Differences in virus and immune dynamics for SARS-CoV-2 Delta and Omicron infections by age and vaccination histories (preprint)

Vaccination against COVID-19 was integral to controlling the pandemic that persisted with the continuous emergence of SARS-CoV-2 variants. Using a mathematical model describing SARS-CoV-2 within-host infection dynamics, we estimate differences in virus and immunity due to factors of infecting variant, age, and vaccination history (vaccination brand, number of doses and time since vaccination). We fit our model in a Bayesian framework to upper respiratory tract viral load measurements obtained from cases of Delta and Omicron infections in Singapore, of whom the majority only had one nasopharyngeal swab measurement. With this dataset, we are able to recreate similar trends in URT virus dynamics observed in past within-host modelling studies fitted to longitudinal patient data. We found that Omicron had greater infection potential than Delta, indicating greater propensity to establish infection. Moreover, heterogeneities in infection dynamics across patient subgroups could be recreated by fitting immunity-related parameters as vaccination history-specific, with or without age modification. Our model results are consistent with the notion of immunosenescence in SARS-CoV-2 pathogenesis in elderly individuals, and the issue of waning immunity with increased time since last vaccination. Lastly, vaccination was not found to subdue virus dynamics in Omicron infections as well as it had for Delta infections. This study provides insight into the influence of vaccine-elicited immunity on SARS-CoV-2 within-host dynamics, and the interplay between age and vaccination history. Furthermore, it demonstrates the need to disentangle host factors and changes in pathogen to discern factors influencing virus dynamics. Finally, this work demonstrates a way forward in the study of within-host virus dynamics, by use of viral load datasets including a large number of patients without repeated measurements.

Predicting public mental health needs in a crisis using situational indicators and social media emotions: A Singapore big data study (preprint)

Mental health issues and needs have increased substantially since the onset of the COVID-19 pandemic. However, health policy and decision-makers do not have adequate data and tools to predict population-level mental health demand, especially amid a crisis. This study investigates whether situational indicators and social media emotions can be effectively used to predict public mental health needs. We collected time-series data from multiple sources in Singapore between 1 July 2020 and 31 December 2021, including daily-level records of situation indicators, emotions expressed on social media, and mental health needs measured by the number of public visits to the emergency room of the country's largest psychiatric hospital, and use of government-initiated online mental health self-help portal. Compared to mental health needs data alone, social media emotions were found to have significant Granger-causality effects with as early as four to five days lag length. Each resulted in a statistically significant enhancement in predicting the public's visits to the emergency room and the online self-help portal (e.g., Facebook Anger Count on emergency room visits, χ2 = 13·7, P = ·0085**). In contrast, situational indicators such as daily new cases had Granger-causality effects (χ2 = 10·3, P = ·016*) with a moderate lag length of three days. The findings indicate that emotions algorithmically extracted from social media platforms can provide new indicators for tracking and forecasting population-level mental health states and needs.