Compulsory admission (CA) refers to the process of hospitalization of an individual to a psychiatric ward or hospital, without their consent, due to severe mental health conditions. While it is an established and legally framed practice in many countries, it raises a number of ethical issues in terms of personal liberty and risks of potential misuse. Ethnicity, male gender and psychosis are the main risk factors for compulsory admission to psychiatric wards/hospitals in general population. Previous studies documented that CA is even more frequently used in prison, yet its determinants are still unknown. To address this issue, we explored the clinical, demographic and criminological determinants of compulsory admission in 317 detainees admitted to a psychiatric acute care secure ward located in the central prison of Geneva. We distinguished three groups: voluntary admissions only (VA), CA only and mixed admissions (MA). Judicial status and types of offenses were also recorded. Sociodemographic data included age, gender, and origin. Clinical parameters included previous compulsory admissions, previous hospital stays, number of admissions, length of stay, number of suicide attempts, short-term seclusions and ICD-10 clinical diagnoses. Fisher's exact, Chi² and Kruskal Wallis tests were used for group comparisons. Logistic regression analysis was used to explore the association between the legal status of hospital stays and clinical diagnosis. There was no significant difference between the three groups in respect to sociodemographic factors. Past compulsory admissions were significantly more frequent in the CA compared to the VA group. Both the mean number of admissions and length of stay differed significantly between the three groups. The number of suicidal attempts was significantly higher in MA compared to both VA and CA groups. Short-term seclusion was significantly more frequent in CA and MA compared to VA. Psychotic disorders were much more frequent in CA (55.1%) and MA (54.8%) compared to VA cases (23.9%). In contrast, depressive and anxiety disorders were significantly less frequent in CA (12.3%) and MA (14.3%) than VA (29.5%). Of importance, neither the type of offenses nor the judicial status differed between the three groups. In regression models, CA was strongly and positively associated with psychotic disorders. The inverse was true for depressive, anxiety as well as adjustment disorders. The present findings reveal that, in contrast to the general population, sociodemographic factors have no impact on the frequency of CA in a population of detainees. The main risk factors for the adoption of this measure are past CA and presence of acute psychosis. In contrast, the presence of depressive and anxiety symptoms limits the recourse to this disputed measure.