Table 1.

Summary of types, metrics, and proposed applications of mobile phone data.

Data type/information	Metrics	Applications	Advantages	Limitations
Call data records (CDR):
• Collected routinely by mobile phone operators • Consists of a time stamp, GPS location of local cell tower, and unique identifier for all subscribers	• Origin-Destination Matrix • Radius of Gyration • Subscriber Density	• Assess changes to population-level mobility and clustering behaviors • Understand risk of importation from different regions • Retrace likely introduction and spread of an outbreak in new areas • Inform projections of disease risk or burden across space	• Typically readily available • High coverage to estimate large, population-level mobility patterns for entire countries or region • Available frameworks provide aggregated, anonymized metrics	• Assumes aggregate mobility behaviors represent that of infected/potentially infectious individuals • Cannot distinguish high vs low risk of transmission • Limited data in Internet-enabled or low cell-tower-density areas • Limited use in understanding transmission chains • Selection bias for whom data is available (mobile phone user)
GPS location data:
• Collected passively through some smartphone applications • Consists of time stamp, GPS location of phone, and unique identifiers for all application users	• Origin-Destination Matrix • Radius of Gyration • User Density, Proximity	• Assess changes to population-level mobility and clustering behaviors • Understand risk of importation from different regions • Retrace likely disease introduction and spread in new areas • Inform projections of disease risk or burden across space	• Provides higher resolution spatial data than CDRs • Provides population-level insight into the average clustering and movement of individuals	• Selection bias in the population for whom data is available (smartphone users who opted into app) • Fewer standardized frameworks for managing privacy and anonymization of potentially sensitive information
Bluetooth data:
• Collected passively by Bluetooth-enabled phones • Consists of the time stamp, distance, and duration of interaction between two devices with unique identifiers	• User Density, Proximity • Proximity Network Characteristics (degree, clustering)	• Assess changes to population-level clustering behaviors due to NPIs • Assess changes to pairwise contact rates in a given population over time	• Large-scale collection of data on pairwise interactions and clustering • Interactions potentially more relevant to disease transmission	• Selection bias in the population for whom data is available (mobile phone user, Bluetooth enabled, interacting with another Bluetooth enabled device) • Cannot distinguish proximity with high vs low risk of transmission
Opt-in application data:
• Applications using Bluetooth and/or GPS location data to track interactions between individuals collect data passively through enabled phones and/or actively when users respond to prompts • Application specific, but could consist of time stamp, distance, duration of interaction, questionnaire responses	• Proximity Network (identified contact chains)	• Contact tracing to facilitate quarantine of potentially infected persons	• Enable rapid tracing and quarantining of exposed individuals with fewer resources • Allow for measured behavior to be linked to an individual’s infection status	• Low tolerance for missing data; unclear ability to sufficiently scale up • Cannot distinguish proximity with high vs low risk of transmission • Selection bias in the population for whom data is available (smartphone users, possibly Bluetooth enabled, opted into and compliant with application, interacting with another user opted into and compliant with application)