The two health behaviors examined were associated
with one QOL dimension, but neither was associated with
HQOL. Less physical activity was associated with better
emotional well-being. Thus, these older adults experienced
more emotional well-being when they did not have to do
higher levels of physical activity such as heavy lifting or
strenuous activity on a regular basis. The second health
care behavior examined, greater health care utilization, was
linked to lower social functioning. This association may
have been a function of limited time for social activities
when more time is spent on contacts with heath care providers.
Another plausible explanation is that higher health
care utilization may have been a proxy for more extensive
health problems, which themselves could limit social
functioning. More in-depth examination of physical activity
using other measures as well as health care utilization
and QOL is warranted.
There were several limitations to our study. First, the
findings were based only on subjects who had complete
data. The initial sample was 1,189, but 278 had missing
data, When we compared those with missing data and those
included in the study, we did find differences. Since people
with missing data scored worse on QOL, using only the
sample with no missing data provided an estimate of better
QOL than it should be for the whole representative sample
of community-dwelling older adults. In addition, results
presented from regression analyses assumed that data were
missing at random; however, we cannot test for that, so our
results might have been different if this assumption was not
met. In our analysis to examine missing data patterns, we
found that people with missing data were older, less likely
to be married, had worse ADL function, more memory
problems, lower physical activity, and lower QOL. Second,
all our QOL measures were generated from populationbased
measures. Thus, the study was limited to testing
associations between variables reflecting computed and
transformed QOL measures. While the two item HQOL has
been validated against longer HQOL measures [27, 28], our
transformations have not been validated. Also, the twoitem
index we computed for social functioning did not
capture the multifunctionality of social functioning. It
would have been preferable for the NHANES data to
include all four domains in the Berkman–Syme Social
Network Index, including church or other religious group
membership and membership in other community organizations
[33]. Third, in the NHANES survey, memory
problems were measured by self-report, rather than an
objective measure. Thus, it is possible that older adults
underreported or were unaware of their memory problems.
Fourth, because the NHANES database limits the response
choices for the categorization of age that includes a category
of over the age of 85, it is possible that the oldest-old
are not adequately represented in our findings. Finally, we