Researchers generally agree that Geriatric Evaluation and Management (GEM) Units are effective only when they are targeted at a specific group of frail, elderly patients who are most likely to benefit. Such patients are those who are neither too sick (eg, severely demented or moribund) nor too well. Various strategies for identifying such patients have been employed by investigators with little consensus on the most efficient targeting criteria. Criteria most often used for inclusion in GEM programs are various combinations of patient age, degree of functional impairment, presence of geriatric conditions (eg falls, incontinence, confusion), particular diagnostic conditions (eg, multiple disorders), and psychosocial conditions (eg, living alone, recent bereavement, low income). Commonly used exclusion factors are severe dementia, inevitable nursing home placement, and terminal illness. Outcome studies suggest that beneficial effects of GEM care are most apparent when patients are selected using specific clinical criteria. Future research on targeting should address the potential need for differing criteria in different settings (eg, inpatient vs outpatient GEM units), simplifications of criteria for greatest ease of application, and prospective evaluation of which criteria best predict functional improvement, longer survival, and reduced health care expenditures in response to GEM care.