Article Impact Level: HIGH Data Quality: STRONG Summary of The Lancet Digital Health, 100886. https://doi.org/10.1016/j.landig.2025.100886 Dr. Amir Hadid et al.
Points
- Researchers utilized wearable biosensors and AI to predict systemic inflammation resulting from viral infections, aiming to detect illness before symptoms become apparent, enabling earlier intervention.
- The study monitored 55 healthy adults using smart rings, watches, and shirts for 12 days after administering a live attenuated influenza vaccine to collect physiological data.
- A machine learning model using night-time smart ring data predicted inflammatory surges with nearly 90% accuracy within a 24-hour window, showing high sensitivity to infection.
- This AI-driven approach significantly outperformed traditional symptom-based detection, which was less reliable due to cases of asymptomatic inflammation and reports of symptoms without underlying infection.
- This technology could enable earlier warnings for VRTIs, such as influenza or SARS-CoV-2, allowing for rapid intervention before critical health events occur in vulnerable populations.
Summary
The WE SENSE prospective controlled trial investigated presymptomatic prediction of viral upper respiratory tract infection (VRTI)-induced systemic inflammation using wearable biosensors and machine learning. The study recruited 55 healthy adults (49% female), who were monitored for 12 days surrounding inoculation with a live attenuated influenza vaccine. Data were collected using an Oura ring, Biobeat watch, and Astroskin-Hexoskin shirt, in conjunction with systemic inflammatory biomarker mapping, PCR testing, and self-reported symptoms. Machine learning algorithms using gradient-boosting techniques were trained on data from 35 participants to predict inflammatory surges without relying on symptoms.
A candidate model using only night-time data from the Oura ring predicted inflammatory surges with a receiver operating characteristic area under the curve (ROC-AUC) of 0.73 (95% CI 0.71–0.74) for real-time prediction and 0.89 (95% CI 0.87–0.90) within a 24-hour tolerance window. Incorporating both night-time and daytime data from the Astroskin-Hexoskin shirt maintained a real-time ROC-AUC of 0.73 (95% CI 0.71–0.75) but improved the 24-hour tolerance ROC-AUC to 0.91 (95% CI 0.90–0.92), with enhanced specificity (0.83, 95% CI 0.79–0.87) and F1 score (0.65, 95% CI 0.58–0.71).
All wearable-based models significantly outperformed the model based on symptoms alone, which yielded ROC-AUC values of 0.66 (95% CI, 0.63–0.68) for real-time predictions and 0.79 (95% CI, 0.77–0.82) for 24-hour tolerance predictions. The study concludes that machine learning analysis of physiological data from wearables provides a more objective and timely method for detecting systemic inflammation than symptom-based approaches, offering potential for earlier intervention in VRTIs, even in asymptomatic individuals.
Link to the article: https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00068-8/fulltext
References Hadid, A., McDonald, E. G., Ding, Q., Phillipp, C., Trottier, A., Dixon, P. C., Jlassi, O., Cheng, M. P., Papenburg, J., Libman, M., & Jensen, D. (2025). Development of machine learning prediction models for systemic inflammatory response following controlled exposure to a live attenuated influenza vaccine in healthy adults using multimodal wearable biosensors in Canada: A single-centre, prospective controlled trial. The Lancet Digital Health, 100886. https://doi.org/10.1016/j.landig.2025.100886
