JMIR Publications

ABSTRACT

Background:

Early detection of health threats is an objective of public health surveillance and event-based surveillance (EBS) using unstructured ad hoc information from diverse sources has served an increasingly important role in achieving this objective. However, evaluation of EBS systems has been hindered by the lack of reference data on outbreak onsets.

Objective:

In this study, we introduce the concept of an “early detection interval” and create a dataset of these intervals in multiple countries for the epidemic caused by the Omicron variant.

Methods:

We define the early detection interval as the time between the date of introduction of an infectious agent to a country and the date at which an increase is detectable in traditional public health surveillance data. To determine the date of introduction of the Omicron variant, we analyzed phylogenetic studies and genome databases. We estimated the end of the interval by applying Bayesian online change point detection to reported COVID-19 case counts. In addition to the early detection intervals, this dataset also contains variables indicating data quality, such as discordance between interval start dates using phylogenetic estimation and data from genomic sample collection & submission.

Results:

This dataset includes early detection intervals for 117 countries with a median length of 28 (IQR: 18 – 44) days, highlighting the potential value of early outbreak detection.

Conclusions:

This dataset can serve as a reproducible reference framework to evaluate the timeliness of alerts generated by EBS.