Literature Review on Epidemiological Modelling, Spatial Modelling and Artificial Intelligence for COVID-19

Introduction: Following the outbreak of Coronavirus (COVID-19) in Wuhan, China in December 2019, the World Health Organisation (WHO) has declared this infectious disease as a pandemic. Unlike previous infectious outbreaks such as Severe Acute Respiratory Syndrome (SARS) and Middle Eastern Respiratory syndrome (MERS), the high transmission rate of COVID-19 has resulted in worldwide spread. The countries with the highest recorded incidence and mortality rates are the US and UK.

Rationale/Objective: This review will compare studies that have used epidemiological models for disease forecasting and other models that have identified sociodemographic factors associated with COVID-19. We will evaluate several models, from basic equation-based mathematical models to more advanced machine-learning ones. Our expectation is that by identifying high impact models used by policy makers and discussing their limitations, we can identify possible areas for future research.

Evidence Review: The bibliographic database google scholar was used to search keywords such as ‘COVID-19’, ‘epidemiological modelling’ and ‘machine learning’. We examined data review articles, research studies and government-released articles.

Results: We identified that the current SEIR model used by the UK government lacked the spatial modelling to enable an accurate prediction of disease spread. We discussed that machine-learning systems which can identify high-risk groups can be used to establish the disparities in COVID-19 death in BAME groups. We found that most of the data hungry AI models used were limited by the lack of datasets available.

Conclusion: In conclusion, advances in AI methods for infectious disease have overcome challenges presented in mathematical models. Whilst limitations do exist, when optimised, these highly advanced models have a great potential in public health surveillance, particularly infectious disease transmission.