Cell Phone Data Helps Track Urban Infectious Disease Outbreaks

More than half of Earth’s population currently live in cities, with continued urbanization expected in Africa and Asia over the course of the next century. Circumstances presented by the urban environment increase risk for the spread of epidemics, particularly in rapidly growing cities with poor housing conditions.

In a study, published in Scientific Reports, investigators compared simulation models to actual case reports of 2013 and 2014 dengue outbreaks in Singapore. The study team found that having access to mobile phone location data can provide helpful information on an outbreak’s spread. Investigators emphasized that human mobility is a very important factor in understanding the spread of vector-borne diseases.

Previous research has been focused on vector-borne illness outbreaks on the national or regional level, but the Scientific Reports article focused on intra-city human mobility and dengue spread.

Investigators used 4 different mobility models based in different datasets: census records, random mobility, theoretical modeling, and mobile phone location data. Mobile phone data was anonymized after sourcing from a Singaporean mobile operator. Text, call, and other activity records were used to pinpoint users’ home and work addresses.

Each mobility model focused on 2 locations—a work and home location, which agents studied were assumed to commute between daily. The models differed in how assignment was made.

The pure mobile phone data model used anonymized call detail records to estimate home and work locations for 2.3 million agents. The random work location model still used home locations from the mobile phone data but assigned work location randomly. The Levy-distribution model assigned each agent a random home location based on the mobile phone data with a work location assigned so that commuting distance followed a truncated Levy-distribution. The radiation model used census data to distribute home locations, and work locations were chosen according to a radiation model derived from past research on mobility and migration patterns.

Flows of people correlated highly among the mobile phone data and the radiation model (r=0.938). Flows of people were somewhat less correlated with the Levy-distribution model (r=0.901), and were significantly less correlated among mobile phone data and random mobility (r=0.304). On these grounds, the study authors concluded that the radiation, Levy-distribution, and random mobility models were successively worse approximations of real urban mobility.

Study authors emphasized that greater preparation was necessary for the 80% of the world’s population at risk from at least 1 vector-borne disease. While most at-risk individuals live in poverty in tropical and subtropical regions, the case of Singapore demonstrates that highly developed cities still need continued efforts to halt outbreaks.

The study authors explained that mobile phone data can give real-time information on urban mobility that can be synthesized with infectious disease surveillance data and seasonal environmental data to map changing patterns of vulnerability in cities.

“Phone location data is better than annual census records,” Emanuele Massaro, PhD, lead author said in a press release. “The problem is that the data is owned by private companies.”

Massaro further pointed out that reducing the fragmentation of this data could be beneficial for both scientific research and wider public health.