Cell-phone traces reveal infection-associated behavioral change

Ymir Vigfusson; Thorgeir A. Karlsson; Derek Onken; Congzheng Song; Atli F. Einarsson; Nishant Kishore; Rebecca M. Mitchell; Ellen Brooks-Pollock; Gudrun Sigmundsdottir; Leon Danon

doi:10.1073/pnas.2005241118

Edited by Nils Chr. Stenseth, University of Oslo, Oslo, Norway, and approved December 16, 2020 (received for review March 19, 2020)

Significance

Infectious disease control critically depends on surveillance and predictive modeling of outbreaks. We argue that routine mobile-phone use can provide a source of infectious disease information via the measurements of behavioral changes in call-detail records (CDRs) collected for billing. In anonymous CDR metadata linked with individual health information from the A(H1N1)pdm09 outbreak in Iceland, we observe that people moved significantly less and placed fewer, but longer, calls in the few days around diagnosis than normal. These results suggest that disease-transmission models should explicitly consider behavior changes during outbreaks and advance mobile-phone traces as a potential universal data source for such efforts.

Abstract

Epidemic preparedness depends on our ability to predict the trajectory of an epidemic and the human behavior that drives spread in the event of an outbreak. Changes to behavior during an outbreak limit the reliability of syndromic surveillance using large-scale data sources, such as online social media or search behavior, which could otherwise supplement healthcare-based outbreak-prediction methods. Here, we measure behavior change reflected in mobile-phone call-detail records (CDRs), a source of passively collected real-time behavioral information, using an anonymously linked dataset of cell-phone users and their date of influenza-like illness diagnosis during the 2009 H1N1v pandemic. We demonstrate that mobile-phone use during illness differs measurably from routine behavior: Diagnosed individuals exhibit less movement than normal (1.1 to 1.4 fewer unique tower locations; P<3.2×10−3), on average, in the 2 to 4 d around diagnosis and place fewer calls (2.3 to 3.3 fewer calls; P<5.6×10−4) while spending longer on the phone (41- to 66-s average increase; P<4.6×10−10) than usual on the day following diagnosis. The results suggest that anonymously linked CDRs and health data may be sufficiently granular to augment epidemic surveillance efforts and that infectious disease-modeling efforts lacking explicit behavior-change mechanisms need to be revisited.

Footnotes

↵¹Y.V., T.K., and D.O. contributed equally to this work.
↵²To whom correspondence may be addressed. Email: ymir.vigfusson{at}emory.edu.

Author contributions: Y.V. and L.D. designed research; Y.V., T.A.K., D.O., C.S., N.K., R.M.M., G.S., and L.D. performed research; Y.V., T.A.K., D.O., and C.S. devised models; G.S. contributed data; Y.V., T.A.K., D.O., C.S., A.F.E., N.K., R.M.M., and L.D. analyzed data; and Y.V., T.A.K., D.O., A.F.E., E.B.-P., and L.D. wrote the paper.
The authors declare no competing interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2005241118/-/DCSupplemental.

Data Availability.

All study data are included in the article and/or SI Appendix. The code and documentation used in our analysis are available at https://github.com/SimBioSysLab/cdr-open-code.