Objective. To measure parents’ and other adults’ values for preventing disease associated with pneumococcal infection and to evaluate how including these values changes the economic appraisal of pneumococcal conjugate vaccine.
Methods. Data on preferences and willingness to pay to reduce risk of illness were collected for 6 illnesses that are preventable by pneumococcal conjugate vaccine (simple otitis media, complex otitis media, moderate pneumonia, severe pneumonia, bacteremia, and meningitis) and 1 vaccine-related adverse event (fever and fussiness after vaccine). Interviews were conducted with 2 groups of respondents: 1) parents of children who had experienced 1 or more of the outcomes described in the survey (n = 101) and 2) a US community sample (n = 109). The 30-minute telephone interview used modified time trade-off questions and willingness-to-pay questions. Values from the interview were incorporated in an existing decision-analytic model that simulated the cost-effectiveness and cost-benefit of pneumococcal conjugate vaccine in a hypothetical cohort of newborns.
Results. Among parents, the median amount of time that respondents said that they would be willing to trade to avoid diseases ranged from 0 days for otitis media to 1 year for severe pneumonia and 2 years for meningitis. Among the US community sample, the median amounts of time traded were higher, ranging from 7 days for otitis media to 3 years for meningitis. Median willingness-to-pay amounts varied from $100 to prevent 1 episode of otitis media and $500 to reduce the risk of meningitis from 21 in 100 000 to 6 in 100 000 and were similar between parents and community members. Incorporating time trade-off amounts into the existing economic model for pneumococcal conjugate vaccine resulted in cost-effectiveness ratios <$10 000 per quality-adjusted life year at a vaccine cost of $58 per dose.
Conclusions. Both parents and community members assign relatively high values to preventing meningitis, pneumonia, and complex otitis media. When the value of preventing pneumococcal diseases is incorporated into economic analyses, pneumococcal conjugate vaccine has a cost-effectiveness ratio in the range of other widely used health interventions.
Dear Editor:
Dr. Beutels and Ms. Viney suggest in their letter that we have departed from accepted methods for temporary health states in children. Valuing temporary health states has not received much attention in either the theoretical or applied health state valuation literature and, contrary to their assertion, there are no widely-accepted standards for valuing temporary health states. Applying utilities from standardized instruments such as the Health Utilities Index (HUI) or the EQ-5D which were developed to value chronic health states in adults (and children 6 and older in the case of the HUI) are unlikely to be accurate for valuing temporary or transient health states in very young children.[1,2] Therefore we chose to collect primary data for preferences for health states prevented by pneumococcal conjugate vaccination.
There is a small but growing body of literature in the area of valuing temporary health states. Alternatives such as the waiting- tradeoff, conjoint analysis, “chained” health states, and other modifications of the time-tradeoff method have been proposed without any clear consensus on a preferred method.[3-6] Our approach draws on one suggested method modified for application to children’s health by using the parent as the respondent. This approach basically converts the temporary state into a short-term chronic state to calculate the utility (or disutility) associated with a particular health state, then this utility weight is included in the cost-effectiveness model only for the duration of the temporary health state.[7] We have modified this method by asking respondents to value a short-term health state for a hypothetical child. The calculation we used to convert time traded off into the change in quality-adjusted life years assumes that the time traded off (DAYS_F, or time foregone) is traded off against the timeframe of the temporary health state (DAYS_HS, or days in the health state), resulting in a disutility value of (DAYS_F)/(DAYS_HS). When this weight was included in the model, we prorated by the fraction of time spent in a health state, e.g.:
(DAYS_F/DAYS_HS)* (DAYS_HS/365) = (DAYS_F/365)
An alternative approach, which we did not use in this study, is that described by Beutels & Viney in which the utility is calculated by dividing the number of days traded off by the respondent’s remaining lifetime. This would be appropriate if the respondent had been asked to value the health state described as, for example, “7 days of otitis media followed by a lifetime of perfect health”. In that case, it would be appropriate to scale the response as suggested: DAYS_F/(365*LE) where LE is the life expectancy of the respondent to calculate the disutility associated with a temporary health state.
The suggestion to specify that the time traded off will come from the end of a respondent’s life is valid, and, indeed, we have already incorporated this into more recent questionnaires and in future studies will likely use discounted amounts in the base case analysis. Debriefing of pre-test study subjects suggested that most respondents were trading time from the end of life, which is why we also provided discounted TTO amounts in our paper.
Valuing the health of very young children introduces additional challenges to the valuation task, including that of whose perspective should be used.[2] There has been increasing recognition of family spillover effects (i.e., the effect of one family member’s illness on other family members) on health-related quality-of-life. The potential importance of including these effects in economic analyses can be quite significant for illnesses in the very young and the very old.[8,9] Our approach of valuing changes in health-related quality-of-life for both parent and child is consistent with the inclusion of family spillover effects in the economic evaluation. Clearly more research will be needed to establish the optimal method for valuing family spillover effects.
Valuing temporary health states will continue to have increasing significance as more screening programs and preventive interventions for children that reduce morbidity rather than mortality are introduced. We offer a new approach for providing health preferences in children for whom no standardized scores exist. To our knowledge, this study is the first to address both of these methodological challenges simultaneously. Certainly more research should be done to reach consensus in the field regarding optimal methods for valuing temporary health states in children.
Lisa A. Prosser, PhD, Assistant Professor, Center for Child Health Care Studies, Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, Boston, MA
Tracy A. Lieu, MD, MPH, Associate Professor, Center for Child Health Care Studies, Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, Boston, MA
References:
1. Bala MV, Wood LL, Zarkin GA, et al. Are health states “timeless”? The case of the standard gamble method. Clin Epidemiol 1999;52(11):1047- 1053.
2. Petrou P. Methodological issues raised by preference-based approaches to measuring the health status of children. Health Econ 2002;12(8):697-702
3. Swan JS, Fryback DG, Lawrence WF, et al., A time-tradeoff method for cost-effectiveness models applied to radiology. Med Decis Making 2000;20:79-88.
4. Swan JS, Sainfort F, Lawrence WF, et al. Process utility for imaging in cerebrovascular disease. Acad Radiol 2003;10:266-274.
5. Phillips KA, Maddala T, Johnson FR. Measuring preferences for health care interventions using conjoint analysis: An application to HIV testing. Health Services Research 2002;37(6):1681-1703.
6. Johnston K, Brown J, Gerard, et al. Valuing temporary and chronic health states associated with breast screening. Social Science Medicine 1998;47:213-222
7. Bennett J, Torrance GW. Measuring health state preferences and utilities: Rating scale, time trade-off, and standard gamble techniques. In: Quality of Life and Pharmacoeconomics in Clinical Trials, Second Edition. Ed: Spilker B. Philadelphia: Lippincott-Raven Publishers. 1996.
8. Basu A, Meltzer D. Spillover effects of patient’s health on family members and its implications to cost-effectiveness analysis (abstract). Med Decis Making 2003;23:564.
9. Langa KM. An illness in the family: Accounting for the complex effects of illness on other family members. Am J Manag Care 2004;10(5):305 -306.
Prosser et al[1] use a novel approach to tackle an important question for the economic evaluation of new vaccines – how to place quantitative values on the benefits of disease prevention in children. They propose both a modified time trade-off (TTO) and willingness to pay approach to measure preferences for (often transient, temporary) vaccine preventable states of ill-health in children. The results are used to estimate costs per quality-adjusted life-year (QALY) gained and net monetary benefits, respectively.
The QALY approach using a standard TTO task, requires the respondent to state a time period (x years) such that s/he is indifferent between say, 10 years in a given health state followed by death, and x years in full health followed by death, giving a QALY weight of x/10, where a year of full health has a QALY weight of 1 and death has a QALY weight of 0. For temporary health states respondents are asked to trade-off between a time period in the given temporary health state followed by recovery to full health, and a longer (shorter) period in a less (more) severe health state followed by full health. A ‘chain’ approach is used to relate the temporary health states to full health and death.[2] Valuing health states by this method relies on well-documented, strong assumptions about individuals’ preferences, including independence of preferences for survival duration, health states and all other factors (such as non-health related consumption).[3-5] Calculating QALY weights also requires an experimental design that is explicit about health state durations and what follows them (i.e. death or another health state). We argue that Prosser et al’s experimental design seems to have ignored this, and that therefore their TTO results could only provide an ordinal ranking. We also note a number of other methodological problems.
Prosser et al asked adults (sampled from parents of children who experienced the health states under study, as well as from the general population) to state the portion of their own life that they would be willing to trade off to prevent their child, or a hypothetical child, experiencing the given health state. Respondents were explicitly asked to consider their own time spent caring for their child and associated stress, as well as the time that the child would spend suffering in the undesirable health state, and trade it off with any amount of their own (presumably healthy) life. The expected duration of their own life was not stated. While the approach seems intuitively appealing, because parents make health care decisions for their children, it is problematic for several reasons.
Adults were asked to trade off their own time, for an improvement in a child’s time. This raises issues of guilt and interview bias as noted in the discussion. Such biases could be largely avoided by asking the respondent to make trade-offs between the child’s quality of life and the child’s survival time only. Respondents would make a judgment by trading off healthy and unhealthy time for one and the same person. This would not only be more consistent with the theory on which TTO is based and more comparable with other QALY-based studies, but would allow an interpretation that is analogous to other quality of life (QOL) research in which respondents are asked to make judgments about the QOL of others.[6] This approach is also likely to increase consistency between different groups of respondents (eg parents and non-parents), and interpretability of remaining differences. Even though Prosser et al report that pilot survey respondents had difficulty separating their own from their child’s QOL (probably more so if they were a parent), by asking to trade off the child’s time in the child’s life only and the adult’s time in the adult’s life only, the response task would seem easier to comprehend and the results more straightforward to interpret. In the Prosser et al TTO task, preferences are defined over the respondent’s own QOL, the respondent’s own survival duration and over another’s (the child’s) QOL and survival duration, all of which may vary over the different scenarios. The task recognizes but does not define the nature of this interdependence. Without this it is impossible to calculate QALYs from the responses. Prosser et al are quite correct to say the responses do not represent utilities. But without specifying their assumptions about the nature of the respondent’s utility function and how others’ utilities enter it, they are incorrect to use these responses to quantify preferences for different health states.
Furthermore, even when assuming that these responses are valid for QOL measurement, Prosser et al appear to have grossly overestimated the benefits. They state that if a respondent was willing to trade off 7 days to prevent simple otitis media, this equates to a 1-time loss of 0.02 QALYs (=7/365). Such a result fails to account for the respondent’s estimate of their remaining life span. The questions in Prosser et al explicitly state that respondents should trade off any portion of their remaining life. Parents and community respondents’ average ages were 37 and 40 years respectively, and therefore could reasonably expect to survive for another 30 years. Assuming that they also expect these years to be healthy, forgoing 7 days equates to 0.00064 QALYs (=7/(365*30)). The calculation by Prosser implies that respondents would forgo 7 days of each remaining year of life. This error may explain why the costs per QALY gained, are less than one twentieth of the costs per life-year gained. This difference is orders of magnitude greater than in other comparable studies.[7-9]
Finally, Prosser et al noted an interpretation problem related to time preference, and calculated “discounted” TTO values in sensitivity analysis. In a TTO task, respondents can only forgo time at the end of the time span they trade off. This is usually made explicit by specifying a time period followed by death for chronic health states, or by full health for temporary health states. In calculating QALY weights from TTO tasks time preference is usually ignored, but QALYs gained in the future are usually discounted in economic evaluation (avoiding ‘double discounting’). Because Prosser et al have not explicitly stated which part of their lives respondents are meant to trade off, it would seem more appropriate to use the discounted values as the baseline results. But what remains unclear is whether the respondents had already taken into account their time preferences when they nominated the amount of time they were willing to trade-off. If this is the case, the responses may already have been discounted. As respondents were asked to consider their own QOL while trading off their time with that of a child’s, they may trade-off time for their own QOL in the present (i.e. a trade off purely in terms of the quality of their time assuming the child is sick now) versus time for the child’s QOL at the end of their life (i.e. a trade off in terms of sacrificing a quantity of their time for the child’s time). The associated time preferences can only be disentangled by limiting each time trade-off task to the lives which experience the undesirable time traded off. Though parents make many decisions about health care use for their children, when a child falls ill the QOL impact on the parent’s time is clearly of a different nature from the QOL impact on the child’s time. The Prosser et al experiment seems to require the respondents to imagine them to be parents (even if they are not), while a societal perspective requires preferences from the whole of society, not just the parents. These are all further arguments to not conflate these two types of preferences.
Taken together, these problems suggest that the modified TTO task introduces more problems than it solves, and that the Prosser et al cost per QALY results are invalid. The alternative willingness to pay approach is potentially more promising because it does not a priori impose assumptions about the nature of preferences. However, the interpretation of the reported results is not straightforward. In the parent sample, the median WTP for the vaccine which specified all risk reductions together was $250, but it was much higher for risk reductions of specific health states on their own ($400 for severe pneumonia and $500 for meningitis). It is not clear which of these values were used in the cost-benefit analysis, and if the specific values were used, whether these were incorporated using a purely additive model.
The QOL impact will play a significant role in the analysis of new vaccines, which are much more costly and often more aimed at reducing morbidity than mortality, in comparison to currently widely established childhood vaccines. The USA is often the first country to introduce new vaccines in their routine programs, and the economic evaluations on which US vaccine policy is based may therefore become influential throughout the world. It is, however, important for diverse audiences to appreciate the different methodological approaches, including those related to measuring preferences. In order to safeguard the credibility of economic evaluation in this field, and thus provide a basis for consistent policy making, the methods to evaluate preferences for ill-health in children need to be methodologically sound and the assumptions on which they rest must be explicit.
Philippe Beutels, PhD Senior Research Fellow The National Centre for Immunisation Research and Surveillance (NCIRS) Royal Alexandra Hospital for Children & University of Sydney and Visiting scholar Centre for Health Economics Research & Evaluation (CHERE) University of Technology, Sydney (UTS) Sydney, Australia
Rosalie C Viney, MEc Deputy Director and Senior Lecturer Centre for Health Economics Research & Evaluation (CHERE) University of Technology, Sydney (UTS) Sydney, Australia
References
1. Prosser LA, Ray GT, O'Brien M, Kleinman K, Santoli J, Lieu TA. Preferences and willingness to pay for health states prevented by pneumococcal conjugate vaccine. Pediatrics 2004;113(2):283-90. 2. Drummond M, O’Brien B, Stoddart G, Torrance G. Methods for the economic evaluation of health care programmes (second edition). Oxford: Oxford University Press, 1997, 305 pp. 3. Pliskin J Shepard D, Weinstein M. Utility functions for life years and health status. Operations Research 1980; 206-224. 4. Bleichrodt H, Wakker P and Johannesson M. Characterizing QALYs by risk neutrality. Journal of Risk and Uncertainty 1997; 15: 107-14 5. Bleichrodt H, Quiggin J. Life-cycle preferences over consumption and health: when is cost-effectiveness analysis equivalent to cost-benefit analysis? Journal of Health Economics 1999; 18: 681-708. 6. Sprangers MA, Aaronson NK. The role of health care providers and significant others in evaluating the quality of life of patients with chronic disease: a review. Journal of Clinical Epidemiology 1992, 45: 743- 60. 7. De Wals P, Petit G, Erickson LJ, Guay M, Tam T, Law B, Framarin A. Benefits and costs of immunization of children with pneumococcal conjugate vaccine in Canada. Vaccine 2003;21(25-26):3757-64. 8. Melegaro A, Edmunds WJ. Cost-effectiveness analysis of pneumococcal conjugate vaccination in England and Wales. Vaccine, in press. 9. Milne RJ, Lennon D. An economic evaluation of pneumococcal vaccination of New Zealand children less than 2 years of age. Paper presented at the 25th Australian Health Economic Society meeting, Canberra, 2nd October 2003.