James Hay
@jameshay.bsky.social
Research Fellow at the Pandemic Sciences Institute, University of Oxford. Using maths and stats to understand infectious disease dynamics, mostly viral kinetics and serology. https://hay-idd.github.io/
We developed an age-stratified SIR model, visually calibrated to 2022/23 data (an H3N2 season). We ran scenarios varying intrinsic transmissibility, loss of immunity overall or in children, and the season start date. We accounted for changes in contacts over school holidays and Christmas.
(5/18)
(5/18)
November 25, 2025 at 1:51 PM
We developed an age-stratified SIR model, visually calibrated to 2022/23 data (an H3N2 season). We ran scenarios varying intrinsic transmissibility, loss of immunity overall or in children, and the season start date. We accounted for changes in contacts over school holidays and Christmas.
(5/18)
(5/18)
Rapid spread in children – the usual culprits driving seasonal flu transmission. Compared to the past 2 seasons, there has been a larger difference in growth rate between <14 yos and adults. Half term caused a drop in growth and may have acted as a circuit break for the early season.
(4/18)
(4/18)
November 25, 2025 at 1:51 PM
Rapid spread in children – the usual culprits driving seasonal flu transmission. Compared to the past 2 seasons, there has been a larger difference in growth rate between <14 yos and adults. Half term caused a drop in growth and may have acted as a circuit break for the early season.
(4/18)
(4/18)
Case numbers and epi dynamics don’t look too unusual when you align the curves by date of peak growth rate (* to date*). The flu season has started very early but is otherwise not completely unprecedented (at least in terms of symptomatic cases).
(3/18)
(3/18)
November 25, 2025 at 1:51 PM
Case numbers and epi dynamics don’t look too unusual when you align the curves by date of peak growth rate (* to date*). The flu season has started very early but is otherwise not completely unprecedented (at least in terms of symptomatic cases).
(3/18)
(3/18)
We analysed public data sources (Resp DataMart, RGCP, WHO FluNet) from England to compare growth rates and reproduction numbers to previous seasons. So far, peak growth rates have been higher than the past 10 seasons, but Rt estimates are in line with the upper end of previous seasons.
(2/18)
(2/18)
November 25, 2025 at 1:51 PM
We analysed public data sources (Resp DataMart, RGCP, WHO FluNet) from England to compare growth rates and reproduction numbers to previous seasons. So far, peak growth rates have been higher than the past 10 seasons, but Rt estimates are in line with the upper end of previous seasons.
(2/18)
(2/18)
An exciting output of our inference is the well-known relationship between antibody titer and probability of infection. Using our method, we can understand not just serological patterns, but also immunity patterns using these multi-antigen serology panels.
April 6, 2024 at 8:32 AM
An exciting output of our inference is the well-known relationship between antibody titer and probability of infection. Using our method, we can understand not just serological patterns, but also immunity patterns using these multi-antigen serology panels.
We find:
1. Serology-based attack rates are high, at around 18% infected per year.
2. Influenza A/H3N2 infection rates are highest in children, decrease with age and plateau in adulthood.
3. Incidence rates are highly correlated at this small spatial scale.
1. Serology-based attack rates are high, at around 18% infected per year.
2. Influenza A/H3N2 infection rates are highest in children, decrease with age and plateau in adulthood.
3. Incidence rates are highly correlated at this small spatial scale.
April 6, 2024 at 8:32 AM
We find:
1. Serology-based attack rates are high, at around 18% infected per year.
2. Influenza A/H3N2 infection rates are highest in children, decrease with age and plateau in adulthood.
3. Incidence rates are highly correlated at this small spatial scale.
1. Serology-based attack rates are high, at around 18% infected per year.
2. Influenza A/H3N2 infection rates are highest in children, decrease with age and plateau in adulthood.
3. Incidence rates are highly correlated at this small spatial scale.
Fitting serosolver gave us estimates for: 1) each individual’s sequence of lifetime influenza infections; 2) incidence at a fine spatial scale; and 3) parameters of an antibody kinetics model describing boosting, waning, cross-reactivity and measurement error.
April 6, 2024 at 8:32 AM
Fitting serosolver gave us estimates for: 1) each individual’s sequence of lifetime influenza infections; 2) incidence at a fine spatial scale; and 3) parameters of an antibody kinetics model describing boosting, waning, cross-reactivity and measurement error.
This new paper brings these pieces together: we fit serosolver to our massive dataset of over 70,000 HI titers, summarizing antibody profiles against 20 A/H3N2 strains for 1,130 individuals from Guangzhou, China.
April 6, 2024 at 8:31 AM
This new paper brings these pieces together: we fit serosolver to our massive dataset of over 70,000 HI titers, summarizing antibody profiles against 20 A/H3N2 strains for 1,130 individuals from Guangzhou, China.