Anxiety and depression can profoundly and persistently impact the well-being and daily functioning of young adults, often leading to serious long-term consequences. Research on mental health trajectories in this age group has typically emphasized either large-scale population-level analyses or qualitative interpretations of individual experiences. To bridge these perspectives, we have conducted connected studies combining quantitative and qualitative approaches.

Using electronic health record (EHR) data comprising 644,827 healthcare encounters among 12,040 young adults, we examined healthcare trajectories over a period of three years preceding the onset of anxiety or depression and 1.5 years of follow-up. Applying a clustering-based machine learning approach, we identified two subgroups with distinct patterns of healthcare utilization and comorbidities. The first subgroup, predominantly female, frequently attended primary care for non-psychiatric conditions, particularly recurrent pain and reproductive health concerns such as contraceptive management and pregnancy-related care. The second subgroup showed a markedly different pattern, characterized by higher rates of psychiatric comorbidities, including personality and hyperkinetic disorders. These findings demonstrate the value of EHR-based trajectory mapping and clustering analytics in uncovering clinically meaningful and gender-specific subgroups within the young adult population. AI and EHR data analytics hold promise for identifying early signals of risk and guiding proactive, personalized interventions for anxiety and depression.

While large-scale analyses offer valuable insights for clinicians and healthcare organizations, understanding young adults’ subjective experiences of seeking help for mental health problems remains essential. Qualitative interviews provide a complementary lens, revealing lived experiences, perceptions, and decision-making processes underlying the quantitative patterns and highlighting aspects not captured by the analysis of EHR data alone. Integrating these approaches underscores the importance of person-centred, coordinated, and multidisciplinary models of care. To effectively support young adults presenting with unspecific psychiatric and non-psychiatric symptoms, healthcare systems must move beyond siloed structures toward more flexible and collaborative service delivery.

We hope that the presentation will provide valuable insights that can make a real difference in healthcare. By integrating AI models with longitudinal healthcare data, clinicians can identify at-risk young adults earlier and with greater precision, enabling more timely assessment and intervention. When complemented by qualitative insights from lived experiences of young adults, these predictive approaches gain crucial contextual depth, ensuring that early-identification strategies are not only clinically robust but also aligned with the actual needs and expectations of young adults.