Descriptive analyses provided estimates of
pharmacologic treatment of non-cancer pain
documented on two consecutive MDS assessments,
including frequency and severity of
pain. To evaluate factors associated with receipt
of any scheduled analgesic (yes/no), we developed
a logistic regression model. Before constructing
a model, we evaluated the potential
for (and ruled out) multicollinearity among
the potential factors of interest. Logistic regression
models provided estimates of the independent
association of the determinants of interest
and the prevalence of scheduled use of analgesics
while simultaneously adjusting for the clustering
effects owing to the correlation of
residents living within the same home.25 We
assumed that resident characteristics were similarly
related to the outcomes across all facilities.
This method adjusted for confounding effects
by NH and state. We derived the adjusted odds
ratios (AORs) and corresponding 95% confidence
intervals (CIs) from the final model.