Calculations for Estimating Emergency Department Coverage
In 2014, Coates et. al.1 reported that the National Syndromic Surveillance Program (NSSP) covered 45% of U.S. nonfederal hospital emergency department (ED) visits across all 50 states, including Washington D.C., based on survey data published by the American Hospital Association (AHA). Since that report, we have streamlined new-site onboarding; improved data collection, management, and reporting; and developed detailed site reports for improving data quality. Given these significant improvements, we revised our calculation for estimating ED visit coverage to use improved NSSP data.
Our revised approach allows for inclusion of all ED facilities that actively contribute to NSSP’s BioSense Platform, regardless of their participation in the AHA Annual Survey. We use data from facilities whose primary facility type category is “Emergency Care,” and we define “active” participants for this estimation as facilities that have reported data within the past 90 days.
First, we calculate annualized ED visits for all facilities that actively submit data. To account for the changing number of facilities actively submitting data (mostly increasing) and for occasional interruptions and fluctuations in data submission, we compute annualized ED visits by calculating the median number of ED visits for each facility during the previous 12 months and multiplying it by 12. This approach minimizes weekly and monthly variability and provides a relatively stable estimate of annualized number of visits for each facility based on data actually submitted to the NSSP BioSense Platform. We use the annualized number of ED visits as the numerator for estimating ED visit coverage of active facilities.
The total number of annual visits for each facility is not readily available for use as a denominator. For facilities that participated in the 2016 AHA survey (the most recent available), we compared the annualized visit count based on NSSP data (described above) with annual number of visits reported in the AHA survey. The NSSP data showed higher visit counts for many facilities, which may be attributed to multiple factors. We use the larger of the two estimates (AHA or NSSP) for the denominator for these facilities. This approach underestimates our results for some facilities but prevents reporting coverage greater than 100% for other facilities. For facilities that did not participate in the AHA survey, we used the annualized NSSP ED visits for the facility as the denominator, being the only available source of data, although it may overestimate results for some facilities.
To estimate overall ED-visit coverage, we divide the total of the annualized NSSP ED visits by the total estimated ED visits (if both AHA and NSSP data were available for a facility, the higher value was used; if data were available from only one source, that value was used).
Highlights of our revised approach are summarized below:
- Primary facility is registered as Emergency Care.
- Facility reported visit data within past 90 days.
- Numerator is based on NSSP data.
- Median monthly number of visits are calculated by using previous 12 months of NSSP data and are multiplied by 12 to get a yearly estimate of visits. (Note: Months with no data are ignored.)
- Numerator is calculated for all active ED facilities.
- If the estimated number of annual visits for a facility is available from both NSSP data and AHA data, the larger of the two values is used in the denominator.
- If the estimated number of annual visits for a facility is available from only one dataset (NSSP or AHA), the available data are used in the denominator.
We believe our revised approach provides the best estimate available at this time. The estimate emphasizes use of data actually submitted—rather than self-reported survey data. Further, the estimate uses current-year NSSP data, which are timelier.
Limitations include the use of multiple sources (NSSP data and AHA Annual Survey data), which may define the data elements differently. There is some lag in availability of AHA data, as the most recent data available are for 2016. Further, there may be facilities not included in either AHA data or NSSP data, and the inability to include these visits in the denominator may affect coverage estimates. Unfortunately, we do not know how many facilities or visits could be missing. As we gather larger quantities of high-quality NSSP data, we will revise our approach to further improve the accuracy of this estimate.