Among the different types of heavy use of mental health services, frequent inpatient admission in acute care units of individuals unable to return to their usual environment refers to as revolving-door (RD). RD in prisoners is related to increased violence (acted and supported) and suicidal recidivism. We explored the determinants of RD in 200 inmates from the Swiss-French speaking areas who were admitted to the sole acute psychiatric care unit for all of the Swiss-French counties, located in Geneva. The Cuzick's test for trend across ordered groups, Kruskal-Wallis test and oneway ANOVA were used to compare demographic and clinical variables between single (one admission, N = 100), frequent (3–7, N = 69) and RD (more than 8, N = 31) during a 12 months period. In addition, univariate and multivariable ordered logistic regression modes were built to examine the determinants of RD. The sample included 27 women (mean age: 31.2 years) and 173 men (34.5 years) who were admitted during the period 2014–2019. The vast majority were single (65%) with low level of education (<6 years, 78%). Suicidal behavior was the more frequent reason for admission (57%). Psychiatric history was positive in 77.5% of cases and in 54.5% of cases there was at least one episode of inpatient psychiatric care. The more frequent ICD-10 psychiatric diagnosis in the last admission were psychotic disorder (38%), personality disorder (29.5%) and adjustment disorder (19.5%). In contrast, depressive episodes (7%) and bipolar disorder (4.5%) were rare. Group comparison showed that the presence of court-ordered treatments, suicidal behavior, personality and psychotic disorders was associated with significantly increased frequency of RD use. In univariate models, the same factors were positively associated with RD, the highest odds ratio being found for court-ordered treatments (5.77) and personality disorders (2.14). In contrast, the diagnosis of adjustment disorders was related to decreased RD use (OR 0.25). Court-ordered treatments and personality disorders were the only factors to predict RD in multivariable regression models. These findings suggest that acute psychiatric care in these patients did not depend of environmental stressors but rather represents the expression of a long-lasting vulnerability related to their psychological profile and criminal status.