Background: Patients with Alzheimer's disease (AD) who deteriorate rapidly are likely to have a poorer prognosis. There is a clear need for a clinical assessment tool to detect such a decline in newly diagnosed patients. Objective: To identify the predictive factors of rapid cognitive decline (RCD) in a cohort of patients with mild to moderate AD ; and to validate a self-questionnaire for caregivers as a diagnostic tool for rapid decline. Design and analysis: An open-label, observational, 12-month, multicenter, French study. Physicians were asked to record data of three eligible rivastigmine naïve (or on rivastigmine for < 1 year) AD patients. Risk factors of RCD and the detection power of the Détérioration Cognitive Observée scale (Deco), a 19 item self-questionnaire for caregivers, were assessed at endpoint using regression analyses. Results: Out of the 361 patients enrolled in the study, 91 (25.2%) were excluded due to loss of follow-up. Among subjects using cholinesterase inhibitors or memantine, 161 (59.6%) experienced a stabilization (29.2%) or an improvement (30.4%) in global functioning as measured by the CGI-C. Sixty of the remaining 204 patients retained for analysis (29.6%, CI 95% [23.4; 35.8]) lost three or more points on the MMSE score between the inclusion and one of the follow-up visit. In the multivariate logistic regression analysis, institutionalization, higher level of education and the loss of 3 points or more on the MMSE were found to be significant predictors of a rapid cognitive loss in this population. The threshold which maximizes the predictive values of the Deco score as a diagnostic tool of rapid cognitive decline was significantly different according to the age of the patient (below or over 75 years old). A score below 16 for patients < 75 years old and below 14 for patients ≥ 75 years old consistently predicted a RCD within the next year. Conclusion: The Deco test appears to be a simple tool to alert the physician to the possibility of an aggressive course of the disease which warrants particular management.