Aim: The aims of the present study were to report on the psychometric properties of the Pain Assessment in Advanced Dementia (PAINAD) scale by comparing it with the gold standard method of self-reporting on a numerical rating scale (NRS), and to provide a categorical version of the PAINAD scale comparable with the verbal descriptor scale of the NRS. Methods: Six hundred elderly patients with various degrees of cognitive impairment consecutively admitted to the acute geriatric section at Padua University were evaluated. Cognitive, functional, and health statuses were evaluated using the Mini-Mental State Examination (MMSE), activities of daily living, and the Cumulative Illness Rating Scale (CIRS), respectively. Pain measurements were obtained by administering the NRS and the PAINAD scale. Results: Cognitive decline was recorded in 310 subjects (52%). The internal reliability of the PAINAD scale was adequate for all items, both in patients with dementia (α = 0.90) and in those with no cognitive impairment (α = 0.94). The psychometric evaluation demonstrated a stronger level of concurrent validity (Kendall’s τ = 0.73, p < 0.0001) and inter-rater agreement (ĸ = 0.74, p < 0.0001) for the PAINAD compared with the NRS. Conclusion: Our findings clearly indicate that the PAINAD scale is a reliable and easily administered tool for assessing pain intensity also in elderly patients with advanced dementia.