When is travel-health AI inference transportable? Mobility-conditioned admissibility for cross-border disease risk and decision support.

BACKGROUND

Artificial intelligence is increasingly used for infectious disease surveillance, traveller risk assessment, outbreak forecasting and decision support. Travel creates a distinctive transportability problem because people move between epidemiological, geographic, climatic and healthcare contexts in which the conditions supporting an inference may change rapidly.

METHODS

This conceptual/theoretical methodological-development study was informed by a targeted narrative review of travel medicine, infectious disease surveillance, human mobility, border health, mass gatherings, antimicrobial resistance and health-AI validation literature.

RESULTS

We develop a mobility-conditioned admissibility framework that treats travel as movement across origin, transit, destination and return regimes. Interpretation is conditioned on mobility state, epidemiological regime, exposure history, time-dependent constraints and clinical or biological constraints. Five proposed operational states are retained: supported inference, conditionally transportable inference, active monitoring, restricted interpretation and withholding. The framework also distinguishes low probability from infeasibility and requires preservation of rare but epidemiologically feasible events.

CONCLUSIONS

Model availability and conventional predictive validity do not establish inferential transportability across changing travel contexts. Mobility-conditioned admissibility provides a complementary use-boundary layer for travel health AI by making cross-regime support explicit before outputs are interpreted or released. Its incremental value requires prospective evaluation against calibrated uncertainty, external validation and dataset-shift monitoring.