Sociodemographic, health, and health care utilization data on a large representative sample of elderly and multiple logistic regression were used to compare persons making no visits to physicians for 2 years with those making few (one to three) visits but in contact with health practitioners. Results suggest that elderly nonusers are more likely than low users to be single, to have some degree of mental impairment, and to have low educational attainment. Using Cox's proportional hazards model to compare outcomes over the next 7 years, no differences were found in the subsequent hospitalization rate of the two groups, but nonusers were at greater risk of a hospitalization episode of 16+ days and appeared to die sooner than low users. However, they were at no greater risk of poor health outcomes than elderly making four or more physician visits in 2 years. The policy implications of the findings are discussed.