This study examines housing status at the time of hospital discharge in a national sample of 3,502 veterans who were homeless at admission to a Veterans Health Administration (VHA) medical center. A supplemental survey on homelessness was added to the annual VHA inpatient census over a 4-year period (1995 through 1998). Data from this survey were used to identify homeless veterans hospitalized on the final day of the fiscal year and to document their housing status at discharge. A signal detection technique, Receiver Operating Characteristic analysis, was used to identify predictors of housing status at discharge. At discharge, only 13% of the veterans in our analytic sample were literally homeless, 40% were doubled up, and 33% were transferred to another an institution; only 13% were living independently. Housing status at admission (specifically, staying with friends or family temporarily rather than literally homeless), treatment in a psychiatric or substance abuse program rather than a medical program, and a greater income level or access to VHA benefits were found to be associated with better housing outcomes. As one might expect, homelessness was infrequently resolved during an acute inpatient hospitalization, but few veterans returned to literal homelessness at the time of discharge.