The early prediction of patients at risk may facilitate the efficient use of interventions that have been demonstrated to reduce readmissions. The aim of the study was to analyze variables during first admissions associated with further high use of an inpatient hospitalization psychiatric unit in Málaga, Spain. The risk of having three or more psychiatric admissions was analyzed in a sample of 1535 first-time admissions with multivariate Cox regression. In the multivariate model, the variables associated with the risk of high use were age at admission (p < 0.001), length of stay (p < 0.001), place of residence (p < 0.001), and previous history with mental health services (p < 0.001). The results suggest that there are several easily accessible characteristics at first admission that are potentially useful in detecting patients at risk.