Rates of compulsory detention in psychiatric hospitals have risen over several decades in England and some other higher income countries. This study explores clinicians’ perspectives on how service users come to be compulsorily detained in psychiatric hospitals and their suggestions for reducing these detentions in the future. Semi-structured qualitative interviews were conducted with 23 clinicians working with individuals who have been compulsorily detained under the Mental Health Act in England. Interviews were carried out by telephone or videoconference. Data was analysed using template analysis, which involved developing a structured framework to organise and analyse data and to develop themes. Three major themes were identified, with multiple sub-themes (a) service user factors that increase the risk of compulsory detention, including high levels of risk, previous/underlying trauma, medication non-adherence, service user perceptions of their mental health state, disadvantage and discrimination, and lack of stability and involvement from family and social networks; (b) service-level reasons for being detained, including lack of communication and continuity of care, historical inability to obtain trust and confidence from parts of population, clinician biases and assumptions, lack of resources, lack of treatment and care variety, and systemic/institutional barriers to engagement; and (c) potential pathways to reducing compulsory detention, including increasing care quality and patient/family level interventions, investing in services, offering choice regarding medication, offering alternatives to detention, and improving discharge planning. Our study advances the literature by highlighting systemic, patient-level, and service-level factors perceived as driving practice. Better-resourced community services and care planning and strategies to address unconscious bias are identified as potential routes to reducing detentions. However, significant limitations are a preponderance of London-based psychiatrists in our sample, which may affect the generalisability of the findings to other roles and locations, and a lack of corroboration of perceived causality with more objective data.