utility” values and have several important mathematical properties that make them ideal for statistical analysis. Foremost, quality-of-life weights are based on an interval scale and scored from 0 to 1, where 0 represents dead and 1 represents perfect health. The properties of an interval scale are important because they allow for evaluating the difference between values, and the difference between values represents an interpretable magnitude. This property does not hold for many general health status instruments.
In addition, quality-of-life weights can be used to determine QALYs, a measure which combines morbidity and mortality into a single measure of effectiveness. Mortality is measured in life years, and morbidity is measured by the quality-of-life weights. The quality-of-life weights are used to assess morbidity over the years of life. Studies that use QALYs as the outcome measure are referred to as cost-utility analyses.
Several off-the-shelf health status questionnaires exist that can be used to derive quality-of-life weights. The most common questionnaire is the EuroQol-5 Dimensions (EQ-5D). The EQ-5D consists of five domains (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and each question has three responses (no problems, some problems, and extreme problems) (Shaw, Johnson, & Coons, 2005). Another popular health status questionnaire is the Health Utility Index Mark III (HUI-III). The HUI-III is more complex than the EQ-5D, and consists of eight domains (vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain) and five or six response levels for each dimension (Horsman,
Furlong, Feeny, & Torrance, 2003). Finally, the Quality of Well-Being Index represents another health status questionnaire that can be used to derive quality-of-life weights (Kaplan, 1998). Any of these instruments can be included as part of a prospective study design. Unfortunately, many large national surveys (in the United States) have not included these instruments as part of their design. However, there are published studies that report quality-of-life weights for common instruments for individuals with various conditions. For example, the Beaver Dam study provides an inventory of quality-of-life weights for individuals with and without various chronic conditions (Fryback et al., 1993).