OBJECTIVES: To develop a simple method for identifying community‐dwelling vulnerable older people, defined as persons age 65 and older at increased risk of death or functional decline. To assess whether self‐reported diagnoses and conditions add predictive ability to a function‐based survey. DESIGN: Analysis of longitudinal survey data. SETTING: A nationally representative community‐based survey. PARTICIPANTS: Six thousand two hundred five Medicare beneficiaries age 65 and older. MEASUREMENTS: Bivariate and multivariate analyses of the Medicare Current Beneficiary Survey; development and comparison of scoring systems that use age, function, and self‐reported diagnoses to predict future death and functional decline. RESULTS: A multivariate model using function, self‐rated health, and age to predict death or functional decline was only slightly improved when self‐reported diagnoses and conditions were included as predictors and was significantly better than a model using age plus self‐reported diagnoses alone. These analyses provide the basis for a 13‐item function‐based scoring system that considers age, self‐rated health, limitation in physical function, and functional disabilities. A score of ≥3 targeted 32% of this nationally representative sample as vulnerable. This targeted group had 4.2 times the risk of death or functional decline over a 2‐year period compared with those with scores <3. The receiver operating characteristics curve had an area of .78. An alternative scoring system that included self‐reported diagnoses did not substantially improve predictive ability when compared with a function‐based scoring system. CONCLUSIONS: A function‐based targeting system effectively and efficiently identifies older people at risk of functional decline and death. Self‐reported diagnoses and conditions, when added to the system, do not enhance predictive ability. The function‐based targeting system relies on self‐report and is easily transported across care settings.