Several clinical and pathologic factors appear to affect melanoma recurrence and survival. While much attention has been directed at identifying prognostic factors, few researchers have developed predictive models for survival and recurrence. Two major clinical questions are of interest in the management of melanoma: 1) what is the patient's chance of surviving for a given period, e.g., 5 or 10 years, after diagnosis of melanoma; and 2) after a patient has been disease free for a period of time, e.g., 5 years, what is his or her chance of melanoma recurrence or death in the following interval, e.g., 5 years or 10 years. In this paper, a generalized multivariate prognostic model to address both of these clinical questions is presented. Tables of the estimated probabilities of melanoma recurrence and death for prognostic subgroups are shown to facilitate prediction of an individual patient's outcome. The model was based on a database of 4,568 patients with localized melanoma, one of the largest melanoma databases in the world with detailed clinical and pathologic information, and long-term follow-up. Tumor thickness at diagnosis was the single most important prognostic factor for all outcomes. Tumor ulceration, Clark's level, lesion location, and sex had an impact on overall survival from diagnosis for some of the subgroups defined by tumor thickness. Tumor thickness at diagnosis was strongly indicative of melanoma recurrence and death even after a disease free interval of 2, 5, or 10 years. Lesion location and ulceration were of prognostic importance after disease free intervals up to 5 years, but their impact on melanoma recurrence and death diminished after longer disease free intervals. Prediction models for melanoma outcome at diagnosis and after a disease free period can provide useful information to clinicians in the management of melanoma patients. Utilization of the model will be valuable in identifying patients at high risk for melanoma recurrence and death. © 1992 the Société Internationale de Chirurgie.