Assessing household mixing during the pandemic
Key data showing how people adhered to the UK government’s COVID-19 laws preventing household mixing in England have been analysed for the first time.
Leeds researchers used regional mobile phone location data to assess the level of household mixing between 23 March 2020 and 23 May 2021.
The data was provided by Cuebiq, which collects location information through apps for use by marketers and researchers. Users of these apps opted in and consented to sharing location data with Cuebiq. Leeds’s team worked with Cuebiq’s Data for Good programme, accessing bespoke anonymous regional level data rather than household or individual data.
Spatial data analysis can help improve the effectiveness of pandemic policy to reduce spread of the virus and limit the negative impacts of lockdowns.
The findings, published today in Scientific Reports, showed:
- People changed their behaviour markedly during all lockdown periods, and household visitation dropped significantly from early 2020 levels
- Later lockdowns in November and January showed higher rates of household mixing, indicative of a 'diminishing returns' effect from repeated imposition of restrictions
- Household visitations increased significantly after February 2021, in line with vaccination rates increasing.
Research supervisor Ed Manley, Professor of Urban Analytics in the University of Leeds’s School of Geography and a Fellow of The Alan Turing Institute, said: “Mixing within enclosed spaces is a major cause of COVID-19 transmission, and interventions have been imposed to limit non-home household visitation.
“Despite discussion of policy adherence and behavioural fatigue, to date there has been no nation-scale analysis of household mixing, and its variation over time and regions.
“Our findings show that there was a lot of variation in household visitation during 2020 which corresponded with events throughout the year, such as the vaccine rollout, relaxing of outdoor mixing rules and national lockdowns.
“Spatial data analysis can help improve the effectiveness of pandemic policy to reduce spread of the virus and limit the negative impacts of lockdowns.”
Researchers constructed national and regional profiles by aggregating individual data within the Cuebiq platform. Within the closed analysis workflow, ‘Home and Work’ locations were established as the two most frequently visited locations by each individual within a 2-week period, calculated at the beginning of each month. Visits to points of interest, such as churches, restaurants and sports stadiums, were removed, as were visits to parks and woodland.
The removal of home, work, points of interest and green space locations left a remaining set of 'unclassified' visit points, 89% of which were within 50 metres of a residential building. These were classified as household visitations – which could also include visits to private gardens, which was permitted under the policy guidance at various points during the pandemic.
Household mixing levels during the pandemic were compared with January and February 2020, before lockdown rules came into force, to establish a baseline.
Clear geographic differences in the extent of household mixing were observed, with consistently higher levels in the London and South-East regions and some urban areas including Manchester, Cambridge and Leicester.
The lowest levels were found in rural authority areas, such as North-East Derbyshire, West Devon and Mid Suffolk.
Eat Out to Help Out
There was a flattening of household mixing in August 2020, which coincided with the Eat Out to Help Out programme, suggesting that encouraging people to mix in safe spaces may have been beneficial in terms of reducing contact between different households.
Professor Manley said: “Household visitation doesn’t equate to malicious noncompliance - issues such as caring for relatives or children or helping with shopping could be behind it. But there may be some areas where household visitation is high and a more nuanced approach is needed.
“This data provides a monitoring tool which can be used to inform policy decisions; providing a better understanding of where and when different policy is needed. By understanding how regions differ we can dig deeper into what is happening and hone the message accordingly.”
The research was supported by the ESRC Consumer Data Research Centre (CDRC), located in the Leeds Institute for Data Analytics (LIDA).