OBJECTIVES: To develop a scale that can assist in predicting likelihood of decline from mild dementia over 1 year in patients with Alzheimer's disease (AD). DESIGN: Retrospective cohort study. SETTING: University Memory and Aging Center. PARTICIPANTS: Patients with probable or possible AD and Clinical Dementia Rating (CDR) of 1 at baseline, divided into development and validation cohorts (n = 118 each). MEASUREMENTS: The CDR and neurological and neuropsychological assessments were given at baseline and 1 year later. RESULTS: In the development cohort, high education, low Mini-Mental State Examination score, poor insight, psychotic symptoms, and greater activity of daily living impairment predicted decline in CDR from 1 to 2 or 3. Receiver operating characteristics (ROC) curve analysis identified cutoff scores that maximized sensitivity and specificity for each significant predictor of decline. Based on the cutoff, raw scores were recoded to reflect risk for decline, weighted, and summed to create a final scale score. ROC curve analysis established a cutoff to indicate risk for decline on the final scale score. Sensitivity, specificity, and area under the ROC were 0.76, 0.74, and 0.83 in the development cohort and 0.77, 0.69, and 0.80 in the validation cohort, respectively. Positive and negative predictive values were 0.71 and 0.78 in the development cohort and 0.68 and 0.78 in the validation cohort, respectively. CONCLUSIONS: Decline from mild to moderate or severe impairment represents significant clinical change, with implications for patient and caregiver quality of life and treatment options. The clinical scale developed uses data to enhance prediction about change from mild to moderate or severe stages of AD.