To determine which of the many clinical parameters routinely collected influence mortality in patients with low left ventricular ejection fraction (LVEF) (< 45% at radionuclide ventriculography), 128 elderly patients (mean age 79 +/- 3 years) with various heart diseases were prospectively followed for 3 years. Twenty-eight-percent had coronary heart disease, 16% hypertensive heart disease, 7% valvular heart disease. The remaining 62 patients (48%) made up a group comprising patients with primitive cardiomyopathy, cor pulmonary with no evidence of coronary heart disease, valvular disease or hypertensive heart disease. Thirty-four-percent of all patients were classified as having congestive heart failure (CHF). Age, sex and 37 clinical variables were analyzed using a Cox proportional model. Forty-four patients died, 36 (82%) of sudden cardiac death. Ten characteristics at study entry predicted an increased mortality risk: S3 gallop, number of clinical signs >or= 3, LVEF <or= 25%, New York Heart Association (NYHA) class >or= III, dyspnea, digoxin treatment, rales, number of symptoms >or= 4, asthenia, associated pulmonary disease. Long-term survival of very elderly patients with low ejection fraction is related to the functional capacity, the severity of symptoms and the number of clinical signs. Moreover a LVEF <or= 25% selects a subgroup of patients at higher risk. Our results suggest that these variables may influence the long-term survival of elderly patients with heart disease. Further studies with a greater number of patients are necessary to better delineate the prognostic value of the clinical and instrumental variables routinely collected in these patients.