Abstract Previously reported data from a randomized controlled trial showed that admission to the geriatric evaluation unit (GEU) and follow-up clinic at the Sepulveda VA Medical Center leads to significantly improved outcomes for frail elderly hospital patients-including a 50% reduction of one-year mortality ( p < 0.005). In the present paper, two-year survival curves for GEU and control groups are reported. In addition, we subdivided the population by potential baseline risk factors (both patient- and treatment-related) and examined one-year survival using 12-month survival curves and odds ratios. There is evidence for GEU-related survival effects in specific subgroups of patients (e.g. patients with heart and pulmonary disease, patients with low baseline scores in functional status and mental status, and patients with high baseline morale scores). Finally, employing stepwise logistic regression, we determined the predictors of one-year survival in the pooled study population. These factors were: assignment to the GEU (adjusted odds ratio = 2.45; p < 0.001); not having a heart diagnosis (2.24; p < 0.001); and having primarily “geriatric/rehabilitation” problems (1.95; p < 0.005). A predictive model derived from the regression defines patient subgroups likely to survive only when assigned to the GEU: cardiac patients with primarily “geriatric” or “rehabilitation” problems, and non-cardiac patients whose problems are primarily “medical”. The dramatic effect of the GEU on survival appears to be concentrated on certain identifiable subgroups of patients who might be targeted to maximize program cost-effectiveness.