On Monday 25 April 2022, statisticians, epidemiologists, and data scientists from across the island of Ireland gathered at NUI Galway for a workshop on various aspects of the statistics of the Covid-19 pandemic, including data issues, statistical modelling, and communication of statistical findings. The workshop was funded by the Irish Statistical Association, with support from the Medical Section of the Royal Statistical Society.
The day was to be introduced by Professor Máire Connolly of NUI Galway, who leads the PANDEM-2 project on pandemic preparedness, but unfortunately she was struck down by Covid herself. So while she joined the workshop online, her collaborator, Professor Jim Duggan of NUI Galway, kindly stepped in and gave an excellent introduction, emphasising the importance of data, statistics and modelling for understanding and responding to Covid-19.
The first speaker was Dr Darren Dahly from the HRB Clinical Research Facility at University College Cork, who spoke on research waste in Covid-19 prediction models. A BMJ review article on prediction models for diagnosis and prognosis of Covid-19 found serious issues in the majority of published research included in the review, including poor reporting, small sample sizes, and high risk of bias. Worryingly, poorly calibrated prediction models are not only wasteful, but might also contribute to patient harm.
This was followed with two talks on modelling the spread of the SARS-CoV-2 in Ireland, arising from research funded by SFI Covid-19 Rapid Response Grants. Dr James Sweeney from the University of Limerick spoke on a compartmental model that incorporated age structure. This allowed for more accurate predictions of case numbers, by accounting for the different levels of social mixing that occur across various age groups. Dr Nicola Fitz-Simon from the HRB Clinical Research Facility at NUI Galway described a discrete-time compartmental model that allows incorporation of more realistic assumptions about durations groups of people spend in each compartment. The model was fitted using open-source data from both Northern Ireland and the Republic of Ireland. Dr Fitz-Simon showed how the model could be used to explore various scenarios.
The morning session finished with a talk by Dr Anthony Masters, RSS statistical ambassador and co-author with Professor David Spiegelhalter of Covid by Numbers. Dr Masters spoke about the challenges of communicating Covid-19 statistics, and included many striking examples of when statistics were poorly and even erroneously communicated. For example, the number of covid tests posted out to the public was used to claim that the UK testing target had been met - however posting out tests does not mean that they were completed. His talk concluded with lessons for how to improve the communication of statistical findings, especially to the public via the media.
After lunch, Professor Adele Marshall of Queens University Belfast spoke on an agent-based model incorporating change-point detection, and gave a live demonstration of the model predictions. The model captures the impact of different types of non-pharmaceutical interventions and their effectiveness during different waves of the pandemic. While the initial model required 18 days to converge, once it had learned the parameters it took only a matter of seconds to predict new scenarios.
This was followed by three talks focusing more on Covid-19 data. Dr Catherine Timoney of HSE West described a network analysis of a Covid-19 outbreak in the student population of Galway in early 2021, using the Computerised Infectious Disease Reporting (CIDR) data. Dr Timoney presented some very interesting network visualisations showing links detected between individuals. Dr Lisa Domegan of the Health Protection Surveillance Centre spoke about mortality surveillance in the Republic of Ireland. This included excess mortality and how the major causes of excess mortality varied during different waves of the pandemic. Dr Jos Ijpelaar of the Northern Ireland Statistical Research Agency (NISRA) was the one remote speaker of the day, and he described mortality surveillance in Northern Ireland. Dr Ijpelaar discussed excess mortality from the Northern Ireland perspective, which provided a nice comparison with the Republic of Ireland approach.
After a break for coffee, the final speaker of the day, Professor Cathal Walsh from the University of Limerick, spoke on varieties of compartmental models and agent-based models that had been used to understand the dynamics of disease spread and make scenario projections in the Republic of Ireland. An important aspect of the approach to understanding the pandemic and inform decision-making was the network of existing research projects, public health and other relevant bodies working together. Recalling the advice of Richard James Hayes, the brilliant Irish WWII code-breaker, there is a need for adequate resourcing to deal with ongoing and future emergencies.
The floor was then opened for questions and discussion. One of the main issues for discussion was around the quality and availability of Republic of Ireland Covid-19 data for researchers. The question: ‘can we learn from other countries that have greater availability of data for research?’ was posed as a problematic and ongoing issue. Food for thought and further discussion and research was provided by comments from other participants, on a variety of topics from modelling assumptions to the role of statisticians in the response to the pandemic.
The workshop ended with applause and thanks to all the speakers, participants both in person and online, and to Dr Nicola Fitz-Simon and Dr Laura Boyle for organising the event.