Patients with head and neck cancer are treated through surgery, radiotherapy (RT), and chemotherapy (CT). Carotid artery damage and neurotoxicity were previously observed in these patients. This study estimated the dementia risk associated with different treatment modalities in a head and neck cancer population with long-term follow-up. Taiwan's National Health Insurance claims database and a cancer registry database from the Collaboration Center of Health Information Application were linked for the present analysis. Patients with head and neck cancer, treated from January 1, 2002 to December 31, 2010, were included in the study. The follow-up duration was the period from the index date to December 31, 2012. Inclusion criteria were head and neck cancer; an age >20 years; and having undergone surgery, CT, concurrent CT, or surgery with adjuvant treatment. Exclusion criteria were another cancer diagnosed before the head and neck cancer, death or being diagnosed with dementia within 2 years after the treatment of the head and neck cancer, stroke before the index date, distant metastasis, in situ carcinoma, sarcoma, head and neck cancer recurrence, an unknown sex, and an age <20 years. In total, 20,135 patients were included. In patient groups that underwent surgery alone, surgery and adjuvant chemoradiotherapy, and chemoradiotherapy alone, the dementia incidence per 1000 person-years was 1.44, 1.04, and 1.98, respectively. The crude hazard ratio (HR) of dementia was 1.84 (95% confidence interval [CI] 1.21-2.81) in the RT with or without CT group. After adjustment for age, sex, clinical stage, and comorbidity, the HR was 1.92 (95% CI 1.14-3.24). Examining the dementia risk in patients who received different treatment modalities according to the Cox proportional-hazard model revealed that an age >65 years and having undergone RT with or without CT were risk factors (P < 0.001 and P = 0.015; and HRs of 16.5 and 1.92, respectively). The dementia risk in patients at different clinical stages was not significantly different among the various treatment groups, regardless of whether the patients received RT. However, younger (<65 y) patients who received RT with or without CT had a 2.96-fold (95% CI 1.24-7.08) higher risk of dementia and a 3.54-fold (95% CI 1.32-9.51) higher adjusted HR compared with the surgery-alone group. Patients who received a total radiation dose >6660 cGy exhibited a 1.69-fold (95% CI 0.97-2.95, P = 0.063) higher dementia risk compared with those who received a total radiation dose <6660 cGy. Receiving a higher radiation dose increased the dementia risk and persistently escalated the dementia incidence even 9 years after RT. Younger (<65 y) patients have a high risk of dementia after RT. The selection of young patients for dose de-escalation requires improvement for reducing irradiation to the neck and areas near brain tissues, particularly in Taiwan, where the median patient age is 53 years.