BACKGROUND Most studies on readmission have focused on patients with severe mental disorders such as schizophrenia and subjects were mostly collected in traditional psychiatric settings. This study investigated the predictors of readmission to a medical-psychiatric unit among patients with minor mental disorders in a general hospital. METHODS All 164 patients admitted to a stress ward within a one-year period were recruited into this study. Data were retrospectively obtained by reviewing medical charts. Variables used to analyze and compare between patients who were and were not readmitted included patient characteristics, clinical features and multiple axes diagnoses of DSM-IV. RESULTS Of 164 patients recruited, 135 were hospitalized only one time. The remaining 29 patients accounted for 61 admissions. The readmission rate was 17.7% (29/164). The significant predictors for readmission included Global Assessment of Functioning scores lower than 50 at admission, age under 40 years, previous psychiatric admission, alcohol dependence/abuse, dysthymia and comorbidity of borderline personality disorder. Using logistic regression to predict readmission within one year, we found the first four variables accurately differentiated 84.15% of patients who were readmitted from those who were not. CONCLUSIONS Our results indicated the incidence of relapse in certain minor mental disorders is similar to that of severe mental disorders. By identifying predictors of readmission and intervening appropriately, unnecessary readmission may be prevented.