BACKGROUND AND PURPOSE: Few studies have examined why some patients with dementia stop attending medical consultations. We conducted a retrospective study to investigate factors associated with discontinuous clinic attendance. METHODS: Participants were 988 patients with dementia from university hospital (UH) clinics and affiliated local hospital (LH) clinics. We compared continuous and discontinuous attenders on cognitive and affective functions and activities of daily living (ADL), and also compared UH and LH patients (UH: continuous, n = 176; discontinuous, n = 207; LH: continuous, n = 418; discontinuous, n = 187). RESULTS: The total annual rate of discontinuation was 8.0%, and the mean period of attendance before discontinuation was 2.2 ± 2.4 years (UH, 2.8 ± 3.0; LH, 1.5 ± 1.3, P < 0.01). Scores for the Mini-Mental State Examination, Hasegawa Dementia Scale - Revised, Geriatric Depression Scale, apathy scale, Abe's behavioral and psychological symptoms of dementia (BPSD) score, and ADL were significantly worse in the discontinuous group than the continuous group for both UH and LH patients (P < 0.01). The best predictor of discontinuation was ADL decline (UH and LH) and Abe's BPSD score (UH). The most common reason for discontinuation was returning to the family doctor (39.1% for UH), and cessation of hospital attendance at their own discretion (35.3% for LH). CONCLUSIONS: We identified the main reasons for discontinuation of attendance as returning to the family doctor and cessation of hospital attendance at their own discretion. The best predictors of discontinuation were ADL decline and worsening BPSD. There were significant differences in discontinuation between UH and LH patients with dementia.