Background: Donepezil is indicated for the management of mild to moderate dementia, particularly in Alzheimer's disease. Several studies have described low adherence rates with donepezil. Aim: To examine and measure donepezil adherence, persistence and time to first discontinuation in older New Zealanders. Methods: An inception cohort of 1,999 new users of donepezil, aged 65 years or older, were identified from the Pharmaceutical Collections and National Minimum Dataset from 1 November 2010 to 31 December 2013. Kaplan-Meier curves and Cox regression analysis were used to estimate the cumulative probability and risk of time to first discontinuation of donepezil therapy. Results: The mean age of the cohort was 79.5 ± 6.4 years and included 42.7% females. Adherence was high (89.0%), while the proportion of donepezil dispensings (81.0-32.5%) declined between 6 and 36 months. Persistence between the 1st and 6th dispensing visit decreased by 19.0%, and 11.0% of the total cohort had a gap of 31 days or more. The adjusted risk of time to first discontinuation in the non-adherent group was 2.2 times (95% CI 1.9-2.6) that of the adherent group. Conclusions: The non-adherent new donepezil users, on average, discontinued faster than the adherent group. Time to first discontinuation in this study was higher compared to discontinuation rates observed in clinical trials.