ارزیابی های زمان-تجارت کردن اتوماتیکی بهداشت کنونی در سلامت زنان

Objective: To motivate the role for preference assessment in women’s health and to report pilot data addressing the performance of automated time-trade-off (TTO) valuations of current health, which were developed to estimate quality-adjusted life years (QALYs) in the women’s health setting. Methods: Values for current health relative to perfect health and death were assessed using an annual time trade-off (1-year horizon and sleep as the trading metaphor), a lifetime time trade-off, and a visual analog scale (VAS). All instruments were administered twice within a 12- to 14-day window among a convenience sample of 27 women. Results: Valuation of health was similar for both time trade-offs (mean of 0.95 for both), but was significantly lower for the VAS (mean of 0.84, Wilcoxon signed-rank p-value < 0.001). Reliability using the intraclass correlation coefficient was 0.67 ± 0.09 and 0.75 ± 0.07 for the annual and lifetime time trade-offs, respectively, and 0.89 ± 0.03 for the VAS. Construct validity was supported by consistent trends in time-trade-off utilities across tertiles of the Short Form 36 (SF-36) general health subscale (trend test p-value < 0.001). Conclusion: Automated time trade-offs for current health provide a promising approach for use in women’s health studies where impact on QALYs must be measured. Natural areas of application include the economic evaluation of preventive interventions in postmenopausal women. Keywords: cost-effectiveness, osteoporosis. qualityadjusted life years, utility, women’s health.

As more women consider long-term use of postmenopausal pharmacological agents, such as hormone replacement therapy (HRT) or selective estrogen receptor modulators (SERMs) to prevent osteoporosis and other diseases, it will become increasingly important to understand the effects of these drugs on health-related quality of life. To assess the economic value of such interventions using quality-adjusted life years (QALYs) as the health end point of interest, quantitative estimates reflecting how women value their health are required. These values, which are sometimes referred to as utilities, are scaled from 0 (death or worst imaginable health state) to 1 (perfect health or best imaginable health state). When estimating QALYs, each year of life is weighed according to its quantitative value estimate or utility. Von Neumann and Morgenstern [1] described the axiomatic basis for utility assessment in the 1940s. Since then a number of approaches to assessing values for health states have been developed, including the timetrade- off (TTO) technique described by Torrance et al. [2]. In contrast to health-status instruments such as the Women’s Health Questionnaire [3], the purpose of health valuation is to measure how women feel about their health rather than to characterize their particular functional health state. In the conceptual model of patient outcomes described by Wilson and Cleary [4], values and preferences are noted as influ-encing both general health perceptions and overall quality of life. Valuations of current (or global) health aim to measure each individual’s preferences for her overall health state [5]. Thus, two women in identical health states as measured by a healthstatus instrument such as the Short Form 36 (SF- 36) [6] who value their health differently would have different utilities for current health. The importance of such valuations in the economic evaluation of interventions in women’s health is nicely exemplified by the findings of the first study to address the cost-effectiveness of estrogen replacement therapy in postmenopausal women. Weinstein [7] showed that the long-term benefits of HRT could be outweighed among asymptomatic women if they felt that HRT side effects reduced quality of life by 2 to 3 days per year. Despite the impact of such side effects on the results of economic evaluations in women’s health, very few data reflecting their affect on overall health valuation (i.e., QALYs) are available. Feeny and Torrance [8] have described the role for utility assessment as an adjunct to usual end points in clinical trials. However, the large resources required for deploying highly trained interviewers at multiple sites are barrier to widespread implementation of such measures. Automated preferenceassessment tools such as U-Titer [9] and IMPACT [10] facilitate consistent utility elicitation without a trained interviewer and thereby provide one solution to this problem [11,12]. Another approach to incorporating health-state valuation in clinical trials is to use self-administered preference classification systems, such as the Health Utilities Index (HUI) [8,13–15], EQ-5D [16–18], or SF-6D [19]. A primary advantage of such instruments is that they provide health-state valuations based on societal preferences, which are most appropriate for cost-effectiveness evaluation [20]. For diseases such as osteoporosis, which have a large impact on physical function, preference classification systems may be sufficient for characterizing health valuation. This conjecture is supported by a cross-sectional study of women with fracture where HUI provided similar health-state valuations compared with directly assessed TTO values for current health [21]. However, a primary disadvantage of generic preference classification systems is that they may not be sensitive enough to characterize the influence of menopause and other factors on women’s health (e.g., the influence of vasomotor symptoms on health valuation). This is of particular concern in the prevention trial setting where it is essential to account for the influence of treatment side effects on health valuation. Given this uncertainty, preference classification systems should be considered as an adjunct rather than a replacement for direct utility assessment. As one step toward the economic evaluation of women’s health interventions, we developed an automated computer-based interview to assess current health valuation using two TTO techniques. Both an annual TTO, appropriate for valuing overall health or transient health problems, and a more traditional lifetime TTO, appropriate for valuing long-term or chronic health conditions, were implemented. In this paper, pilot data addressing the reliability and validity of these two current health assessments among women are reported. Applications of these instruments in women’s health studies are also reviewed.

When assessing the economic value of women’s health interventions, it is desirable to measure their impact on QALYs. Automated preferenceassessment tools [11,12] make it possible to collect patient preference data in a consistent fashion in multicenter studies. While others have reported on the validity of automated instruments for measuring health-state preferences [5,27], little has been reported on their reliability. Most reliability studies of TTO assessments have focused on the valuation of health-state descriptions (e.g., EQ-5D health states) rather than current health [28–30] and have been interviewer administered. The reported 2- week reliability for both current health TTO assessments implemented in U-Titer (intraclass correlation coefficients of 0.67–0.75) was reasonable relative to previous reports of classic intervieweradministered TTO assessments (intraclass correlation coefficients of 0. 62–0.87 for retest periods of one year to one week or less) [28]. In addition, using the reliability of the general health and vitality subscales as benchmarks for comparison, the annual TTO had poorer reliability than either health status subscale, while reliability for the lifetime TTO and vitality subscale were similar. The current health TTO values also tracked appropriately with general health as measured by the SF-36 general health subscale. These data suggest that automated TTOs described here may be appropriate for use in selected female populations. In practice, choice of time horizon will be dictated by the study objective and setting. It is important that a time horizon that is meaningful to study participants be used. One advantage of the annual TTO over the lifetime assessment is that the influence of time preference on utilities is minimized. The annual TTO described here is most appropriate when one wishes to characterize the potentially transient effects of an intervention on QALYs. This situation naturally arises in clinical trials of pharmaceutical agents where potential drug side effects may adversely or beneficially affect QALYs. It is Figure 3 Box and whisker plots showing values for current health for lifetime and annual time trade-offs across tertiles of the SF-36 general health subscale. The boxes indicate the 25th and 75th percentiles (i.e., interquartile range-IQR) with the middle line indicating the median value. Lines extending from the boxes (i.e., whiskers) mark the upper and lower adjacent values, which are defined as the 75th percentile +1.5*IQR and the 25th percentile -1.5*IQR. More extreme values are individually plotted.important, however, to note that the sleep metaphor may not be appropriate for use in assessing healthstate values among women experiencing acute menopausal symptoms involving sleep disturbances. Investigators should consider collecting data on perceived sleep needs and recent sleep experience when trade-offs involving the sleep metaphor are implemented. Alternative TTO approaches to valuing transient health states include use of an annual TTO in which convalescence serves as the trading metaphor [25], the use of chained TTOs [31], or the evaluation of health paths [32,33]. Our pilot study had several important limitations that must be addressed. First, our convenience sample size of 27 women employed by Dartmouth College was relatively small and participants likely had higher socioeconomic status, educational attainment levels, and health status than the general population. As a result, our estimates of reliability may represent a best-case scenario. Second, because of our primary interest in estimating the reliability of the annual TTO, this instrument was always presented following the training task. Although women were not given a final preference score at the end of each assessment task, we cannot rule out anchoring as a factor contributing to higher reliability scores for the lifetime TTO and VAS. While some have reported higher reliability scores for visual analog scales [30], this finding is not consistent across all studies [28]. Third, our evaluation of validity was limited by the amount of additional health-status information that we collected and by the number of health states that were assessed. The latter precluded our use of recently proposed methods for measuring the validity of utility assessments by considering the consistency of rank orderings for multiple health states across instruments [27,34]. However, when the primary focus is on current-health valuation rather than valuation of imagined health states (i.e., health-state descriptions), use of consistency rankings may have limited applicability. Several aspects of the TTO assessments that we evaluated also deserve comment. First, although development of the automated instruments was motivated by applications in women’s health, current health assessments are global measures of health preference and may find application in a wide variety of disease areas. Our pilot data, however, do not support use of these instruments in the general female population, male populations, or specific disease areas. Further investigation of the reliability of these instruments is needed prior to their use in other settings. Second, by using death as a lower anchor in the lifetime TTO and VAS tasks, we did not allow participants to rate their own health state as worse than death [35]. Fortunately, this is unlikely to be of practical importance when used to assess the value of preventive health interventions. Differences in scaling of utilities across studies could, however, influence the comparability of cost-utility analyses [36]. Finally, more recent versions of U-Titer allow investigators to specify alternative search algorithms. Because of the documented effect of the search algorithm on the health-state values obtained [37] we note that our pilot data apply only to the TTOs implemented using a bisection search algorithm. To date, the automated TTO assessments of current health described here have been used in several studies among women. The annual TTO (using the modified choice A format) was used in a health-economic trial comparing raloxifene and hormone replacement therapy for postmenopausal osteoporosis prevention [11]. The annual trade-off was implemented because the study’s objective was to evaluate the effect of a pharmaceutical intervention on the valuation of near-term health. This UTiter interview also included a valuation of side effects, which are often transient, and required a short rather than lifetime time horizon. In this setting, use of the annual rather than lifetime current health assessment had the added practical advantage of requiring introduction to only one trading metaphor. Although no longitudinal differences in health valuation between treatment groups were noted [38], the detailed preference data collected prospectively in this trial allowed for an analysis of the impact of specific treatmentemergent side effects on the incidence of low valuation for current health [39]. The lifetime TTO for current health was used in a continuing study whose objective is to identify cost-effective approaches to osteoporosis prevention and treatment in the elderly [21,24,40]. An analysis of 382 women aged 50 to 96 years addressed the impact of fractures on quality of life using the classic TTO assessment and indicated that women with hip and/or vertebral fractures would give up 6 to 51% more of their remaining life years to attain prefect health relative to women without fracture [40]. In another analysis, which investigated the impact of HRT on current health valuation, current HRT users were found to have significantly higher values when compared with never and past HRT users [24].In summary, when assessing the economic value of preventive interventions on health, it is particularly important that the effect of these interventions on quality of life be considered. The automated current-health TTO instruments described here provide one approach for collecting preference data when impact on QALYs must be addressed. Applications of these instruments in the women’s health setting appear promising. We conclude that automated utility assessment instruments should be considered as one approach to characterizing the health impact of women’s health interventions. This research was supported in part by grant AG12262 from the National Institute on Aging.