Karl,
You asked about SF-36 / Basis in mental health applications. Assuming
that the second of these is specific to MH then your choice might be
influenced by whether your analysis is from a clinical(only), from an
economic perspective (only) or both.
In a clinical mode you might prefer to operate with a measure that has a
demonstrable clinical pedigree, of which there are hundreds from which
to choose. In addition to the condition-specific measures that abound in
MH, several of the generic measures have been used - including SF-36,
QWB and EQ-5D. You might however want to select a generic measure that
enable you to capture information that reflected patient self-assessment
of their health status, and in a UK context you'd want to exclude the
QWB. We have used both SF-36 and EQ-5D alongside GHQ-12 and Hospital
Anxiety Depression Scale, and a colleague has similar twin-track data
with SF-36/EQ-5D and SCL-90.
In a pragmatic world, your selection might be influenced by other
considerations - such as whether anyone else has used the measure
recently, or whether there is a (declared or undeclared) consensus that
a particular measure is as good as 'standard', or whether NOT having a
particular measure would jeopardise future reporting or publication
.....
If your analysis is now (or may later) involve an economic evaluation,
then your choice will be constrained by the need to represent health
outcomes in terms of a single index. Hence all profile measures (whether
generic or condition-specific) can be put to one side since by
definition they fail this test. NHP for example, was designed so as to
resist 'conversion' into a single total score, although some folk
persist in abusing this in practice. SF-36 in its original format and
SF/18 / 20 and 2) fail the test for similar reasons. Your choice for
generic measures rapidly reduces to QWB / HUI or EQ-5D, of which only
the latter is calibrated in terms of UK population preference weights.
So in summary.
Q. How do you choose ?
A. It all depends on what purpose you want to use the data for
BUT
.... in the UK / for NHS reporting / for economic evaluation
.... you ought to give serious consideration to using EQ-5D as a minimum
Paul Kind
Centre for Health Economics
University of York
|