[Frontiers in Public Health, 4 January 2023]

The co-circulation of two respiratory infections with similar symptoms in a population can significantly overburden a healthcare system by slowing the testing and treatment. The persistent emergence of contagious variants of SARS-CoV-2, along with imperfect vaccines and their waning protections, have increased the likelihood of new COVID-19 outbreaks taking place during a typical flu season. Here, we developed a mathematical model for the co-circulation dynamics of COVID-19 and influenza, under different scenarios of influenza vaccine coverage, COVID-19 vaccine booster coverage and efficacy, and testing capacity. We investigated the required minimal and optimal coverage of COVID-19 booster (third) and fourth doses, in conjunction with the influenza vaccine, to avoid the coincidence of infection peaks for both diseases in a single season. We show that the testing delay brought on by the high number of influenza cases impacts the dynamics of influenza and COVID-19 transmission. The earlier the peak of the flu season and the greater the number of infections with flu-like symptoms, the greater the risk of flu transmission, which slows down COVID-19 testing, resulting in the delay of complete isolation of patients with COVID-19 who have not been isolated before the clinical presentation of symptoms and have been continuing their normal daily activities. Furthermore, our simulations stress the importance of vaccine uptake for preventing infection, severe illness, and hospitalization at the individual level and for disease outbreak control at the population level to avoid putting strain on already weak and overwhelmed healthcare systems. As such, ensuring optimal vaccine coverage for COVID-19 and influenza to reduce the burden of these infections is paramount. We showed that by keeping the influenza vaccine coverage about 35% and increasing the coverage of booster or fourth dose of COVID-19 not only reduces the infections with COVID-19 but also can delay its peak time. If the influenza vaccine coverage is increased to 55%, unexpectedly, it increases the peak size of influenza infections slightly, while it reduces the peak size of COVID-19 as well as significantly delays the peaks of both of these diseases. Mask-wearing coupled with a moderate increase in the vaccine uptake may mitigate COVID-19 and prevent an influenza outbreak.

[MBio, 18 April 2023]

Immune imprinting is a driver known to shape the anti-hemagglutinin (HA) antibody landscape of individuals born within the same birth cohort. With the HA and neuraminidase (NA) proteins evolving at different rates under immune selection pressures, anti-HA and anti-NA antibody responses since childhood influenza virus infections have not been evaluated in parallel at the individual level. This is partly due to the limited knowledge of changes in NA antigenicity, as seasonal influenza vaccines have focused on generating neutralizing anti-HA antibodies against HA antigenic variants. Here, we systematically characterized the NA antigenic variants of seasonal A(H1N1) viruses from 1977 to 1991 and completed the antigenic profile of N1 NAs from 1977 to 2015. We identified that NA proteins of A/USSR/90/77, A/Singapore/06/86, and A/Texas/36/91 were antigenically distinct and mapped N386K as a key determinant of the NA antigenic change from A/USSR/90/77 to A/Singapore/06/86. With comprehensive panels of HA and NA antigenic variants of A(H1N1) and A(H1N1)pdm09 viruses, we determined hemagglutinin inhibition (HI) and neuraminidase inhibition (NI) antibodies from 130 subjects born between 1950 and 2015. Age-dependent imprinting was observed for both anti-HA and anti-NA antibodies, with the peak HI and NI titers predominantly detected from subjects at 4 to 12 years old during the year of initial virus isolation, except the age-independent anti-HA antibody response against A(H1N1)pdm09 viruses. More participants possessed antibodies that reacted to multiple antigenically distinct NA proteins than those with antibodies that reacted to multiple antigenically distinct HA proteins. Our results support the need to include NA proteins in seasonal influenza vaccine preparations.

[BMJ Open Respiratory Research, 2 February 2023]

Introduction: We compared the population rate of COVID-19 and influenza hospitalisations by age, COVID-19 vaccine status and pandemic phase, which was lacking in other studies.

Method: We conducted a population-based study using hospital data from the province of British Columbia (population 5.3 million) in Canada with universal healthcare coverage. We created two cohorts of COVID-19 hospitalisations based on date of admission: annual cohort (March 2020 to February 2021) and peak cohort (Omicron era; first 10 weeks of 2022). For comparison, we created influenza annual and peak cohorts using three historical periods years to capture varying severity and circulating strains: 2009/2010, 2015/2016 and 2016/2017. We estimated hospitalisation rates per 100 000 population.

Results: COVID-19 and influenza hospitalisation rates by age group were ‘J’ shaped. The population rate of COVID-19 hospital admissions in the annual cohort (mostly unvaccinated; public health restrictions in place) was significantly higher than influenza among individuals aged 30–69 years, and comparable to the severe influenza year (2016/2017) among 70+. In the peak COVID-19 cohort (mostly vaccinated; few restrictions in place), the hospitalisation rate was comparable with influenza 2016/2017 in all age groups, although rates among the unvaccinated population were still higher than influenza among 18+. Among people aged 5–17 years, COVID-19 hospitalisation rates were lower than/comparable to influenza years in both cohorts. The COVID-19 hospitalisation rate among 0–4 years old, during Omicron, was higher than influenza 2015/2016 and 2016/2017 and lower than 2009/2010 pandemic.

Conclusions: During first Omicron wave, COVID-19 hospitalisation rates were significantly higher than historical influenza hospitalisation rates for unvaccinated adults but were comparable to influenza for vaccinated adults. For children, in the context of high infection levels, hospitalisation rates for COVID-19 were lower than 2009/2010 H1N1 influenza and comparable (higher for 0–4) to non-pandemic years, regardless of the vaccine status.

[Science Advances, 3 November 2023]

The seasonal influenza (flu) vaccine is designed to protect against those influenza viruses predicted to circulate during the upcoming flu season, but identifying which viruses are likely to circulate is challenging. We use features from phylogenetic trees reconstructed from hemagglutinin (HA) and neuraminidase (NA) sequences, together with a support vector machine, to predict future circulation. We obtain accuracies of 0.75 to 0.89 (AUC 0.83 to 0.91) over 2016–2020. We explore ways to select potential candidates for a seasonal vaccine and find that the machine learning model has a moderate ability to select strains that are close to future populations. However, consensus sequences among the most recent 3 years also do well at this task. We identify similar candidate strains to those proposed by the World Health Organization, suggesting that this approach can help inform vaccine strain selection.

[Frontiers in Digital Health, 28 June 2023]

Background: The use of social media data provides an opportunity to complement traditional influenza and COVID-19 surveillance methods for the detection and control of outbreaks and informing public health interventions.

Objective: The first aim of this study is to investigate the degree to which Twitter users disclose health experiences related to influenza and COVID-19 that could be indicative of recent plausible influenza cases or symptomatic COVID-19 infections. Second, we seek to use the Twitter datasets to train and evaluate the classification performance of Bidirectional Encoder Representations from Transformers (BERT) and variant language models in the context of influenza and COVID-19 infection detection.

Methods: We constructed two Twitter datasets using a keyword-based filtering approach on English-language tweets collected from December 2016 to December 2022 in Saskatchewan, Canada. The influenza-related dataset comprised tweets filtered with influenza-related keywords from December 13, 2016, to March 17, 2018, while the COVID-19 dataset comprised tweets filtered with COVID-19 symptom-related keywords from January 1, 2020, to June 22, 2021. The Twitter datasets were cleaned, and each tweet was annotated by at least two annotators as to whether it suggested recent plausible influenza cases or symptomatic COVID-19 cases. We then assessed the classification performance of pre-trained transformer-based language models, including BERT-base, BERT-large, RoBERTa-base, RoBERT-large, BERTweet-base, BERTweet-covid-base, BERTweet-large, and COVID-Twitter-BERT (CT-BERT) models, on each dataset. To address the notable class imbalance, we experimented with both oversampling and undersampling methods.

Results: The influenza dataset had 1129 out of 6444 (17.5%) tweets annotated as suggesting recent plausible influenza cases. The COVID-19 dataset had 924 out of 11939 (7.7%) tweets annotated as inferring recent plausible COVID-19 cases. When compared against other language models on the COVID-19 dataset, CT-BERT performed the best, supporting the highest scores for recall (94.8%), F1(94.4%), and accuracy (94.6%). For the influenza dataset, BERTweet models exhibited better performance. Our results also showed that applying data balancing techniques such as oversampling or undersampling method did not lead to improved model performance.

Conclusions: Utilizing domain-specific language models for monitoring users’ health experiences related to influenza and COVID-19 on social media shows improved classification performance and has the potential to supplement real-time disease surveillance.

[Science Advances, 3 November 2023]

The seasonal influenza (flu) vaccine is designed to protect against those influenza viruses predicted to circulate during the upcoming flu season, but identifying which viruses are likely to circulate is challenging. We use features from phylogenetic trees reconstructed from hemagglutinin (HA) and neuraminidase (NA) sequences, together with a support vector machine, to predict future circulation. We obtain accuracies of 0.75 to 0.89 (AUC 0.83 to 0.91) over 2016–2020. We explore ways to select potential candidates for a seasonal vaccine and find that the machine learning model has a moderate ability to select strains that are close to future populations. However, consensus sequences among the most recent 3 years also do well at this task. We identify similar candidate strains to those proposed by the World Health Organization, suggesting that this approach can help inform vaccine strain selection.

[Vaccine, 29 September 2023]

Background: Influenza vaccines prevent influenza-related morbidity and mortality; however, suboptimal vaccine effectiveness (VE) of non-adjuvanted trivalent inactivated influenza vaccine (naTIV) or quadrivalent formulations in older adults prompted the use of enhanced products such as adjuvanted TIV (aTIV). Here, the VE of aTIV is compared to naTIV for preventing influenza-associated hospitalization among older adults.

Methods: A test-negative design study was used with pooled data from the 2012 to 2015 influenza seasons. An inverse probability of treatment (IPT)-weighted logistic regression estimated the Odds Ratio (OR) for laboratory-confirmed influenza-associated hospitalization. VE was calculated as (1-OR)*100% with accompanying 95% confidence intervals (CI).

Results: Of 7,101 adults aged ≥ 65, 3,364 received naTIV and 526 received aTIV. The overall VE against influenza hospitalization was 45.9% (95% CI: 40.2%–51.1%) for naTIV and 53.5% (42.8%–62.3%) for aTIV. No statistically significant differences in VE were found between aTIV and naTIV by age group or influenza season, though a trend favoring aTIV over naTIV was noted. Frailty may have impacted VE in aTIV recipients compared to those receiving naTIV, according to an exploratory analysis; VE adjusted by frailty was 59.1% (49.6%–66.8%) for aTIV and 44.8% (39.1%–50.0%) for naTIV. The overall relative VE of aTIV to naTIV against laboratory-confirmed influenza hospital admission was 25% (OR 0.75; 0.61–0.92), demonstrating statistically significant benefit favoring aTIV.

Conclusions: Adjusting for frailty, aTIV showed statistically significantly better protection than naTIV against influenza-associated hospitalizations in older adults. In future studies, it is important to consider frailty as a significant confounder of VE.

[Zoonoses and Public Health, 8 February 2024]

Introduction: Public health preparedness is based on timely and accurate information. Time series forecasting using disease surveillance data is an important aspect of preparedness. This study compared two approaches of time series forecasting: seasonal auto-regressive integrated moving average (SARIMA) modelling and the artificial neural network (ANN) algorithm. The goal was to model weekly seasonal influenza activity in Canada using SARIMA and compares its predictive accuracy, based on root mean square prediction error (RMSE) and mean absolute prediction error (MAE), to that of an ANN.

Methods: An initial SARIMA model was fit using automated model selection by minimizing the Akaike information criterion (AIC). Further inspection of the autocorrelation function and partial autocorrelation function led to ‘manual’ model improvements. ANNs were trained iteratively, using an automated process to minimize the RMSE and MAE.

Results: A total of 378, 462 cases of influenza was reported in Canada from the 2010–2011 influenza season to the end of the 2019–2020 influenza season, with an average yearly incidence risk of 20.02 per 100,000 population. Automated SARIMA modelling was the better method in terms of forecasting accuracy (per RMSE and MAE). However, the ANN correctly predicted the peak week of disease incidence while the other models did not.

Conclusion: Both the ANN and SARIMA models have shown to be capable tools in forecasting seasonal influenza activity in Canada. It was shown that applying both in tandem is beneficial, SARIMA better forecasted overall incidence while ANN correctly predicted the peak week.