Evolving to ILINet 2.0

BACKGROUND:  A Neolithic transformation is underway in public health, where the ubiquity of digital healthcare (HC) data is changing public health’s traditional role as data hunter-gatherers to one of data farmers harvesting huge reserves of electronic data.  ILINet 1.0 is an example of a 1980s hunter-gatherer approach to influenza syndromic surveillance involving the independent efforts of all state health departments.  Weaknesses of the ILINet 1.0 model include duplicated costs of provider recruitment and data management, low practice coverage, duplicated efforts in provider practices, inconsistent weekly provider compliance, slow data turn-around, lack of publicly available MSA-level ILI data, and lack of  forecasting capability during the current Epi-week.  We demonstrate how these weaknesses can be overcome using electronic healthcare reimbursement claims (eHRCs), which are widely accepted standard business practice records throughout the HC industry. eHRCs are available in centralized “data farms” that harvest them daily from office-based providers and retail pharmacies throughout the U.S.

METHODS:  We obtained Georgia ILI-related eHRC data for the 2006-2010 influenza seasons from IMS Health (IMS), a global HC information company.  We obtained comparable ILINet 1.0 data for the same period from the GA Department of PH.  The eHRC sample included 85+ million patient visits to providers’ offices and 213+ million prescriptions dispensed at retail pharmacies.  We derived ILI case definitions from ICD-9 and standardized drug codes and analyzed state- and MSA-level data by provider specialty and patient age.

RESULTS:  eHRCs that reached  the IMS data warehouse by the close of an Epi-week were sufficient to generate accurate, reliable, and descriptive ILI signals (% ILI and % antivirals) at state and MSA levels compared to only state-level  % ILI  for GA ILINet 1.0.  eHRC data are turned around within 1-2 days after the end of an Epi-week vs. 6 days for GA ILINet 1.0.  % ILI and % antivirals signals are equivalent for ILI surveillance purposes, suggesting that % antivirals signals are a reliable proxy for ILI.  MSA-level eHRC analyses demonstrate that the Fall 2009 wave of influenza A H1N1 in GA was comprised of multiple MSA sub-epidemics that began concurrently with asynchronous school openings.  The immediacy between drug dispensing and eHRC generation in pharmacies allows reliable ILI forecasting from eHRCs at the beginning of each Epi-week.

CONCLUSIONS:  Big eHealth data can be harvested immediately to begin the evolution towards ILINet 2.0 and faster and more granular ILI surveillance for the U.S. public health and national security communities.