Where in Bradford has been most vulnerable during the COVID-19 pandemic?
With high levels of deprivation, ill-health, and many other risk factors, Bradford Metropolitan District is recognised as one of the most vulnerable and hardest-hit parts of the country during the coronavirus pandemic; the Joint Biosecurity Centre identifies it as one of three local authorities caught in a “perfect storm” of “interconnected factors” that have kept infection rates “stubbornly high”. Yet stark inequalities within the district mean that different areas have been impacted in different ways, and a neighbourhood-level understanding of the pandemic is key to mitigating this impact as effectively as possible.
As part of the COVID-19 Scientific Advisory Group at the Bradford Institute for Health research, this project aimed to develop local insights in two ways: analysing spatiotemporal positive COVID-19 test data to examine how the pattern of infections in the district evolved over time, and using individual-level health and sociodemographic data to build an area-level picture of coronavirus mortality risk. Its findings will be communicated to policymakers, and used to inform local interventions and services planning, including the ongoing vaccine rollout.
Data and methods
Positive COVID-19 test data in Bradford Metropolitan District were provided by the local authority, and grouped by week and census area (lower layer super output areas, or LSOAs). This level of aggregation provided counts large enough to provide robust statistics, while retaining the spatial and temporal granularity necessary to conduct effective local analysis. A variety of statistics were used to analyse the spatial and temporal structure of these data, including the radius of gyration of positive tests about the population weighted centroid of the district, semivariograms, and local and global Moran’s I.
Routine anonymised event-level data for patients’ interactions with healthcare services, along with individual-level sociodemographic and geographic data, are hosted on the Connected Yorkshire platform, in line with the Observational Medical Outcomes Partnership Common Data Model. Using server-side queries and data processing, a cohort of Bradford individuals was constructed with linked health and sociodemographic fields corresponding to risk factors for coronavirus mortality. Hazard ratios for these risk factors were taken from Williamson et al. (2020).
For each individual in the cohort, the product of their hazard ratios for any desired combination of factors gives a personal score representing COVID-19 mortality risk due to those chosen factors. These scores were averaged to LSOA-level to produce maps of mean risk score by LSOA. For example, on average a cohort member in an LSOA with a mean risk score of 6 would be at twice the risk of COVID-19 mortality, compared to someone in an LSOA with a risk score of 3.
There was a clear trend in the positive test data, whereby cases were more spatially disordered during the COVID-19 peaks in spring 2020 and winter 2020/2021, and were more ordered during the intervening ‘lull’ in the pandemic over the summer of 2020. For example, values of the spatial autocorrelation coefficient (global) Moran’s I suggest a more random distribution of cases during these pandemic peaks, and more spatial structure during the months in between. That is to say, when things were bad they were bad more or less everywhere, but when district-wide infections were down there were clusters that remained disproportionately affected; more must be done to prevent sustained community transmission in these areas.
Mapping overall COVID-19 mortality risk, including all risk factors included in the dataset, shows a few individual high-risk LSOAs dotted around the district but no obvious overarching pattern. However, the factors contributing most to this risk vary across the district; some of these factors (age, certain health conditions) are accounted for in vaccine eligibility criteria and some (other health conditions, sociodemographic risk) are not. In particular, urban areas are at much higher risk due to factors not part of these criteria.
Value of the research
There are evident and immediate local policy implications from this research. Firstly, it must be remembered that when infections appeared to be dwindling at the overall level, at a smaller scale certain areas remained vulnerable. This fact should inform policy decisions made as we move between stages of the pandemic going forward, or indeed of future pandemics – for example, a transition to targeted local interventions may be appropriate as large-scale measures are eased.
Along with social distancing and other non-pharmaceutical interventions, an effective vaccination programme is critical in managing the effects of the pandemic. This project shows how the two largest population centres in the district, Bradford and Keighley, are not well served by the national vaccination strategy. Despite being at similar or higher overall risk to more rural parts of the district, they are de-emphasised in the vaccine rollout because of the particular combination of risk factors present.
Both strands of research speak to inequalities in the district – existing inequalities have begotten new ones in the age of COVID-19, and policy must be designed with redressing these inequalities as a primary objective.
Quote from project partner
“Harry Tata has contributed to the work of the Bradford COVID-19 Scientific Advisory Group, producing analysis that has been of use in understanding the spatial and temporal impact of COVID-19; in addition Harry has also provided support to the analysis of COVID risk factors in the Bradford population. Harry has achieved this in a six month placement, and has impressed with his analytical capabilities and application; making a real impact to urgent and contemporary public health issues in Bradford.”
Professor John Wright
Director, Bradford Institute for Health Research
- Even as overall case numbers dropped, localised areas remained badly affected.
- By ignoring important sociodemographic risk factors, the current vaccine strategy leaves at-risk areas behind.
- Health informatics
Harry Tata – Data Scientist Intern, Leeds Institute for Data Analytics
Dr Brian Kelly – Senior Research Fellow, Bradford Teaching Hospitals NHS Foundation Trust
Dr Bo Hou – Senior Research Fellow, Bradford Teaching Hospitals NHS Foundation Trust
Bradford Institute for Health Research (BIHR)
Williamson, E.J., Walker, A.J., Bhaskaran, K. et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature 584, 430–436 (2020). https://doi.org/10.1038/s41586-020-2521-4