| |
The relationship between
health status and directly elicited utility for current health
is complex.
Revicki found that
physical function correlated with a categorical rating scale (RS)
measure of utility, but not with the standard gamble (SG) (Revicki,
1992). Revicki explained the difference by the introduction
of risk with the SG. That is, in order to provide a rating that
is meaningful in the VNM utility theory context, the SG requires
respondents to consider to what extent they will put life at risk
to avoid the health state being rated (Weinstein
and Fineberg, 1980), a very different task from the RS.
Current health status
is only a partial predictor of current health utility (Llewellyn-Thomas,
Sutherland et al., 1991; Tsevat,
Goldman et al., 1991). Among survivors of myocardial infarction,
the TTO correlated poorly with measures of cardiac functional
status (Tsevat,
Goldman et al., 1991). Over time, their TTO utilities remained
stable and did not correlate with changes in Karnofsky index or
measures of cardiac function (Tsevat,
Goldman et al., 1993).
In a survey of seriously
ill hospitalized patients, Tsevat and co-workers found only weak
correlations between current health status and current health
utility. There was a weak association between changes in TTO scores
over time and changes in ability to perform ADLs (Tsevat,
Cook et al., 1995).
Those two studies corroborate
many others showing that the relationship between one's current
state of health and one's utility for that health state is at
best modest (Bombardier,
Ware et al., 1986; Churchill,
Torrance et al., 1987; Llewellyn-Thomas,
Sutherland et al., 1991; Tsevat,
Goldman et al., 1991; Llewellyn-Thomas,
Thiel et al., 1992; Revicki 1992; Fryback, Dasbach et al.,
1993; Tsevat, Goldman et al., 1993; Fowler, Cleary et al., 1995;
Nease, Kneeland et al., 1995; Revicki, Wu et al., 1995; Tsevat,
Cook et al., 1995; Bosch and Hunink, 1996; Chen, Daley et al.,
1996; Nease, Tsai et al., 1996; Nichol, Llewellyn-Thomas et al.,
1996; Stigglebout, de Haes et al., 1996; Tsevat, Solzan et al.,
1996; Perez, McGee et al., 1997; Bartman, Rosen et al., 1998;
Bult, Hunink et al., 1998; Patrick, Mathias et al., 1998)
There is no a priori
reason that health utilities should correlate strongly with health
status. Many individuals with health problems prefer to live as
long as possible, even with the health problems, rather than trade
time for perfect health, as health may not be the only thing that
matters to everyone (Tsevat, 2000).
Utility for current
health is multifactorial. It is affected by, but not well predicted
by, each of several factors, including major diagnosis, number
of chronic conditions, affective state, and functional status.
Previous studies that have examined the correlation of physical
function with utility for current health have relied on summary
scores of physical function and thus have not provided a mapping
of specific states of ADL impairment to utility for life with
those impairments. Methods developed by Goldstein and colleagues,
as described below, address these problems by using the theoretically
preferred SG rather than the TTO, and by providing detailed descriptions
of ADL dependency to allow for mapping to utility.
|