Prediction of Dengue Incidence Using Search Query Surveillance
PLoS Neglected Tropical Diseases·
- DOI
- 10.1371/journal.pntd.0001258
- PMID
- 21829744
- PMCID
- PMC3149016
- OpenAlex
- W2079515288
- Study type
- Journal article
- Publisher
- Public Library of Science (PLoS)
- Article type
- journal-article
- Integrity
- current
Why this research matters now
Search-query surveillance offers a readily available and cost-effective tool that may support dengue monitoring in settings where traditional surveillance infrastructure is limited.
Structured evidence summary
Research question
The study evaluates whether internet search query data can accurately predict dengue incidence and identify periods of elevated case counts.
Study design
The researchers applied multiple predictive modeling approaches (linear regression, generalized boosted regression, negative binomial regression, logistic regression, and Support Vector Machine) to internet search data and dengue incidence records spanning 2004-2011. Model performance was assessed through cross-validation.
Population and setting
The analysis drew on weekly dengue incidence data from Singapore and monthly incidence data from Bangkok during 2004-2011.
Main findings
Step-down linear models achieved high predictive accuracy in both cities: r² values of 0.948 (Singapore) and 0.943 (Bangkok), with correlations between fitted and observed incidence of 0.931 and 0.869, respectively. Support Vector Machine models successfully identified high-incidence periods with area-under-curve values of 0.906 (Singapore) and 0.960 (Bangkok) at the 75th percentile threshold.
Public-health relevance
Search-query surveillance offers a readily available and cost-effective tool that may support dengue monitoring in settings where traditional surveillance infrastructure is limited.
Important limitations
This summary is derived solely from the supplied single-article abstract and metadata. Full assessment of model generalizability, internet penetration effects, lag periods, search-term selection rationale, and validation across additional settings requires review of the complete published paper.
GIDS interpretation
The article is indexed under dengue, influenza, Thailand, Singapore, surveillance, transmission dynamics, and treatment, which aids discovery in the context of infectious disease monitoring and digital epidemiology research. These classifier links do not constitute a live outbreak signal.
Related GIDS surveillance
Literature context does not validate, explain, or change a surveillance signal. Exact and contextual relationships are shown separately.
Evidence relationships
This article has 10 auditable classifier relationships to diseases, places, topics, and study design.