The strengths of our study include high generalizability due to the use of nationally representative data involving
non-institutionalized patients. MEPS collects information on prescription medication use that is validated by the
pharmacies where the prescriptions were purchased, enhancing the validity of these data over standard survey
data [24]. However, the current study findings should be seen in the light of several limitations. First, this study
used cross-sectional data; therefore, causal relationship cannot be established between factors and anticholinergic
medication use. Second, the study used a secondary database, so the possibility of data collection or data entry
errors could not be ruled out. Additionally, the study did not examine the doses of anticholinergic medications
received by the patients; however, a medication analysis by Carnahan et al. [26] found that adding dose-related
information to high-level anticholinergics did not improve predictability of SAA with ADS. There is also lack of
information on some potential predictors in the dataset,such as patient perceptions, physician characteristics, and
certain comorbidities so these predictors could not be analyzed in the present study. A post hoc power calculation
was conducted to investigate the non-significant association for the predisposing, enabling, and need factors
in the final multivariable model. It was found that, except for FPL, education, and usual source of care, our study was
sufficiently powered (power[0.80) for the other variables. Because MEPS is a self-reported data, we cannot rule out
the possibility of data reporting errors and also, we cannot differentiate between proxy or self-reported data. Finally,
the study did not evaluate clinical outcomes of anticholinergic use