Research of Interest in Travel Health
Published on 01/08/2026
Abstract
Machine learning (ML) shows promise as a modern diagnostic ally in post travel medicine, offering strong performance for malaria and meaningful support for complex gastrointestinal syndromes. By integrating clinical and travel data, ML can sharpen diagnostic focus for non specialists, though safe adoption demands rigorous validation, transparency, and ethical safeguards. This summary highlights the key findings of the article by Furuya Kanamori et al., demonstrating how ML may modernise diagnostic pathways for returned travellers.
Introduction
Health care professionals are faced with a challenge when travellers return home sick because of the non-specific nature of most travel-related illnesses: there is a long list of differential diagnoses.
Machine learning (ML) has been used in research, modelling and risk prediction tools, however this paper by Australian travel health specialists1 explores an untapped opportunity and shows how ML could possibly be useful in post-travel consultations.
Be that by aiding (not replacing) clinicians in diagnostic decision-making and narrowing the list
of differential diagnoses or stream-lining traditional travel histories to improve clinical outcomes. There is also the potential to implement this in general practice or in an emergency setting, where clinicians may lack experience in tropical illnesses.
The paper compiled evidence on the use of ML in post-travel clinical care as well as identifying key gaps and opportunities for future research and clinical implementation. It is a perspective, 1870-word article based on a systematic review and diagnostic meta-analysis of five studies. Specifically, the paper encompasses both conventional ML approaches and emerging artificial intelligence (AI) methods.
Methods
To identify relevant articles, a systematic search was performed across five databases (PubMed, Embase, Scopus, Web of Science, and the Cochrane Library) and the analysis was based on data from 3,065 cases from European referral centres, UK hospitals, an Israeli medical centre, US travel clinics, and Southeast Asian clinics – providing a diverse data set.
This paper focused on returning travellers presenting with either febrile illnesses (where ML models were evaluated on their ability to predict a diagnosis of malaria from initial, non-specific symptoms) or gastrointestinal conditions. Just three malaria studies and two gastrointestinal studies met the criteria for inclusion into the meta analysis. The benchmarks used in this study were established diagnostic standards and clinical outcomes.
Results
ML and AI models have had a good to excellent performance in diagnosing in post-travel health setting, effectively identifying both malaria and gastrointestinal conditions among returned 2 travellers.
Regarding the malaria findings, a low sensitivity value suggests that there are false negatives with models that rely solely on blood-sampling, but performance was significantly enhanced when detailed travel itinerary data was included alongside clinical findings. Gastrointestinal findings showed diagnostic capabilities but less reliability overall than for malaria.
Interpretation
The paper suggests that there is great potential for ML in post-travel consultations and that is clinically meaningful. The analysis indicates that ML can flag malaria reliably, but it also shows how important it is to include a travel history as sensitivity increases substantially when this is done, reducing the risk of false negatives; turning ML into an actionable diagnostic support, if used correctly. The comparatively poorer outcome for gastrointestinal cases could be due to heterogeneity and the syndromic nature of the conditions, however, even a modest performance still suggests that ML could aid reducing differential diagnoses or support in triage.
ML could become a tool used by many health professionals in a travel health setting. This idea is supported by other recent research using ML models to diagnose influenza in international travellers2, and another study which outlined a framework for using AI-driven chatbots for personalized travel health assistance.3
Implications for UK Clinical Practice
Across NHS and private sectors, ML could provide non-specialists with vital decision support. By narrowing differential diagnoses and prioritising investigations, these tools could reduce diagnostic uncertainty where specialist expertise is unavailable. The UK has high outbound travel but is not a tropical country, so healthcare professionals may have little experience in identifying tropical diseases.
Where returning travellers’ diagnoses are in doubt, ML would be particularly useful. However, it is important to note that ML would only augment and not replace a clinician’s judgement and findings support ML as a diagnostic aid and not a diagnostic solution. The authors point out a number of limitations to robustly validating and embedding ML in clinical practice, including the relatively limited evidencebased research currently available. Overall, ML represents a promising direction for the modernisation of post-travel diagnostics, providing decision-support for clinicians.
References
1. Furuya-Kanamori L, Henderson A, Vos G, Farnham A, Mills DJ, Lau CL. Beyond pre-travel risk assessment: machine learning for clinical diagnosis in returned travellers. J Travel Med 2026;33(3):taag016. https://doi.org/10.1093/jtm/taag016
2. Asai Y, Yamamoto K, Nomoto H, Nakagawa H, Sahara T, Yamato M, et al. Development of machine-learning models to diagnose influenza among international travellers based on symptoms and epidemiological information. Travel Med Infect Dis. 2026 Mar;102966. https://doi.org/10.1016/j.tmaid.2026.102966
3. Baglivo F, De Angelis L, Cruschelli G, Rizzo C. A decalogue for personalized travel health assistance with
AI-driven chatbots. J Travel Med. 2024 Jun;31(4):taae026. https://doi.org/10.1093/jtm/taae026