Mental health disorders affect a substantial portion of the global population. Despite preferences for psychotherapy, access remains limited due to various barriers. Digital mental health interventions (DMHIs) have emerged to increase accessibility, yet engagement and treatment completion rates are concerning. Evidence across healthcare where some degree of self-management is required show that treatment engagement is negatively influenced by contextual complexity. This article examines the non-random factors influencing patient engagement in digital and face-to-face psychological therapies. It reviews established models and introduces an adapted version of the Cumulative Complexity Model (CuCoM) as a framework for understanding engagement in the context of digital mental health. Theoretical models like the Fogg Behavior Model, Persuasive System Design, Self-Determination Theory, and Supportive Accountability aim to explain disengagement. However, none adequately consider these broader contextual factors and their complex interactions with personal characteristics, intervention requirements and technology features. We expand on these models by proposing an application of CuCoM's application in mental health and digital contexts (known as DiCuCoM), focusing on the interplay between patient burden, personal capacity, and treatment demands. Standardized DMHIs often fail to consider individual variations in burden and capacity, leading to engagement variation. DiCuCoM highlights the need for balancing patient workload with capacity to improve engagement. Factors such as life demands, burden of treatment, and personal capacity are examined for their influence on treatment adherence. The article proposes a person-centered approach to treatment, informed by models like CuCoM and Minimally Disruptive Medicine, emphasizing the need for mental healthcare systems to acknowledge and address the unique burdens and capacities of individuals. Strategies for enhancing engagement include assessing personal capacity, reducing treatment burden, and utilizing technology to predict and respond to disengagement. New interventions informed by such models could lead to better engagement and ultimately better outcomes.