A major problem with alcoholism treatment is the high rate of early recidivism to drinking and re-admission for alcoholism treatment. The objective of this study was to retest a model or predict early (within 6 months) re-admission to alcoholism treatment using a second independent sample. Additionally, we compared a high-risk alcoholism relapse (HAR) model (defined by chronicity of heavy drinking, daily alcohol consumption and previous treatment history) with three previously defined alcoholism typologies for descriptive and predictive validity. Male alcoholics (N = 299) admitted for treatment at a Veterans Affairs inpatient treatment program were interviewed and then followed for 6 months after discharge. The HAR model identified 107 (35.8%) alcoholics at high-risk for relapse prior to discharge. Of the HAR group 61% were re-admitted within 6 months compared to 28% of the low-risk alcoholism relapse (LAR) group (OR = 4.0, 95% CI = 2.4-6.8). The HAR group was older with a lower socioeconomic status, fewer legal problems, more physical and mental health problems and decreased evidence of social support. The HAR model was more successful than were the typologies for predicting early relapse. The HAR model demonstrates descriptive and predictive validity and compares favorably to existing typology models.