Flying, Phones and Flu: An evaluation of Keflavik International Airport and its role in the introduction of pandemic H1N1 into Iceland in 2009 using anonymized call records Open Access

Kishore, Nishant (2016)

Permanent URL:


INTRODUCTION: We studied the introduction of pandemic H1N1 to Iceland; an isolated island with centralized national health records, near-ubiquitous mobile phone use, and one likely port of entry: Keflavik International Airport. Using anonymized call detail records linked with health records we evaluated the role that international travelers played in the introduction and propagation of pandemic H1N1 in Iceland by quantifying the association between airport exposure and influenza-like-illness (ILI) diagnosis.

METHODS: This was a nested case-control study comparing odds of exposure to Keflavik International Airport amongst cases and matched controls producing a longitudinal two-week matched odds ratios (mORs). We further evaluated rates of infection among 1st degree connections of cases compared to their matched controls.

RESULTS: The longitudinal two-week mOR produced for individuals who were exposed to the airport in the 4 days before ILI diagnosis showed an increase in the two-week mOR in the early stages of the epidemic from August 17th until August 31st with a mOR of 2.00 (95% CI: 1.4, 2.9). During the two week period from August 17th through August 31st we calculated the two-week IDR of infection among 1st degree connections to be 14.2 (95% CI: 5.7, 35). The IDR decreased steadily to a threshold IDR of approximately 5 during the epidemic peak.

CONCLUSIONS: We find that there is an association between exposure to Keflavik International Airport and incident ILI diagnoses during the initial stages of the epidemic. Our data show a definitive high rate of transmission earlier in the epidemic. However, even during generalized epidemic in the population, 1st degree connections of individuals diagnosed with an ILI get sick at a rate 5 times higher than the 1st degree connections of their matched healthy controls. Our methods were validated through evaluation of domestic airports as negative controls. Bias analyses showed minimal threats to the validity of our measures of association, assuming the validity of our bias model. Through greater collaboration with both mobile network operators and health officials, the techniques described in this study can be used for hypothesis-driven evaluations of locations and behaviors during an epidemic and their associations with health outcomes.

Table of Contents

Introduction. 1

Methods. 5

Study Population and Design. 5

Datasets - MNO Call Detail Records. 5

Datasets - CHS-CDC ILI Diagnoses. 6

Variables. 6

Data Cleaning. 8

Analysis Plan. 9

Identification of Cases of Interest 10

Positive Controls. 10

Negative Controls. 10

Social Network Analysis. 11

Quantitative Analysis of Systematic Error. 11

Results and Analysis. 14

Data. 14

Epidemic Curves. 14

Selection. 15

Evaluation of Exposure. 15

Sensitivity Analysis of continuous OR window.. 16

Positive Controls. 16

Negative Controls. 17

Social Network Analysis. 17

Quantitative Analysis of Systematic Error. 18

Conclusions. 19

Primary Findings. 19

Secondary Findings. 19

Negative Controls. 19

Quantitative Analysis of Systematic Error. 20

Locations and their Roles in an Epidemic. 20

Limitations. 21

Future Analyses. 22

References. 23

Tables. 27

Figures and Figure Legends. 32

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
Subfield / Discipline
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files