This paper reviews the literature on disengagement from mental health services examining how the terms engagement and disengagement are defined, what proportion of patients disengage from services, and what sociodemographic variables predict disengagement. Both engagement and disengagement appear to be poorly conceptualised, with a lack of consensus on accepted and agreed definitions. Rates of disengagement from mental health services vary from 4 to 46%, depending on the study setting, service type and definition of engagement used. Sociodemographic and clinical predictors of disengagement also vary, with only a few consistent findings, suggesting that such associations are complex and multifaceted. Most commonly reported associations of disengagement appear to be with sociodemographic variables including young age, ethnicity and deprivation; clinical variables such as lack of insight, substance misuse and forensic history; and service level variables such as availability of assertive outreach provision. Given the importance of continuity of care in serious mental disorders, there is a need for a consensual, validated and reliable measure of engagement which can be used to explore associations between patient, illness and service related variables and can inform service provision for difficult to reach patients.