A statewide probability sample of 1,625 elders living in Massachusetts are studied prospectively to identify key determinants of long-term care (LTC) institutionalization. One-hundred forty-seven elders, 9 per cent of the original cohort, entered a LTC institution during the six-year investigation. Using logistic multiple regression, we examine the predictive power of 19 independent variables grouped into six categories: demographic characteristics, attitude, social context, long-term care needs, physical disability, and mental/emotional disability. Five variables are significantly related to institutionalization: advancing age, using ambulatory aids, mental disorientation, living alone, and using assistance to perform "instrumental" ADL (activities of daily living). These results may be helpful to those trying to target non-institutional services to elders for use as substitutes for institutional long-term care. They may also help explain why recent experimental tests of substituting non-institutional care for institutional services have been less than successful.