Decomposing Participatory Surveillance Symptom Time Series to Track Respiratory Infections: A Cross-Country Evaluation Using Non-Negative Matrix Factorization

Read full text Summarise with VIRAT
Authors: Gesa Carstens, Mattia Mazzoli, Nicolò Gozzi, Albert Jan Van Hoek, Daniela Paolotti
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Background The annual respiratory season in Europe is marked by the co-circulation of multiple respiratory pathogens, such as influenza viruses, rhinoviruses, and coronaviruses. Effective surveillance is necessary but hampered by heterogeneity of case definitions and limited pathogen specificity in existing systems. This study aims to detect pathogen-specific signals in the participatory surveillance of the Netherlands using a sub-set of samples with virological detection. Additionally, we explore a method to use the findings in the Netherlands to enhance the virological interpretation of participatory surveillance data in Italy. Methods We analyzed symptom data collected through a participatory surveillance platform in the Netherlands and Italy over five years (2020–2025). Symptom-by-week matrices from the Dutch cohort were aggregated into syndromes and their associated time series using Non-negative Matrix Factorization (NMF). We compared the respective time series of the syndromes with influenza virus, SARS-CoV-2, seasonal coronaviruses, RSV, and rhinovirus incidence estimated from nose- and throat swabs of a subsample of symptomatic participants of the participatory surveillance platform in the Netherlands. We tested the transferability of these components by applying the Dutch-derived components to describe Italian symptom data and extract respective incidences. Results NMF identified eight symptom clusters in the Dutch cohort, one aligning with SARS-CoV-2, one aligning with rhinovirus and a third component aligning with influenza virus, RSV and seasonal incidences estimated from collected nose- and throat swabs. Transferring Dutch-derived symptom clusters to Italian data showed consistency in key components between Dutch and Italian cohorts, particularly those associated with SARS-CoV-2. Conclusion This study demonstrates that unsupervised symptom decomposition can identify co-circulating respiratory pathogens trends from syndromic surveillance data, providing timely pathogen circulation insights. Graphical abstract

Fields of Study & Subjects

Data-Driven Disease Surveillance Respiratory viral infections research Influenza Virus Research Studies

Publication details

Journal / source
medRxiv
Publication date
31 March 2026
Language
EN
Document type
Preprint