-Background and Objectives: Dual diagnosis is associated with increased emergency department service use. During recent years, increased emphasis has been given to monitor the decision to hospitalise the most sever patients. Many rating scales have developed based on patient-related factors. To assess the level of severity of dual diagnosis patients measured by the Severity Psychiatric Illness (SPI) Scale and to evaluate it as a tool for predicting discharge in emergency psychiatric practice.Methods: Data on 1,227 consecutive admissions visited in a psychiatric emergency room of a general teaching hospital during a period of six months were collected. A routine computerised protocol was completed which included socio-demographic, clinical and social factors and the SPI scale.Results: 206 admissions (17%) had dual diagnosis disorder, 106 (9%) had substance abuse disorder (SUD), and 906 (74%) had non-substance abuse disorder (NSUD). Differences among groups were found in clinical characteristics, discharge decision, readmissions, previous psychiatric history, reason for referral, DSM-IV diagnosis and illicit drug use. Dual diagnosed patients had the highest scores of severity. Logistic regression analyses revealed the independent contribution of different SPI'items. The model showed a good fit and indicated excellent calibration in the sample studied, predicting 87.6% of discharge decisions.