The decision to start long-term tube-feeding in elderly people is complex. The process by which such decisions are made is not well understood. The authors examined the factors involved in the decision to start long-term tube-feeding in cognitively impaired older people from the perspective of the substitute decision-maker. A telephone survey was administered to the substitute decision-makers of tube-fed patients over 65 years old in chronic care facilities in Ottawa. Subjects were recruited from September 1997 to March 1998. Patients were incapable of making their own decisions about tube-feeding. Data were collected on sociodemographic factors, patients' health status, advance directives, communication between the substitute decision-maker and the health care team, and the decision-maker's perceived goals of tube-feeding and satisfaction with the decision regarding tube-feeding. Among the 57 cases in which the patient was eligible for inclusion in the study, 46 substitute decision-makers agreed to participate. Most of the patients had not given advance directives, and only 26 substitute decision-makers (56.5%) were confident that the patient would want to be tube-fed. A physician spoke with the substitute decision-maker about tube-feeding for 15 minutes or less in 17 cases (37.0%) and not at all in 13 cases (28.3%). Most of the substitute decision-makers (39 [84.8%]) felt that they understood the benefits of tube-feeding, but less than half (21 [45.7%]) felt that they understood the risks. The prevention of aspiration and the prolongation of life were the medical benefits most often cited as reasons for tube-feeding. Just over half (24 [52.2%]) of the substitute decision-makers felt that they had received adequate support from the health care team in making the decision. Substitute decision-makers of patients less than 75 years old were more likely than those of older patients to feel supported (odds ratio [OR] 4.2, 95% confidence interval [CI] 1.0-17.9). Compared with the physician's making the decision independently, substitute decision-makers felt more supported if they primarily made the decision (OR 16.5, 95% CI 2.7-101.4) or if they made the decision together with the physician (OR 5.3, 95% CI 1.0-27.9). Most (20 [43.5%]) of the substitute decision-makers did not feel that tube-feeding improved the patient's quality of life, and less than half (21 [45.7%]) indicated that they would choose the intervention for themselves. The substitute decision-making process for tube-feeding in cognitively impaired elderly people is limited by a need for advance directives, lack of confidence in substituted judgement and poor communication of information to the substitute decision-maker by the health care team.