OBJECTIVES:  Potentially inappropriate medication (PIM) use contributes to increased morbidity and mortality in dementia patients. Prior studies have focused on specific classes of PIMs (e.g. strong anticholinergics) or on nursing home dementia patients. No study has comprehensively evaluated PIM use in community-dwelling dementia patients. The current study thus used Medicare database to examine the prevalence and predictors of PIMs in elderly dementia patients. METHODS:  This retrospective cohort study used the 5% national Medicare data from 2011-2012. The cohort included elderly patients diagnosed with dementia in the baseline year, i.e. 2011. PIMs were defined in the follow-up year, i.e. 2012, using the American Geriatric Society Beers criteria for PIM use due to drug-disease interactions. Predictors were identified in the baseline year based on the Andersen Behavioral Model, and included predisposing (sociodemographic), enabling (dual eligibility) and need factors (Elixhauser comorbidities, medication use and healthcare utilization). Descriptive statistics was used to determine the prevalence of PIMs. Multivariable logistic regression analysis was used to determine predictors of PIMs in dementia patients. RESULTS:  The cohort included 57,469 elderly dementia patients. The mean age was 85±8 years, and most patients were females (77%) and non-Hispanic whites (82%). Overall, 53.1% of dementia patients received PIMs. The prevalence of different classes were as follows: antipsychotics (31.3%), H2-receptor antagonists (11.3%), antihistamines (10.3%), antimuscarinic urinary incontinence (9.1%), antiemetics (6.7%), nonbenzodiazepine receptor agonist hypnotics (6.1%), tricyclic antidepressants (5.7%), antispasmodic (3.3%), skeletal muscle relaxants (1.6%), antiparkinsons (1.5%), benzodiazepines (1.1%). Mutivariable logistic regression found that females (odds ratio [OR], 1.16), Blacks (OR, 1.18), patients with Elixhauser comorbidities (twelve conditions), emergency room visit (OR, 1.1) and more than five prescription medications (OR, 3.0) were associated with higher likelihood of receiving PIMs. CONCLUSIONS:  One out of two dementia patients received at least one PIMs. Predictors identified in the study can be targeted to reduce PIM use in dementia patients.