![[ Visit Client Website ]](images/banner.png)
BACKGROUND: Influenza, while often mild, can lead to hospitalization or death. Surveillance for influenza hospitalizations is important to evaluate the relative severity of disease resulting from the circulating viruses, to detect changes in the virus through virologic surveillance among hospitalized cases, and to detect changes in risk groups for severe disease. As part of New York’s multi-faceted approach to influenza surveillance, all hospitals with licensed medical/surgical beds electronically report weekly the aggregate number of hospitalized patients with laboratory-confirmed influenza by age group. However, accuracy and utility of this data has been questioned because of reporting delays and incompleteness of reporting. We evaluated the performance of the 2012-2013 influenza season Weekly Influenza Hospitalizations Survey, particularly completion rates and timeliness, and based upon the results, modified the survey for the 2013-2014 season.
METHODS: Each week during October-May, 197 hospitals were surveyed via a secure web application. During the 2012-2013 season, the survey was open each Wednesday through the following Tuesday for reporting hospitalizations that occurred during the previous week ending Saturday. Response data were analyzed in Excel.
RESULTS: During 2012-2013, 192 hospitals (97%) responded at least once during the season. Only 16% of hospitals responded in all 33 reporting weeks. On average, 78% of hospitals responded each week, but ranged from 88% during peak influenza activity to 61% in May when influenza activity was low. Among hospitals serving the 15-county Emerging Infections Program (EIP) FluSurvNet catchment area, numbers of hospitalized influenza patients reported via the survey were consistently and unexpectedly lower than numbers of patients reported directly to the EIP. Although 75% of responses were received during the first two days of each survey period, because of survey timing, data lagged one week behind other influenza surveillance systems.
CONCLUSIONS: Poor response rates suggest that influenza hospitalizations were underreported during 2012-2013. Although hospitals responded promptly, the lag in the survey itself limited the usefulness of the data for monitoring severe influenza trends in real time. Improvements during 2013-2014 included advancing the survey period by one week (responses are due on Wednesdays, four days after each week ending Saturday) and sending weekly email reminders to non-responding hospitals. During the first 13 weeks of 2013-2014, the average weekly reporting rate was 93% (range 90-95%) and 66% of hospitals reported every week. Improved response rates and timeliness during 2013-2014 to date are likely attributable to greater oversight and follow-up with non-responding hospitals.