Objective: Though antipsychotics were originally developed to treat schizophrenia, their use in older adults with dementia has grown substantially. Given concern about the safety of these drugs, we assessed the impact of a prior authorization (PA) policy upon use and choice of antipsychotic medication in long-term care. Methods: We conducted a retrospective cohort study using administrative data from two Canadian provinces -- one in which access to newer antipsychotics (risperidone, olanzapine, and quetiapine) was unrestricted (Ontario), and another in which access required PA (British Columbia (BC)). Subjects were all 37,057 Ontario and 13,569 BC residents aged 66 years or older who were newly admitted to a nursing home between April 1, 1998 and March 31, 2002, who had no history of schizophrenia or psychosis in the 5 years preceding admission, and who had no evidence of antipsychotic drug use in the preceding year. We assessed crude and adjusted exposure to antipsychotic medication over the year following nursing home admission, as well as the types of medications used. Results: Nineteen percent of Ontario residents were newly dispensed an antipsychotic within 100 days of nursing home admission vs. 16% in BC. Male sex, younger age, fewer comorbidities, and history of dementia all were strongly associated with receipt of an antipsychotic. Adjustment for these factors reduced the cross-provincial difference in drug use. However, fewer BC residents received newer antipsychotics, particularly after risperidone received an approved indication for the management of behavioural symptoms of dementia. Olanzapine, which required PA throughout the study, was dispensed to 11% and 3% of Ontario and BC residents, respectively. Conclusion: Although BC's PA policy had negligible impact upon the incidence of antipsychotic drug use as a whole, it appeared to influence drug choice. Questions remain about the impact of such policies upon health outcomes and costs.