Tim K Tsang

Measuring population immunity is crucial for epidemic preparedness, but methods to translate individual immunity into population-level immunity metrics remain underdeveloped. We aimed to develop and evaluate population immunity estimators using influenza as a model pathogen. In this multicountry, retrospective observational study, we analysed 41 835 serum samples from six studies in China, Hong Kong, Viet Nam, and the USA across 27 influenza epidemics in 2009-24. We constructed four population-immunity estimators from individual antibody titres: geometric mean titre, proportion of non-naive individuals, proportion of population immune, and relative reduction in reproductive number calculated using a next-generation matrix framework. We evaluated the ability of these estimators to predict which subtype (H1N1 vs H3N2) would dominate, to predict whether epidemics would be larger or smaller than the previous season, and to correlate with subsequent cumulative incidence in a longitudinal Hong Kong cohort spanning eight influenza seasons with within-epidemic serum collection. Subtype-specific relative changes in these estimators from previous seasons correctly predicted 57-86% of H1N1-dominant seasons and 100% of H3N2-dominant seasons, with an area under the receiver operating characteristic curve (AUROC) of 79-93%. The estimators correctly predicted larger epidemics in 67-83% of cases and smaller epidemics in 90-100% of cases, with an AUROC of 83-92%. Performance varied by estimator and influenza subtype, with wide uncertainty intervals for some estimates indicating modest precision. In the longitudinal Hong Kong cohort, all four estimators negatively correlated with subsequent 30-day cumulative incidence for H1N1 (Pearson correlations -0.23 to -0.46), whereas H3N2 correlations were mostly non-significant. We developed and validated four complementary population immunity estimators derived from individual antibody titres that showed predictive utility for influenza subtype dominance and epidemic size direction across diverse settings. These estimators could inform seasonal influenza preparedness, surveillance prioritisation, and health-care resource planning.

The 2015-2016 Zika virus outbreak in the Americas presented significant challenges in understanding the transmission dynamics due to substantial reporting biases, as women of reproductive age (15-39 years) were disproportionately represented in the surveillance data when public awareness of relationship between Zika and microcephaly increased. Using national surveillance data from Colombia during July 27, 2015-November 21, 2016, we developed a Bayesian hierarchical modeling framework to reconstruct the true numbers of symptomatic cases and estimate transmission parameters while accounting for differential reporting across age-sex groups. Our model revealed that the detection rate of symptomatic cases among women of reproductive age was 99% (95% CI: 98.7-100), compared to 85.4% (95% CI: 84.7-86.1) in other demographic groups. After correcting for these biases, our results showed that females aged 15-39 years remained 82.8% (95% CI: 80.2-85.2%) more susceptible to Zika symptomatic infection than males of the same age, independent of differential reporting areas. Departments with medium-high altitude, medium-high population density, low coverage of forest, or high dengue incidence from 2011-2015 exhibited greater Zika risk. This study underscores the importance of accounting for surveillance biases in epidemiological studies to better understand factors influencing Zika transmission and to inform disease control and prevention.

Forecasting influenza activity in tropical and subtropical regions, such as Hong Kong, is challenging due to irregular seasonality and high variability. We develop a diverse set of statistical, machine learning, and deep learning approaches to forecast influenza activity in Hong Kong 0 to 8 weeks ahead, leveraging a unique multi-year surveillance record spanning 32 epidemics from 1998 to 2019. We consider a simple average ensemble (SAE) of the top two individual models, and develop an adaptive weight blending ensemble (AWBE) that dynamically updates model contribution. All models outperform the baseline constant incidence model, reducing the root mean square error (RMSE) by 23%-29% and weighted interval score (WIS) by 25%-31% for 8-week ahead forecasts. The SAE model performed similarly to individual models, while the AWBE model reduces RMSE by 52% and WIS by 53%, outperforming individual models for forecasts in different epidemic trends (growth, plateau, decline) and during both winter and summer seasons. Using the post-COVID data (2023-2024) as another test period, the AWBE model still reduces RMSE by 39% and WIS by 45%. Our framework contributes to comparing and benchmarking models in ensemble forecasts, enhancing evidence for synthesizing multiple models in disease forecasting for geographies with irregular influenza seasonality.

There has long been controversy over the potential for asymptomatic cases of the influenza virus to have the capacity for onward transmission, but recognition of asymptomatic transmission of COVID-19 stimulates further research into this topic. Here, we develop a Bayesian methodology to analyze detailed data from a large cohort of 727 households and 2515 individuals in the 2009 pandemic influenza A(H1N1) outbreak in Hong Kong to characterize household transmission dynamics and to estimate the relative infectiousness of asymptomatic versus symptomatic influenza cases. The posterior probability that asymptomatic cases [36% of cases; 95% credible interval (CrI): 32%, 40%] are less infectious than symptomatic cases is 0.82, with estimated relative infectiousness 0.57 (95% CrI: 0.11, 1.54). More data are required to strengthen our understanding of the contribution of asymptomatic cases to the spread of influenza.

For >70 years, a 4-fold or greater rise in antibody titer has been used to confirm influenza virus infections in paired sera, despite recognition that this heuristic can lack sensitivity. Here we analyze with a novel Bayesian model a large cohort of 2353 individuals followed for up to 5 years in Hong Kong to characterize influenza antibody dynamics and develop an algorithm to improve the identification of influenza virus infections. After infection, we estimate that hemagglutination-inhibiting (HAI) titers were boosted by 16-fold on average and subsequently decrease by 14% per year. In six epidemics, the infection risks for adults were 3%-19% while the infection risks for children were 1.6-4.4 times higher than that of younger adults. Every two-fold increase in pre-epidemic HAI titer was associated with 19%-58% protection against infection. Our inferential framework clarifies the contributions of age and pre-epidemic HAI titers to characterize individual infection risk.

BACKGROUND: When a new infectious disease emerges, appropriate case definitions are important for clinical diagnosis and for public health surveillance. Tracking case numbers over time is important to establish the speed of spread and the effectiveness of interventions. We aimed to assess whether changes in case definitions affected inferences on the transmission dynamics of coronavirus disease 2019 (COVID-19) in China. METHODS: We examined changes in the case definition for COVID-19 in mainland China during the first epidemic wave. We used exponential growth models to estimate how changes in the case definitions affected the number of cases reported each day. We then inferred how the epidemic curve would have appeared if the same case definition had been used throughout the epidemic. FINDINGS: From Jan 15 to March 3, 2020, seven versions of the case definition for COVID-19 were issued by the National Health Commission in China. We estimated that when the case definitions were changed, the proportion of infections being detected as cases increased by 7.1 times (95% credible interval [CrI] 4.8-10.9) from version 1 to 2, 2.8 times (1.9-4.2) from version 2 to 4, and 4.2 times (2.6-7.3) from version 4 to 5. If the fifth version of the case definition had been applied throughout the outbreak with sufficient testing capacity, we estimated that by Feb 20, 2020, there would have been 232 000 (95% CrI 161 000-359 000) confirmed cases in China as opposed to the 55 508 confirmed cases reported. INTERPRETATION: The case definition was initially narrow and was gradually broadened to allow detection of more cases as knowledge increased, particularly milder cases and those without epidemiological links to Wuhan, China, or other known cases. These changes should be taken into account when making inferences on epidemic growth rates and doubling times, and therefore on the reproductive number, to avoid bias. FUNDING: Health and Medical Research Fund, Hong Kong.

Vaccination is an important intervention to prevent influenza virus infection, but indirect protection of household members of vaccinees is not fully known. Here, we analyze a cluster household randomized controlled trial, with one child in each household randomized to receive influenza vaccine or placebo, for an influenza B epidemic in Hong Kong. We apply statistical models to estimate household transmission dynamics and quantify the direct and indirect protection of vaccination. Direct vaccine efficacy was 71%. The infection probability of unvaccinated household members in vaccinated households was only 5% lower than in control households, because only 10% of infections are attributed to household transmission. Even when that proportion rises to 30% and all children are vaccinated, we predict that the infection probability for unvaccinated household members would only be reduced by 20%. This suggests that benefits of individual vaccination remain important even when other household members are vaccinated.