Trauma-Informed Care: An Investigation Into Knowledge, Attitudes, and Practices Among Genetic Counselors.

Mental health disorders affect an estimated 1 in 8 people globally, yet most who need care never receive it []. In the United States, almost half of individuals with mental illness do not receive care, with gaps driven by cost, geographic access, and stigma [,]. These shortfalls have accelerated interest in digital mental health interventions (DMHIs) as scalable, accessible alternatives to traditional care []. Among DMHIs, conversational AI agents, commonly referred to as AI therapists or mental health chatbots, have emerged as a particularly promising modality, offering around-the-clock availability and low cost []. Conversational AI agents can produce meaningful reductions in symptoms of depression and anxiety [,], lending empirical credibility to their growing adoption and positioning them as a practical option in settings where traditional care is hard to access. Despite the promise of conversational AI agents, a critical gap remains in understanding what drives therapeutic benefit within AI-powered care. Engagement in DMHIs is most commonly operationalized as the number of modules, activities, or sessions completed [], largely because these metrics are easy to extract and compare across users. However, these metrics assume that all engagement is equivalent, collapsing meaningful variation in how interactions unfold over time into a single quantity. Higher engagement does not consistently predict better outcomes in DMHIs, with the relationships between volume-based engagement and symptom improvement being weak and inconsistent [,]. Characterizing engagement in DMHIs has proven to be difficult, partially because engagement is multidimensional and encompasses behavioral, cognitive, and affective components that volume-based metrics cannot distinguish []. Prior work in DMHIs has largely focused on adherence, dropout, or total use [,]. As a result, existing approaches fail to capture how engagement unfolds over time, including patterns of depth and timing that may be critical for understanding outcomes. In contrast, human psychotherapy research shows that not only the amount of care but also how it is structured over time, including session intensity and spacing, can influence clinical improvement beyond the total therapy received [,]. For example, twice-weekly cognitive behavioral therapy (CBT) sessions have been shown to produce faster rates of symptom improvement in depression compared to once-weekly delivery, independent of the total sessions attended, suggesting that session concentration matters beyond cumulative dose []. These findings suggest that engagement is not interchangeable across time and that when and how interactions occur may shape therapeutic benefits independent of overall exposure. This distinction is particularly relevant for conversational AI interventions, where engagement is highly flexible and user-driven, leading to substantial variability in how individuals interact with care over time. However, despite this variability, the temporal and structural patterns of engagement have not been systematically examined in AI-powered DMHIs, leaving a significant gap in understanding which forms of engagement are associated with better outcomes and how AI-powered platforms should be designed and evaluated.