Insights & Applications
Mobile operator data can support government and public decision making during the COVID-19 pandemic.
We have identified below five key areas of applications which would benefit from mobility insights extracted from CDR data:
Monitoring the primary effects of mobility and social distancing interventions
Monitoring the side effects of interventions
Identifying routine mobility patterns (to plan interventions and assess risk levels)
Monitoring changes in density of population (dynamic population mapping)
Mobility data as an input to predictive models and analyses with other data
Monitoring the primary effects of mobility
and social distancing interventions
and social distancing interventions
We propose indicators that aim to provide a measure of change in mobility following specific government interventions and their announcement.
These indicators may be used to assess whether restrictions have had the expected effect of reducing travel, dispersion, and population mixing.
They are not a measure of the number of people who do or do not comply with mobility restrictions. Restrictions have a number of exemptions which cannot be quantified using CDR data alone. Exemptions include, for example, key workers (e.g. health sector, law enforcement, military, maintenance of essential service, supply chain of essential products), people returning home, people supporting their families, people in need of health care, etc. Therefore, as we cannot quantify exemptions, we cannot quantify compliance.
Example of applications
Assessing limitations on travel distance
Examples of insights provided:
How much has long-distance travel decreased?
Do regions receive fewer incoming visitors?
Indicators required
Inter-regional travel: distribution of distances travelled
Inter-regional travel: dispersion
Population mixing
Assessing the promotion or regulation about work from home
Examples of insights provided:
How much have urban ‘commuting’-type trips been reduced?
Has population mixing been reduced in urban centres?
Indicators required
Hourly presence and variance, per area
Crowdedness
Intra-regional travel distribution
Number of regions visited, per home location
Population mixing
Assessing the effects of lockdown
Examples of insights provided:
How much have urban trips and population mixing decreased?
Do more people appear to remain at home more than before?
Indicators required
Intra-regional travel distribution
Number of regions visited, per home location
Number of subscribers visiting only their home region
Population mixing: visitors vs residents
Crowdedness
Assessing the effects of curfew
Examples of insights provided:
How much has travel reduced during night time?
Indicators required
Daytime / nighttime presence ratio
Intra- and inter-regional travel
Crowdedness
Assessing the effects of closing of public places
Examples of insights provided:
By how much does the closure of specific establishments (religion, entertainment, shops, education) reduce travel and population mixing?
Indicators required
Hourly presence and variance, per area
Crowdedness
Intra-regional travel
Number of regions visited, per home location
Population mixing
Assessing bans on public gatherings and events
Examples of insights provided:
How much have crowd sizes and crowd frequencies decreased?
Indicators required
Crowdedness: crowd size and frequency
Monitoring the side effects of
interventions
interventions
Interventions designed to reduce mobility and increase social distancing may have unintended opposite effects, which negatively impact the control of COVID-19 or have negative social impacts. For example, research conducted in New York showed that the introduction of school closures in New York resulted in increased activity at grocery, shopping, food and outdoor places. In addition, evidence from Italy, France and Ghana demonstrated that there were large scale movements of people before lockdowns were implemented.
We propose indicators that aim to provide a measure of unintended side effects following government interventions. These indicators will help planners assess whether interventions should be modified or ceased. In addition, such indicators would also help planners form a more accurate estimate of the likely impact an intervention would have, accounting for these side effects and supporting longer term planning efforts.
Example of applications
Home relocations
Examples of insights provided:
How many people have relocated (changed their -presumed- region of residence) as a result of interventions, announcements or news?
How many people are leaving urban centres for rural areas?
Indicators required
Number of subscribers who changed their home location
Number of subscribers who relocated between any two regions (origin-destination flow)
Crowds formation
Examples of insights provided:
Are some places visited by a larger number of people following interventions? (e.g. shops after school closure)
Are some places in which visitor numbers were not assumed to change, visited by a less number of people following interventions? (e.g. hospitals)
Indicators required
Hourly presence, per area
Crowdedness
Intra-regional travel
Population mixing
Duration of disruption
Examples of insights provided:
How long before a ‘new normal’ settles in and daily repeated patterns re-appear?
Indicators required
Hourly and daily variance of subscriber presence, per area
Travel range
Examples of insights provided:
Do people tend to travel longer distances?
Indicators required
Inter-regional travel: distribution of distances travelled
Geographic dispersion
Examples of insights provided:
Do the interventions, announcements, news, result in more dispersion?
Indicators required
Inter-regional travel: dispersion
Number of regions visited, per home location
Population mixing
Effect of border closure
Examples of insights provided:
Are many citizens coming back to the country?
Indicators required
TBC: This may not be measurable with CDRs. It depends on several factors (information on roaming, duration of SIM validity, frequency of phone usage, ..), and we are currently investigating possible indicators of new SIMs and new arrivals to points of entry
Identifying routine mobility patterns
(to plan interventions and assess risk levels)
(to plan interventions and assess risk levels)
We propose indicators which extract current patterns of mobility, helping to identify hotspots (places with receiving large crowds and with high population mixing), most travelled routes and secluded regions.
This information would help decision-makers to plan interventions and restrictions, and to target areas to send information messages to.
In addition, the indicators will enable planners to conduct scenario testing for different types of measures (restrictions on travel, closures of public places) and for the staged relaxation of measures.
Example of applications
Selecting and prioritising locations for implementing mobility restrictions
Examples of insights provided:
Which locations receive large crowds and/or have a large amount of population mixing (hotspots)?
What are the main travelled routes and routes linking hotspots?
Are there areas in the country already partially shielded from the rest (secluded)?
Indicators required
Hotspot locations
Crowdedness - size and frequency
Population mixing - mixing factor
Regional connectivity structure:
Isolated clusters of regions
Main travelled routes
Main travelled routes through hotspots
Targeting locations for communications (where to send information, e.g. via SMS, voice or billboards)
Examples of insights provided:
Which locations receive large crowds and/or have a large amount of population mixing (hotspots)?
Which locations do people who travel a lot come from?
Indicators required
Hotspot locations
Crowdedness - size and frequency
Population mixing - mixing factor
Inter-regional travel - number of regions visited, per home location
Identifying places where people may be at higher risk of spreading or contracting the virus
(based on mobility profile only - not infectious cases)
Examples of insights provided:
Which locations receive large crowds and/or have a large amount of population mixing (hotspots)?
What are the main travelled routes and routes linking hotsports?
Indicators required
Hotspot locations
Crowdedness - size and frequency
Population mixing - mixing factor
Regional connectivity structure:
Main travelled routes through hotspots
Estimating the number of people in each region who may be at higher risk of spreading or contracting the virus?
Examples of insights provided:
How many people in each area visit a hotspot / multiple hotspots?
How many people in each area visit high-mixing areas?
Indicators required
Number of residents of hotspots
Average number of hotspots visited per subscriber, obtained from:
Crowdedness
Population mixing
Inter-regional travel - number of regions visited, per home location
Monitoring changes in density of population
(dynamic population mapping)
(dynamic population mapping)
There can be large scale movements of populations as a result of the COVID-19 epidemic, with, for example, countries experiencing high levels of movements from urban to rural areas. Such changes are important for resource allocation decisions within and outside the health care sector. They can also have unintended impacts on public services, food supplies and critical infrastructure.
We propose indicators that monitor the changes in population densities from CDRs which, combined with existing population estimates, will help provide more accurate estimates of population distributions during the epidemic. This will provide a useful indicator both during the outbreak and after restrictions have been lifted, as planners will be able to understand how long it takes for populations to return to the pre-crisis state.
Example of applications
Changes to the population size of each region during the crisis
(either driven by interventions or news or events related to the epidemic)
Examples of insights provided:
How many people move out of their home and relocate to a new region?
Indicators required
Changes to resident subscriber population of each region (weekly)
Duration of relocations
Examples of insights provided:
Does the population distribution return to its pre-crisis state and how long does it take?
Indicators required
Average time for a region to get back to its pre-crisis subscriber population size
Mobility data as an input to predictive models and analyses with other data
The indicators we propose can be used in further analyses, predictive modelling and research. The indicators reflect all dimensions of mobility and can be used to support decision-making and investigations across a wide range of domains. This may include epidemiological modelling, resource planning, provisioning of services, and longer term research into preparedness for epidemics or the effect of mobility restrictions on the environment.
Example of applications
Prediction of the spatial spread of SARS-COV-2
(assuming good case data)
Examples of insights provided:
Which areas of the country will be most affected? In which order may this happen?
When and where to relax interventions and which ones?
Indicators required
Inter-regional travel
Population mixing
Changes to resident subscriber population of each region
Resource planning and provision of services
(assuming data on resources and location of services)
Examples of insights provided:
How to optimise resource provision (e.g. health care needs) given population movements?
Indicators required
Population mixing
Changes to resident subscriber population of each region
Longer term research
e.g. preparedness to epidemics, economics, ecology, social sciences
Examples of insights provided:
How long did it take for mobility patterns to settle to a new normal after interventions? How disruptive has the crisis been and for how long (resilience)?
How are changes in economic variables (production, seasonal agricultural work, poverty) and environment (air pollution, noise pollution, electricity consumption) related to changes in mobility?
How changes in air pollution affect the transmission and severity of the disease?
Indicators required
Hourly and daily variance, per area
Changes to resident subscriber population of each region
Intra and inter-regional travel
Crowdedness