OBJECTIVES: To describe the process of care of a program of in-home comprehensive geriatric assessment (CGA) and to determine: (1) if there are major findings in all domains of CGA (medical, functional, mental health, and social/ environmental), (2) if there is a continued clinical yield when CGA is repeated annually, and (3) factors that affect patient adherence with recommendations from CGA. DESIGN: Descriptive prospective study of subjects allocated to the intervention group of a 3-year randomized trial of preventive in-home CGA. SETTING: Homes of participants living in an urban setting. PARTICIPANTS: Persons aged 75 years or older and living at home who received the intervention (N = 202 subjects, mean age 80.8 years, 70% female, 95% white, 64% living alone). INTERVENTION: Annual in-home CGA and quarterly home visits by gerontologic nurse practitioners for 3 years. MEASUREMENTS: Detailed data were collected prospectively on clinical problems detected by CGA, and specific recommendations were made for these problems using an instrument developed explicitly for this project to study the CGA process. Subject adherence with these recommendations was also recorded. RESULTS: Major problems were identified in all domains of CGA; the most common problems were medical. In the first year, 76.7% of subjects had at least one major problem identified that was either previously unknown or suboptimally treated. One-third of subjects had additional major problems identified during the second and third years. A constant number of therapeutic and preventive recommendations was made each year (11.5 per subject annually). Subject adherence varied by type of recommendation (ANOVA, F = 108.4, P < .001); adherence was better for referrals to a physician than for referrals to a non-physician professional or community service or for recommendations involving self-care activities (Scheffe's test, P < .001). CONCLUSION: In these community-dwelling older people, there was a continued yield of problems identified and recommendations made when CGA was repeated annually for 3 years, supporting the practice of repeat CGA in older people in the community. Subject adherence with recommendations from CGA varied by type of recommendation, but further work is needed to determine additional factors that affect this adherence and to determine the association between the yield of CGA (i.e., problems identified, recommendations given and adherence with these recommendations) and important clinical outcomes.