Real-time spatial health surveillance: Mapping the UK COVID-19 epidemic

Richard Fry,Joe Hollinghurst,Helen R Stagg,Daniel A Thompson,Claudio Fronterre,Chris Orton,Ronan A Lyons,David V Ford,Aziz Sheikh,Peter J Diggle

doi:10.1016/j.ijmedinf.2021.104400

Abstract

Introduction The COVID-19 pandemic has highlighted the need for robust data linkage systems and methods for identifying outbreaks of disease in near real-time. Objectives The primary objective of this study was to develop a real-time geospatial surveillance system to monitor the spread of COVID-19 across the UK. Methods Using self-reported app data and the Secure Anonymised Information Linkage (SAIL) Databank, we demonstrate the use of sophisticated spatial modelling for near-real-time prediction of COVID-19 prevalence at small-area resolution to inform strategic government policy areas. Results We demonstrate that using a combination of crowd-sourced app data and sophisticated geo-statistical techniques it is possible to predict hot spots of COVID-19 at fine geographic scales, nationally. We are also able to produce estimates of their precision, which is an important pre-requisite to an effective control strategy to guard against over-reaction to potentially spurious features of ’best guess’ predictions. Conclusion In the UK, important emerging risk-factors such as social deprivation or ethnicity vary over small distances, hence risk needs to be modelled at fine spatial resolution to avoid aggregation bias. We demonstrate that existing geospatial statistical methods originally developed for global health applications are well-suited to this task and can be used in an anonymised databank environment, thus preserving the privacy of the individuals who contribute their data.

Highlights

Introduction The COVID19 pandemic has highlighted the need for robust data linkage systems and methods for identifying outbreaks of disease in near real-time
For the results presented in this paper we chose to map four summaries: the mean, a point prediction of prevalence; the 5% and 95% quantiles, which together measure the uncertainty associated with each point prediction; and the probability that the prevalence in the layer Super Output Area (LSOA) in question is greater than the country-wide average for each devolved nation
In the Supplementary Material we describe in detail how we developed and fitted the particular model that we used for our appli cation to the COVID-19 app data

Summary

Introduction

19 pandemic has highlighted the need for robust data linkage systems and methods for identifying outbreaks of disease in near real-time. Methods Using self-reported app data and the Secure Anonymised Information Linkage (SAIL) Databank, we demonstrate the use of sophisticated spatial modelling for nearreal-time prediction of COVID-19 prevalence at small-area resolution to inform strategic government policy areas. On 11th March 2020, the World Health Organization declared a pandemic of COVID-19 caused by the SARS-CoV-2 coronavirus [1]. By this date, the UK had reported 373 confirmed COVID-19 cases and six deaths [2]. A key example of this was the implementation of the first ’local lockdown’ in Leicester on 30th June 2020, in response to a cluster of COVID-19 accounting for approximately one in ten of all new disease cases across the country in the preceding week [6]

Objectives

Methods

Results

Discussion

Conclusion