However, there are limitations to the study. First, a
convenience sampling, rather than a probability sampling,
was used. Although the method of cluster
randomization by study sites into different groups was
applied, without random sampling it was possible that
some older adults who met the sample selection
criteria but represented significant differences from
the sample studied were not recruited because they
were not in the study sites. Second, not all of the participants
in the clusters participated in the study and
clustering was not considered in the analysis, which
may weaken its statistical power and diminish the
reliability of the results. A truly randomized controlled
trial with participants individually randomized to
intervention or control groups might provide more
robust data. Finally, the demographics of the participants
across study sites and in the two groups had no
significant differences; however, four of the nine
outcome variables had significant differences between
the two groups in the baseline data. Although analysis
of covariance was applied in the data analysis to offset
the group differences at baseline and adjusted means
were used for the group comparisons, these differences
might have had an impact on the results of the study
and warrant further investigations of these dynamics.