Implementing a digital solution to support cancer patients with mental health co-morbidities: A need for ai-driven first-line support strategies

Cancer remains a leading cause of morbidity and mortality worldwide, imposing significant physical, emotional, and economic burdens on patients and healthcare systems. Beyond the direct physiological challenges of treatment, individuals with cancer often experience profound psychosocial distress, including symptoms of depression and anxiety, which can hinder treatment adherence and coping capacity. Despite evidence that psychosocial interventions can enhance overall well-being and sometimes even survival rates, integration of mental health into oncology care faces obstacles such as workforce shortages, stigma, and geographical barriers. To strengthen the scientific contribution of this work, we now (i) clarify that the paper presents a mixed methods development evaluation protocol rather than a purely conceptual proposal, and (ii) summarize preliminary feasibility data collected in two European cancer centers (n = 96; see Section 5). Based upon the feasibility study, this paper proposes-empirically based pilots- to measure a digital mental health triage solution underpinned by artificial intelligence (AI). By embedding routine psychosocial screening into oncology workflows, patients are stratified into low, moderate or high-risk tiers. Low and moderate tiers receive AI guided self-management resources; high-risk patients are referred rapidly to specialist care. Psychological distress was assessed at baseline and after four weeks using the 51-item Empowerment for Participation mini assessment (EFP Mini), highly correlated with both GAD-7 and PHQ-9. Group 1 (Cohort & Self-assessment only) achieved a modest but statistically significant reduction in total distress scores (Δ = –28.8 ± 54.6; 95% CI [–54.36, –3.24]; t(19) = –2.36; p = 0.029), Group 2 (self-assessment with Ai support) showed a non-significant trend toward improvement (Δ = –22.8 ± 39.3; 95% CI [–50.90, 5.30]; t(9) = –1.84; p = 0.10), and Group 3 (self-assessment with the blended AI + human coaching arm yielded a large, clinically meaningful decrease (Δ = –121.7 ± 55.3; 95% CI [–161.26, –82.14]; t(9) = –6.96; p < 0.001), with a one-way ANOVA confirming significant differences across groups (F(2, 37) = 12.79; p < 0.001). Potential benefits, implementation challenges and regulatory implications-including the forthcoming EU AI Act-are critically examined. This paper explores the methodological considerations, potential benefits, and implementation challenges associated with an AI-driven mental health triage system in oncology. Future directions include long-term evaluations of effectiveness, cost-benefit analyses, and culturally sensitive adaptations across diverse healthcare settings.