The aim of this paper was to explore the factors necessitating psychiatric hospital care in a Finnish multi-centre study of general hospital in-patients referred for psychiatric consultation. The study group consisted of 1251 patients referred to psychiatric hospital (n = 181) and a comparison group (n = 1070) consisting of subjects who were not referred. Differences between groups were studied by univariate analysis. Logistic regression analysis was used both to assess the factors contributing to referral to psychiatric hospital and to create predictive models. The validity of the models was analysed by means of receiver operating characteristic (ROC) curves in an independent sample. Psychiatric hospital care during the previous 5 years was associated with a 3.7-fold (odds ratio) increased risk of hospitalization. A diagnosis of psychosis was associated with a 2.9-fold increased risk, and attempted suicide as a reason for consultation was associated with a 2.1-fold increased risk. Not being married doubled the risk, and the odds ratio was also high in cases of poor psychosocial functioning (as assessed by Global Assessment of Functioning (GAF) score). The predictive model differentiated reasonably well between those patients who were hospitalized and the other patients. In conclusion, this multi-centre study of factors predictive of referral to psychiatric hospital among general hospital patients revealed that the most important determinants were previous psychiatric care, diagnosis of psychosis or severe depression, attempted suicide, being unmarried, and poor psychosocial functioning as assessed by GAF score.