A study was conducted in Andhra Pradesh, at the Institute of Health Systems (IHS), to measure peoples’ preferences about various health states. The broad goals of this study were:

Two distinct sources of assessment were made. One arm of the study consisted of a series of multi-method deliberative health state valuation (MDHSV) workshops for educated persons from different backgrounds. Participants of these workshops valued health states using more than one procedure card sort, including, visual analogue scale, time tradeoff, and person tradeoff methods. The second arm of the study involved measurement of valuations given by general population drawn from a rural area, done by conducting household surveys in Kondakkal village in Ranga Reddy district of Andhra Pradesh. Respondents in the community survey were requested to give their valuations using card sort followed by visual analogue scales.

Although synthetic health status measures would require health state weights for a large number of conditions, a health state valuation study has its limitations in generating direct valuations for a large number of conditions. Health state valuation exercises are usually characterised by high cognitive load on the valuers. Hence a valuer is usually given a limited set of conditions. Increasing the number of conditions would lead to valuer fatigue, resulting in poor validity and reliability of the measurements. A practical alternative is to use a small set of conditions to measure peoples’ preference and then statistically infer the health state values for other states. The small set of conditions used to gather measurements on peoples, valuation of health states, is referred to as index health states or synonymously indicator conditions.

Before proceeding with the selection of indicator conditions, we take stock of principles and guidelines relevant for choice of index health states. Parts of the health state valuation exercise require the valuer to deliberate about all health states at hand and assign values to each. Hence, the number of health states presented to a single valuer should be kept within manageable limits to minimise incidence of cognitive overload. This limit is a function of human working memory and cognitive behaviour, and is independent of sample size. The limit is all the more significant for population-based surveys, where multiple sessions would not usually be feasible. Gudex et al (1996) limited the number of health states per valuer to 15, for a population-based survey in UK. Sacket and Torrance included 10 health states for their population based survey in Hamilton, Otario Canada (Sackett and Torrance, 1978). The EuroQol group (Brooks, 1996) found, it was feasible for a person to value about 12-16 health states using the EuroQol instrument.

Secondly, study design and data analysis considerations should be kept in view. For example, the set of index health states should represent widest range of health state profiles. This will allow statistical inference of health state weights for conditions for which no direct measurement is available. Hence it is important that the set of indicator conditions maximise the independent variation in each dimension of health status. The need to maximise independent variation along each dimension and the constraint imposed by the ability of valuers in dealing with a number of conditions, run counter to each other. Since the limit on number of health states for valuation by an individual is fixed on psychometric grounds, the only scope to increase variation in dimensions of health status is to increase sample size and present different sets of conditions to different persons. Comparability and linkage between different health states would require that some health states be common for all valuation sets: these are known as core health states. The inclusion of a few pairs of dominating and dominated health states will facilitate inference about the validity of measurement protocol. A state that is unequivocally worse than another state in at least one dimension and at most equivalent in all others, is considered a dominated condition. One that is unequivocally better than another state in at least one dimension, and at the least, equivalent in all, is considered a dominating condition. Examples of dominating and dominated pairs of health states are: amputation of one leg below the knee, and amputation of the both the legs below the knee, quadriplegia and paraplegia. Then the proportion of times that ranks for dominated states are inverted can be used as a measure of (or lack of) the health state comprehension.

Finally, local relevance and familiarity of health states and its effect on valuer motivation has to be considered. Health states corresponding to diseases not found or found very rarely in the local population should be avoided. Since a descriptive label for each health state is retained to facilitate synthetic comprehension by the valuer, going for conditions that are hitherto unheard-of would not be helpful. Such conditions might demotivate valuers due to lack of apparent purpose of the exercise. Instead, at least some health states corresponding to diseases highly prevalent and / or considered important public health problems in the local area should be included. This is required to increase valuer's motivation and allow for a sense of purpose in the valuation exercise. On the other hand, health states associated with disease labels known to be a taboo, to provoke a sense of outrage or subject to widely prevalent stereotypes, should be avoided. The disease label is likely to dominate the valuer’s thought process to near total exclusion of the health state profile. There would appear to be some conflict between the need to avoid stereotypes and the goal of seeking locally prevalent health states, described earlier. Indeed, there would be some diseases which are highly prevalent and carry a strong stereotype. So, quite naturally, the art of selecting indicator conditions requires careful weighing of potential confounding due to various factors, and the feasibility of getting valuer’s co-operation.

The number of health states to be valued by a single valuer was limited to 11, to minimise incidence of cognitive overload, and valuer fatigue. The study, however, sought to directly measure disability weights for a larger set of conditions. According to original plan, 22 health states were chosen for the study, apart from the valuer’s own health state. Six of these health states were used as the core, common to all valuers. The remaining 16 health states were divided into four subsets. Each subset added to the core subset, made a set of health states. Accordingly, the study planned for four sets of health states. The six core conditions were selected to represent a broad range of 6D5L profile and disability severity weights. While making up these lists, conditions known to be prevalent in Andhra Pradesh were included, and conditions not found in AP were excluded, while trying to keep the list as comparable with the ones used at other study sites for international comparability. Thus, the selection of indicator conditions involved several rounds of discussion between local investigators at IHS and study co-ordinators at WHO. Sets were systematically assigned to balance assignment of valuers to different sets of health states.

Previous                     Next