VERDI recommendations for future research

Key takeaway: Pregnant women and children were generally excluded from clinical research early in the SARS-CoV-2 pandemic and, in many cases, the 2022 mpox outbreak. 

COVID-19 vaccines were not approved for pregnant women and children until late in the pandemic, leading to uncertainty about their effectiveness and safety and relatively low uptake. Similarly, these groups were often excluded from treatment trials. In high income settings, the epidemiology of the 2022 mpox outbreak meant that responses focused on adult MSM; however, several trials in other settings did recruit pregnant women and children. 

Key takeaway: Surveillance data were not always disaggregated by age group, and rarely by pregnancy status. 

To estimate age-specific effects of a pathogen and interventions, it is essential that data are available by age. Similarly, data by pregnancy status (and ideally additional characteristics such as trimester and pregnancy outcome) are needed. However, these breakdowns, in particular for pregnancy, were not always available, limiting the analyses that could be performed. 

Key takeaway: Researchers may be hesitant to include pregnant women and children in pandemic responses, including due to perceived ethical issues. 

There has been a recent paradigm shift, endorsed by the WHO and the EMA, which calls for the inclusion of pregnant women and children in clinical research, to protect them through research, rather than from research (link). The WHO developed a toolkit providing a central resource for documents and other materials to support inclusive research (link).

Key takeaway: VERDI capitalised on and strengthened existing infrastructure from previous studies of pregnant women and children, in Europe and globally, allowing faster evidence generation than establishing de novo studies. 

An example of this is the RECoVER household study, which ran in three European countries early in the pandemic, and was reactivated in the Netherlands as part of VERDI to study the newly emerged omicron variant in the household setting. The existing infrastructure meant that the protocol, ethics approval, participant-facing documents and study processes were already in place and so data collection could begin rapidly.  The new “Be Ready Now” consortium and the Partnership for Pandemic Preparedness in Europe will be addressing how to ensure that sufficient studies across at risk populations are kept “warm” so they are ready to respond to an infection risk in a timely way (link). 

Key takeaway: The evolution of the SARS-CoV-2 pandemic and mpox outbreak were impossible to predict and so some of the tasks proposed in the VERDI grant application became less relevant over the course of the study. 

For example, VERDI proposed a serological survey to be conducted amongst people attending sexual health clinics in Europe to monitor incident infections, and a registry of paediatric cases of mpox combining retrospective and prospective data collection. The decline in mpox case numbers by late 2022 meant that the serological survey and prospective data collection for the paediatric registry were no longer feasible. Changes in control measures and testing/reporting policy also had important impacts, for example the changes to notification requirements in Thailand during the pandemic meant that the approach to identifying index cases needed to be adapted. 

Key takeaway: The SARS-CoV-2 pandemic exemplified the importance of ongoing funding for observational studies, as variants of concern (VoC) had different symptoms and health consequences in children compared to adults over the course of the pandemic, and the issue of long COVID gradually emerged. 

Rapid changes in the epidemiological situation during the pandemic necessitated ongoing research studies, including in relation to the changing impacts on children. For example, as schools reopened, new variants emerged with different transmissibility and clinical features, and vaccines were rolled out initially to older age groups, age-specific attack rates and transmission dynamics changed. The European household study conducted in VERDI was able to collect data during periods dominated by different VoC, and to make comparisons between adults and children during these periods. 

People reporting long COVID symptoms early in the SARS-CoV-2 pandemic may have been doubted by health professionals and the media. Analyses of EHRs allowed analysis of longer-term outcomes such as long COVID over time.  Post-acute infection syndromes are a consequence of a wide variety of acute infections and so long-term syndromes following new infection outbreaks should be anticipated and planned for, early on. 

Key takeaway: Sharing of expertise and protocols between VERDI partners allowed the successful set up of studies in Thailand, based on a European protocol and adapted to the local context and stage of the outbreak. 

The protocol for the European RECOVER study was shared with Thai partners, who used their expertise and knowledge of the local context to adapt the design to their settings. Adaptations were needed, for example, to accommodate changes in notification requirements which meant that COVID-19 cases no longer needed to be reported to public health bodies. Laboratory protocols were also shared between Thai partners, and between Thai and European partners. Sharing of modelling expertise via a training workshop run by ISI further aided analysis of the study data. 

Key takeaway: Research on COVID in pregnancy and paediatrics was initially fragmented, with insufficient links between EU-funded consortia; the Cohort Coordination Board (CCB, established in February 2022) facilitated cross-consortia networking. 

The CCB provided an invaluable resource for COVID-19 cohort consortia to come together, present their studies, identify areas of overlap and synergy, and harmonise work. It has subsequently expanded to other infections. Cross-consortia “hop-on” style calls could be piloted to provide a means of funding increased integration between consortia (e.g. conjoining specific tasks) at the set-up phase. Through the CCB, VERDI has been able to advocate for the needs of pregnant women and children and their inclusion in studies within other consortia. 

Key takeaway: Close collaboration between researchers and government departments / agencies enhanced the policy relevance and uptake of research findings. 

Key examples of collaborations between VERDI researchers and Ministries of Health include the projects undertaken in Thailand and Estonia to establish linked pregnancy-infant datasets, using existing national electronic health datasets. In both cases, the process was facilitated by engagement with policymakers to communicate the value of the linked dataset as a resource for research and public health/surveillance bodies both during and outside pandemics. 

Key takeaway: The complex and dynamic nature of SARS-CoV-2 control measures and policy changes was not always thoroughly documented in public facing reports and websites. 

Knowledge of the timeline of local policy measures is crucial for evaluating their impact on transmission and other outcomes, but cannot be reliably reconstructed retrospectively if they are not systematically recorded. While country-level data were well recorded in the Oxford COVID-19 Government Response Tracker¹, some measures specific to countries or to regions within countries were not included. For example, regional information relevant to Estonian analyses that was not available included the timing of school closures in specific regions of the country and the timing of Finland’s border closure specifically to travellers from Estonia. Additionally, information on the lifting on the requirement for mandatory SARS-CoV-2 testing of pregnant women on admission for delivery was not available. Some countries, such as the UK (link), have published details of the timeline of interventions, but these may not be in well-known fora.  

¹Hale, T., et al., A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker). Nat Hum Behav, 2021. 5(4): p. 529-538.

Key takeaway: Lack of knowledge and awareness for community members and people with mpox, and lack of training for healthcare workers, led to intense multi-level stigma during the mpox outbreak. 

Lack of knowledge about mpox amongst the public and healthcare workers led to delays in seeking and receiving appropriate care, which may be particularly important for pregnant women; both patients and healthcare workers experienced stigma and emotional distress. 

Key takeaway: Where healthcare systems are weak and tools lacking, patients may be reluctant to seek care due to lack of trust/confidence in confidentiality, equity and/or effectiveness of services. 

Patients diagnosed with mpox were often reluctant to seek care or were concerned about the implications of doing so, for example isolation from family members. For those wishing to access testing, this was often expensive and time-consuming. However, attitudes towards vaccination were generally positive. Healthcare providers highlighted the need for mpox-specific guidelines and capacity strengthening. 

Key takeaway: COVID-19 vaccine coverage estimates specifically for pregnant women were rarely available. 

COVID-19 vaccine coverage is frequently published for people aged >60 years, but more rarely for pregnant women: for example, for the 2023/24 winter season, Ireland and Spain were the only EU/EEA countries to provide data on vaccination in pregnancy (link). Although defining an appropriate denominator for vaccine coverage in pregnancy is complex (link) developing approaches to this should be prioritised, ideally with stratification by key factors such as age, ethnicity and co-morbidities. Lacking these data makes it difficult to identify and tackle inequalities as well as to estimate the impact of vaccination during pregnancy. 

Key takeaway: Many settings had multiple EHR datasets available for children and pregnant women, but these were not always linked. 

In many cases, important data were collected in separate datasets, e.g. vaccines were often recorded in a vaccination registry and SARS-CoV-2 positive tests in another, while information on primary care contacts and hospital admissions were collected in pre-existing administrative databases. Linkage of these datasets (combining records for the same person based on a unique identifier) allowed robust analysis combining the data from the various sources. Further, data on pregnant women and their infants was rarely routinely linked, making it challenging to rapidly assess infant outcomes in the light of maternal characteristics during pregnancy. 

Key takeaway: In some settings, timelines for researchers accessing EHR datasets were too long for results to inform policy in a timely way; this is particularly an issue when pregnant women and children are excluded from other forms of research. 

Access to these datasets is, rightly, controlled, with access requests requiring a strong rationale, ethical approval, and data sharing agreements. The process of requesting and obtaining access can be lengthy and, in some cases, unclear, leading to delays in the submission and approval of applications.  

Key takeaway: Retrospectively including early cases in EHR-based studies was sometimes not permitted as consent had not been obtained, limiting access to data in the early stages of the pandemic. 

Routine care data from the early stages of an outbreak provide valuable information on clinical characteristics and outcomes of infection. However, in settings where there is no standard process for using routine care data for research purposes, concerns may arise about the use of these data for this purpose if consent was not obtained from the patient at the time of the healthcare contact. Frameworks for rapid review of research during health emergencies have been previously proposed¹.

¹ Tansey, C.M., et al., A framework for research ethics review during public emergencies. CMAJ, 2010. 182(14): p. 1533-7.

Key takeaway: EHR datasets can provide useful analyses of seasonal trends in infections in non-pandemic times, and their predictors, which in turn helps earlier assessment of the impact of a new pandemic infection on seasonal infections. 

VERDI analyses showed marked changes in infection-related hospital admission rates for children during the early stages of the pandemic when non-pharmaceutical interventions were implemented, extending into 2021. Comparison with pre-pandemic baseline, including secular trends, was crucial to properly interpret the rates observed during the pandemic. 

Key takeaway: Collecting sufficient data on an infection that is rare in a particular population (e.g. children or pregnant women) requires a large network and adapted data collection tools. 

Multiple surveillance networks exist for a variety of infections, collecting data including age and pregnancy status. However, the case report forms used in these networks were generally developed with the general population in mind, and are not designed for use with children or pregnant women. For example, very young children are unable to report several important signs and symptoms which are routinely collected for adults (e.g. muscle aches, nausea, psychological disturbances) and measures such as mobility and ability to self-care, while important for adults, are less relevant for infants and young children. Few surveillance systems collect detailed data on pregnancy or neonatal outcomes. Developing and/or adapting CRFs to capture this information is complex and requires expert advice and diverse perspectives, as well as dedicated staff to coordinate data collection; in many settings, additional funding is also needed. Suitable CRFs should be developed in advance to ensure rapid and inclusive data collection during the next outbreak, accounting for pregnant individuals, neonates, and children who are often overlooked. 

Key takeaway: Embedding research skills in clinical care settings, including integrating laboratory capacity, accelerates preparedness. 

For VERDI Partners in South Africa, the presence of an integrated research-clinical care infrastructure with teams already knowing each other, and protocols for approaching patients in place, were crucial to the setting up and running of their COVID-19 cohorts. 

Key takeaway: Whole genome sequencing (WGS) of SARS-CoV-2 provided insights into household transmission being the predominant source of SARS-CoV-2 infections within households in Thailand,  and led to reductions in estimates of household secondary attack rates as compared to estimates not incorporating genomic data. 

In the Chiang Mai Household Study, the SAR was 23.5% when all infections identified in the household were considered secondary cases, and 15.4% when secondary cases were included only if viral genomic sequencing was consistent with transmission from the index case. This demonstrates that in settings where household contacts may acquire infection outside the home, e.g. particularly in contexts in which households are not required to isolate if a member is infectious, transmissibility may be overestimated if this is not accounted for. 

Key takeaway: Some participants in household studies experienced stigma or discrimination from neighbours during enrolment. 

In addition to experienced stigma, the fear of stigma may have deterred eligible households from participating. This highlights the need for sensitive (and culturally appropriate) approaches to recruitment into research studies, but more fundamentally, for public health messaging to reduce stigma associated with infection and encourage individuals to seek healthcare where needed and, where appropriate, to participate in research.   

Key takeaway: Participatory surveillance can complement official channels by providing faster, more granular signals of emerging threats. 

While these citizen-driven platforms can act as early warning systems, their value depends on credibility, integration, and sustainability. On their own, participatory signals may suffer from biases in coverage, representativeness, and verification. However, when combined with traditional epidemiological surveillance, they can significantly improve situational awareness and response times. The challenge lies in bridging informal community data streams with institutional processes, ensuring that signals are validated, trusted, and rapidly translated into action. Further challenges include the need for ongoing maintenance, which requires sustainable funding. 

Key takeaway: Participatory surveillance platforms, if the coverage is adequate, can provide valuable and granular information not only about the prevalence of diseases among the population by age group but also on factors such as behavioural information, opinions, and attitudes towards interventions. This information can be a valuable complement to traditional surveillance data. 

Despite their potential, the effectiveness of participatory platforms relies on sustained engagement from the public and trust in the system. Without adequate communication and visibility, participation often diminishes outside crisis periods, limiting the representativeness and continuity of the data. Moreover, when data collection is not coupled with health education, misinformation can spread unchecked, undermining both the reliability of the platform and broader public health goals. To exploit their full value, these platforms must be positioned as long-term preparedness tools, integrated into policy frameworks and recognized as legitimate complements to official surveillance. This is already the case for some countries, e.g. UK, Germany, France, the Netherlands, Portugal.  

Key takeaway: Corpus linguistic approaches can provide insights into linguistic trends and guide more in-depth qualitative analysis of big data. 

Combining computational linguistic approaches and more traditional qualitative approaches allowed us to describe patterns of reporting on mpox and the impact of change in mpox nomenclature.  These statistical approaches provided insights into the use of language to report mpox, guiding further in-depth analysis.  This resulted in identifying potentially stigmatising discourse on mpox in 2022, with more neutral reporting in 2024. 

Key takeaway: Focus group discussions in the post outbreak/pandemic phase can provide important data on ongoing communication and support needs. 

VERDI focus group discussions were conducted after cases of mpox in the UK had declined substantially and after the introduction of a national vaccination programme. However, conducting research some time after the peak of the outbreak allowed participants to reflect on how it had impacted them, and to identify an ongoing need for information and advice about mpox and prevention. 

Key takeaway: Timely and ongoing information from trusted sources is essential during an outbreak pandemic. Social media is an important source of health information. 

Focus group discussions highlighted the need for trusted providers of information and the avoidance of sensationalism. Participants also reported that the sudden cessation of public health campaigns was confusing and that ongoing communication was needed even after media interest had waned. Several healthcare providers and scientists produced high-quality content on TikTok, although misinformation was also identified. 

Key takeaway: Modelling efforts were often ad hoc and data to inform models were often not available. 

Accurate prediction or evaluation of the impact of interventions on transmission requires robust data on which to base model assumptions. This includes traditional surveillance data and additional datasets such as on mobility and social mixing, all of which should be age-stratified in order to generate results relevant to different age groups (e.g. the contribution to transmission, or impact of vaccination). Baseline data on mobility and social mixing should be collected before the next pandemic. Specifically, key data needs include: i) timely incidence data for the population of interest, such as the weekly number of confirmed cases in schoolchildren, with the purpose of fitting a model to determine transmissibility of the pathogen in different settings; ii) viral genomic sequence data to provide information on transmission events; iii) estimates of contact patterns in the population of interest, e.g. empirical contacts in the school population to improve model accuracy, especially to quantify the impact of the disease transmission and the effectiveness of interventions, in terms of averted cases, and cost of the measure; iv) vaccination coverage among the population of interest, to estimate the immunity of the population to evaluate the impact of the disease spreading more effectively; v) infection/disease parameters, including age-specific estimates of individual susceptibility to infection, transmissibility of the infection, probability of developing symptoms if infected, and probability of a case being detected based on symptoms, to evaluate the dynamics of the disease spreading among the population of interest; vi) adherence rates to interventions, to estimate the effective impact of interventions in the population of interest; vii) mobility data, e.g. mobility fluxes from one spatial area to another or from one social context (home, workplace, school) to another, to accurately simulate the transmission dynamics of an infection in the population of interest. 

Key takeaway: Collaboration between epidemiology and modelling teams was sometimes hampered by the lengthy process of establishing data sharing agreements. 

During a public health crisis, delays in accessing high-quality data directly can result in slower situation assessments and less effective decision-making. Negotiating data sharing agreements in the midst of an emergency not only can delay crucial activities but also creates uncertainty about what can legally or ethically be shared. These bottlenecks create problems for collaborations, as modellers often rely on timely epidemiological data to calibrate forecasts, while epidemiologists benefit from modelling outputs to guide interventions. Establishing clear, trusted agreements in advance is essential to foster seamless cooperation when speed and accuracy matter most. 

Key takeaway: Reusing and adapting models developed for other infections accelerates analysis. 

It is crucial to build and maintain a core set of multi-disciplinary modelling teams that can quickly adapt pre-existing models to a new pathogen’s characteristics. These models should be flexible enough to incorporate various transmission routes, population structures, and intervention types. It would also be important to have a toolbox of targeted interventions for different settings (e.g., schools, workplaces, communities) and test their effectiveness and cost-benefit ratios using models. For schools, this means having protocols for proactive, regular screening ready for immediate deployment. This “ready-to-go” infrastructure would prevent the delay and fragmentation of modelling efforts that occurred during the COVID-19 pandemic. 

Key takeaway: Beyond routine laboratory serology or PCR testing for SARS-CoV-2, omics analysis within the context of household transmission studies added critical insights into the contribution of children to COVID-19 transmission in households. 

Within VERDI, a Netherlands VoC household transmission study was implemented swiftly as it capitalised on a previous protocol already executed in Europe.  The study was later adapted and replicated in Thailand.  When such a study is activated during an active outbreak, recruitment can be swift and genomic results provide critical insights into transmission risk in households. 

Key takeaway: Finger and heel prick techniques offer an attractive method of collecting sera from infants and children but limited volumes and cumbersome laboratory techniques limit the ability to recover adequate samples for detection of neutralising antibodies.  

The challenges of obtaining and testing adequate serum samples from infants and children are well recognised.  Novel technological developments, including microsampling and high-throughput testing platforms, offer alternative strategies which can be used in real world pandemic situations. 

Key takeaway: FAIR data principles must be integrated into health data management systems before a crisis, as ad hoc efforts during a pandemic are usually too slow. Legal and ethical aspects often took time to address during COVID-19, and as a consequence evidence generation was delayed. 

Studies included in VERDI collected data in a variety of formats, often not using recognised data standards. This limited the potential to conduct combined analyses across cohorts. Data sharing and pooled analyses are especially important for some sub-populations such as pregnant women, where case numbers may be small. Where data sharing occurred, the process was often lengthy and complex due to the need to establish data sharing agreements and memoranda of understanding. 

Key takeaway: Synthetic data generation can help epidemic responses by offering a privacy-preserving alternative to data sharing. 

Synthetic data generation is an alternative to data sharing when sharing of a dataset is not possible, for example due to privacy concerns. The aim is to create a dataset which preserves the statistical properties of the original data, but which does not contain information on real people and can therefore be shared without threats to confidentiality. VERDI research has assessed the feasibility of employing state-of-the-art data synthesis and anonymisation techniques to generate high-fidelity and privacy preserving synthetic datasets capable of meaningfully reproducing study results, thus benchmarking the potential of synthetic cohort data for public sharing (link).