A Data-Driven Early Warning System for Disease Outbreaks
Early detection of infectious disease outbreaks is essential for timely public health response, yet local case data are often sparse and noisy, making reliable monitoring difficult. In their paper, “Small-Area Estimation of Case Growths for Timely COVID-19 Outbreak Detection,” Zhaowei She, Zilong Wang, Turgay Ayer, and Jagpreet Chhatwal propose a new statistical learning framework, that is, transfer learning random forests (TLRF), that improves the estimation of epidemic growth rates across small geographic regions.
The approach combines ideas from small-area estimation and transfer learning with modern machine learning tools, specifically random forest models, to borrow information across counties and time periods. This data-driven strategy produces more stable estimates of infection growth even when local observations are limited.
Using COVID-19 data from across the United States, the authors show that their method detects emerging outbreaks more quickly and reliably than conventional approaches. The results demonstrate how advanced analytics can strengthen epidemic surveillance and support faster, better-informed public health decision making.
OBJECTIVES
Systematic literature reviews (SLRs) underpin life sciences research but are resource intensive. Generative artificial intelligence, particularly large language models (LLMs), may accelerate key SLR tasks, yet performance and reliability for evidence synthesis remain unclear. This manuscript aims to review current evidence on GenAI performance across core SLR tasks.
METHODS
We conducted a PRISMA-adapted rapid evidence assessment of English-language biomedical studies published from November 2022 to July 2025 evaluating GenAI or LLMs for systematic literature review tasks, including search strategy development, title/abstract screening, full-text screening, data extraction, risk-of-bias assessment, qualitative synthesis, report writing, and end-to-end review generation. Findings were summarized qualitatively by task.
RESULTS
Among 115 included studies, evidence supporting the use of GenAI was strongest for title/abstract screening (n=51) and data extraction (n=33). Selected high-quality evaluations reported sensitivities ≥90%, workload reductions of 27-71%, and human-comparable or superior performance in calibrated human-in-the-loop workflows. Evidence for full-text screening (n=15) and risk-of-bias assessment (n=17) was more variable, showing gains in structured or fine-tuned implementations but persistent limitations in specificity and nuanced judgment. For search strategy development, qualitative synthesis, and report writing, GenAI was most effective as a supportive tool; fully autonomous end-to-end SLR generation was unreliable.
CONCLUSIONS
GenAI can improve efficiency across multiple SLR tasks when used in hybrid human-AI workflows. Current evidence supports targeted, task-specific adoption with transparent reporting and human oversight, rather than full automation.
R. Fleurence, Riaz Qureshi, Rakesh Aggarwal et al.· Value in Health· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
