Outpatient influenza surveillance: bridging the gap between clinical medicine and public health

Influenza causes considerable morbidity and mortality each year. Surveillance efforts in the United States are helpful for describing the influenza season (October through mid-May) but, due to reporting lags, are limited in the ability to provide early notification of the epidemic onset in a community. Here, we summarize the health impact of influenza and describe how influenza-like illness surveillance is currently conducted in outpatient clinics that contribute data to the Sentinel Providers Network in Utah. We compared data collection methods and found that clinics using automated methods had more complete data reporting with substantially lower personnel effort compared with clinics using traditional methods. Next, we describe the development of an outpatient surveillance system that applied statistical process control to the percentage of daily visits with a positive rapid influenza test. This system was developed using 4 years of historical data stored in discrete, searchable fields contained in the electronic medical records of outpatients visiting 10 clinics in the Utah Health Research Network (a clinic system contributing data to the Utah Sentinel Providers Network). Relative to traditional surveillance of influenza-like illness, rapid test positivity generated earlier signals of clinical influenza activity and had fewer false positive signals. Finally, we implemented our surveillance system prospectively to assess its ability to generate early alerts in real time. The real-time implementation generated a signal 19 days prior to the epidemic onset, which was similar to the median of 16 days estimated using historical data. We describe the impact on clinician behaviors related to influenza prevention that was associated with early notification generated by this novel surveillance system, and present lessons learned from the first year of real-time implementation. In summary, our surveillance system generated a signal 2??3 weeks before the epidemic onset in analysis of both historical data and real-time implementation. This compared favorably with reports by other authors of early signals generated using alternatives to traditional surveillance of outpatient influenza-like illness. The availability of rapid test results in the electronic medical record and simplicity of statistical process control methodology make this surveillance system relatively easy to implement in a clinical setting.