Fig. 3: molFOI varies by participant, not by age.

a Each point represents a single participant, showing molFOI from the 2011 season (x-axis) and cumulative molFOI across the 2012–2016 seasons (y-axis). The axis scales vary due to the difference in time periods and sampling strategy between the two datasets (see Fig. 1). The line represents a linear regression (adjusted R² = 0.38, P = 3.4e⁻¹³). We also computed the two-sided Spearman’s rank correlation between molFOI in 2011 and 2012–2016; we found a highly significant, positive correlation (⍴(111) = 0.53 (95% CI = [0.38, 0.66]), P = 1.51e⁻⁹). b Replicates the figure in (a), but with number of treatments for symptomatic disease in 2011 on the x-axis and number of treatments for symptomatic disease in 2012–2016 on the y-axis. The line represents a linear regression (adjusted R² = 0.03, P = 0.04), but a two-sided Spearman’s rank correlation did not identify a significant association between number of treatments in 2011 and 2012–2016 (⍴(111) = 0.17 (95% CI = [−0.02, 0.35]), P = 0.073). c Average molFOI of all participants of a given age (on the x-axis) during a given season (on the y-axis). d Average number of treatments of all participants of a given age (on the x-axis) during a given season (on the y-axis). e molFOI from the 2011 season, stratified by infection status at enrollment in May 2011. molFOI for participants who were infected at enrollment is significantly higher than for those not infected at enrollment (Kruskal–Wallis test, χ²(1) = 78.42, P = 8.33e⁻¹⁹). f Replicates the analysis in (c), but with number of treatments for symptomatic disease in 2011 on the y-axis. Participants who were infected at enrollment had significantly more treatments than those who were not infected at enrollment (Kruskal–Wallis test, χ²(1) = 7.85, P = 0.0051). Violin plots in (e, f) represent the density of the distributions, with horizontal lines at the 25th, 50th, and 75th percentiles. CI confidence interval, molFOI molecular force of infection.