Monitoring airline pilot mental health: a 3PM framework utilising digital phenotyping and AI

The aeromedical certification of commercial airline pilots relies on periodic, self-disclosure-dependent assessments. However, this reactive model lacks the temporal resolution required to capture the gradual behavioural and physiological perturbations preceding an overt mental health crisis. Concurrently, anonymous epidemiological surveys consistently document subclinical depressive symptoms, insomnia, and burnout among active flight crews. Within the framework of predictive, preventive, and personalised medicine (3PM), these conditions represent suboptimal health states situated in the critical transition zone between health and disease. In this narrative review, we propose a conceptual framework for integrating digital phenotyping and artificial intelligence (AI) into aeromedical mental health monitoring, grounded in the 3PM paradigm. With this aim, we appraise three complementary data streams - including smartphone-derived behavioural markers, wearable-derived physiological signals, and cockpit-derived vocal biomarkers - and examine how aviation-specific confounders modulate their validity. Furthermore, we propose a hybrid federated-plus-on-device AI architecture that reconciles population-level statistical power with individual data sovereignty. Crucially, this framework advances the three constitutive dimensions of 3PM in the aviation context. Regarding prediction, it delivers longitudinally resolved diagnostics capable of detecting health-to-disease trajectories during the prodromal phase, surpassing the temporal limitations of current certification cycles. For prevention, it establishes risk-stratified intervention pathways triggered by deviations from individual baselines, thereby shifting oversight from reactive grounding to pre-emptive peer support. Finally, it affords the personalisation of medical services via on-device AI; by calibrating monitoring to each pilot's specific behavioural baseline, duty patterns, and circadian exposure, it effectively replaces population-normed evaluations with individually tailored longitudinal surveillance.