Page 60 - Htain Manual

With a binary scale, people may prefer to be alive or not; or have a disease or not; but

that is the extent to which their preferences may be known. A scale with ordinal properties is

one where the preferences or choices are ranked in some set increasing or decreasing order.

The catch however is that we know that the preferences are ranked by some difference
between themselves, but not by how much. A cardinal scale eliminates that as it can have

either interval or ratio properties. Now another issue arises that the interval would mean that

equal intervals do exist between the preferences but we cannot make any assumptions about
the absolute size of these preferences. An example of this is that on a scale that has numbers

10, 20, 30 ad 40; we know there exists a specific interval of 10 points but that does not clear

the picture of the absolute change in this interval – is the change of 10 arithmetic or

exponential, etc. Therefore, it is recommended to use ratio scales as these have a true zero

point and are able to compare the size of numbers rather than just the size of the change.

Thirdly, the health measure should be able to reflect the preferences, either of

individual patients or of the public. Another set of characteristics derived from principles of

psychometrics are summarized below:

1. Reliability: how much can the measure produce repeated results from an unchanged

population with minimum random error?

2. Validity: to what extent does the measure capture what it aims to quantify. It may be

in the form of content validity (the appropriateness of items within a tool) or construct

validity (extent to which results correlate with other indicators, measures or concepts
of interest).

3. Practicality: the measure should be acceptable to respondents and ethics committees

as well as being easy to administer, score and interpret findings.
4. Responsiveness: a measure should be able to detect clinically and socially meaningful

changes in the health status over time. In most cases an ‘effect size’ is calculated,

which is the difference between the mean baselines and mean follow-up scores

divided by the standard deviation of the baseline scores.

Now coming to the outcomes of concern (apart from PROs) to us while conducting an
HTA or an economic evaluation. Quality adjusted life years (QALYs), disability adjusted life

years (DALYs) and life years gained (LYG) are all common outcome measures in economic

evaluations of health interventions. While LYG is a pure measure of mortality, QALYs and

48