Abstracts

Advances in high-throughput technologies, digital phenotyping, and the increased accessibility of publicly available datasets offer immense opportunities for leveraging big data across various fields.In the realm of infectious diseases, these applications can enhance the processes of surveillance, clinical diagnosis, and prognosis.Still, analyzing big data presents challenges due to the diversity of data sources, the potential for missing or poor-quality data, and the clinical heterogeneity among different patient demographics and host variations.Consequently, data-driven approaches utilizing machine learning models and techniques are essential to improving surveillance, diagnosis, and prognosis in infectious diseases.Using a systematic literature review, the current best models for each of the processes are exemplified.Current efforts in developing machine learning models primarily focus on improving quantitative performance metrics, with little attention given to translational or clinical relevance.Furthermore, there needs to be a shift in focus from replacing clinicians to complementing their daily workflow.Insights and recommendations for clinically relevant machine learning models are discussed.