The major goals of this project were to assess: (1) How community and clinically generated data compare regarding disease incidence, contributing population demographics, and spatio­temporal coverage; and (2) The value of community infection data for understanding, monitoring and forecasting seasonal and pandemic influenza dynamics.

Major activities in the first year included development of the means for data collection from the community and completing data collection and laboratory testing.  The second year focused on examining the collected data and developing analyses to identify how community infection data can contribute to understanding disease incidence and patterns. To accomplish the second goal, we also reached out to other groups performing influenza research and obtained data from their studies to combine with ours and perform a more comprehensive analysis.

At Columbia, our primary role was to process community samples and perform laboratory testing.  Over the course of the project, we received 314 self-collected nasopharyngeal swab specimens.  All specimens were analyzed using the GenMark respiratory virus panel (RVP) assay for 18 common respiratory viruses.  Of the 314 samples, 84 were positive for 1 or more respiratory virus, including single infections from adenovirus C (n=3), coronavirus 229E (9), coronavirus OC43 (9), coronavirus NL63 (2), rhinovirus (41), human metapneumovirus (1), influenza A/H3N2 (7), influenza B (1), influenza A/H1N1 (1), parainfluenza 1 (1), parainfluenza 3 (1), parainfluenza 4 (2), respiratory syncytial virus (RSV) A (1), and RSV B (3).  There were 2 coinfections: RSV B and coronavirus 229E; RSV B and adenovirus C.

The Columbia team is additionally supporting the analyses by the NYU team on 2 fronts.  Firstly, we are participating in and providing feedback to the ongoing analyses conducted with the aim of demonstrating that different and complementary data provide a more comprehensive picture of global disease incidence than a single data form in isolation.  The intent is to use different observations along the morbidity surveillance pyramid to understand how loss of information accumulates during surveillance.  For example, the difference between the number of reported symptomatic infections and those that actually seek health care represents ‘under-healthcare-reporting’, whereas the proportion of infections that are not symptomatic represent the ‘under-ascertained’.  Analysis of these issues directly addresses the primary aims of this project. 

Secondly, we have provided results from an alternate active surveillance project carried out separately with DoD funding by Dr. Shaman’s team.  Specifically, we have provided the results from nasopharyngeal swab samples taken during April-June 2016 and January-April 2017 from adult visitors at a major New York City tourist institution. This sampling generated specimens from 2685 participants; all samples were assayed using the same GenMark RVP protocol (185 were virus positive).

Initial findings, led by NYU, have been published at the social media and health workshop at ICWSM 2017, and in a paper in BMC research notes; these papers have explored the predictive power of community (social media) data for infection, as well as comparison of community and clinical respiratory infection data in a unique setting (Nigeria). We also have two larger papers in preparation that explore each of the objectives, which we anticipate will be of interest to both the computational and public health communities.

Dr. Chunara, PI at NYU, has requested an NCE for the NYU portion of this collaborative project.  Though not requesting an NCE, Dr. Shaman’s group at Columbia University plans to continue participating in these analyses and supporting the development and submission of manuscripts. 

Last Modified: 11/29/2018
Modified by: Jeffrey L Shaman