Abstract
Prior work has shown that school-age children's influenza activity precedes that of adults by one to two weeks. Whether this is robust across outcomes or constant across a season's course is unexamined.
Using electronic health record data from two health systems in South Carolina (2016-2026) on individuals with an influenza diagnosis or positive test, we tested age-group lead-lag relationships across six measures using differenced cross-correlation and Granger causality, with significance confirmed via heteroskedasticity-robust block permutation. We tested whether relationships differed between a season's growth and decline phases.
Children 5-17 years led every other age group by approximately one week in four out of six outcomes and were the only group showing any nonzero-lag relationship across 60 pairwise combinations. A screen of 150 source-target combinations identified 26 permutation-confirmed instances of children's lower-acuity activity preceding elevated emergency department or inpatient burden in other age groups. The same-outcome lead was concentrated during a season's growth phase, while the cross-severity lead was robust throughout both a season's growth and decline.
School-age children's influenza activity is a robust, multi-outcome leading indicator with distinct transmission-timing and severity-progression signals in different phases of seasonal epidemics, supporting incorporation of age-stratified, phase-aware indicators into influenza forecasting.