Abstract
Quality Adjusted Life Years (QALYs) are a popular outcome measure in cost-effectiveness analyses. QALYs are computed by multiplying follow-up or survival by a scaling factor reflecting health related quality of life, and as such capture quantity and quality gains simultaneously. Issues with regard to the validity of QALYs have been
... read more
the focus of ongoing debate. In this thesis we focus on the validity of QALYs for short-term pain. It was hypothesized that the decomposed Standard QALY Approach (SQA) would result in an underestimation of the impact of short-term pain. Unique annual profiles, containing short episodes of pain (1 week-1 year), were constructed and holistically (meaning: as a whole) valuated by means of Time Trade Off (TTO). 60 respondents each valuated one profile. Standard QALYs were calculated for these 60 profiles as well. If our hypothesis were correct, we should find respondents to be willing to make large trade offs for the profiles, resulting in lower holistic than standard QALYs. It was however observed that a holistic approach did not result in systematically lower QALYs than SQA. The latter only appeared to underestimate the impact of episodes with relatively long pain. Holistic valuations from 68 respondents for a total of 711 unique annual and 103 lifelong profiles were collected as well. For the annual profiles it was analyzed how much time respondents were willing to trade off, and what this meant in terms of implicit values for the pain episodes within the profiles (if one would want SQA to yield similar results as the holistic approach). At this it is important to not that trade offs that exceed the duration of the pain episodes within the profiles imply (extremely) negative values for such episodes. Generally a lower bound of -1 is applied for values. It was observed that exceeding trade offs were made for 181 profiles (25%) by 69% of the respondents. The mean time traded off for high intensity pain did exceed the duration of the pain to a considerable extent, resulting in implicit pain values of up to -25 for pain patients and up to -4 for subjects without pain. For a large number of profiles (274) no time at all was traded off (threshold effect). Usually common regression techniques are applied while modeling valuation data, though, considering the data distribution, techniques explicitly taking account of the threshold effect might be more appropriate in the context of TTO values. Different strategies were compared while modeling the 814 profile valuations and indeed Tobit and two-part approaches resulted in better fit than common regression. Furthermore in this thesis attention is given to the effect of the time horizon applied in the holistic TTO task (respondents' individual life expectancies) on the time traded off. In line with our expectations, time traded off diminished as respondents' life expectancies became shorter. Finally the high cognitive burden associated with trade off tasks and it's consequences in terms of generalizability and validity of results is critically looked at.
show less
