Because EMNet findERnow gives exact directions to facilities in NEDI-USA, we aimed to ensure that coordinates assigned to every facility in NEDI-USA were no more than 1/8th of a mile (approximately one block) from the ER’s actual location.
To achieve this ambitious objective, we embarked on a research-quality “data cleaning.” First, we compared 2 national hospital databases, resolving thousands of identified discrepancies between the addresses listed in each database through web searches and phone calls to facilities. Second, because only one of those hospital databases contained geographic coordinates in addition to addresses, we sent our edited list of ER facility addresses to a geocoding company, which generated geographic coordinates for each. We then compared this generated list with the coordinates in the national hospital database. All facilities whose coordinates differed by more than 1/8th of a mile between the two sources were marked for further investigation. Third, we used Google Maps to resolve these discrepancies and generate correct geocodes because EMNet findERnow utilizes Google Maps. Facility addresses were entered into Google Maps and coordinates were generated using manual geocoding tools available on the Google Maps website.
For the purposes of this project, a geocode was considered to be accurate if the coordinates matched either: 1) the placement of the corresponding hospital symbol or facility outline on Google maps, 2) the image of the facility on Google Maps, or 3) the directions to the facility as reported on the its website. Finally, all coordinates obtained using Google Maps were entered back into the program to ensure they went to the correct facility address. This process yielded the database underlying EMNet findERnow. We are extremely confident that it is more accurate than any other database available today.
Through this process, EMNet staff discovered many factors which complicate the accurate generation of geocodes. Below, we list some of the most common:
In comparing our two geocoding sources, approximately 1300 ERs differed in coordinates by 1/8 of a mile or more, despite listing the same address. In some cases, both databases gave similarly incorrect data, making it hard to identify coordinates as incorrect.
Even major mapping resources do not provide full, accurate data on some ER locations In particular, very rural areas are not always mapped to the degree necessary for us to ensure absolute data accuracy. In other instances, mapping resources may identify an ER facility, but the location is incorrect. Such problems affected dozens of ERs in NEDI 2009.
Every year, many ERs open. For example, approximately 100 ER facilities opened between January 2008 and December 2010. Each year, EMNet obtains information from several national hospital databases on hospital openings and closures. To keep up-to-date on the newest ERs, EMNet researchers also track hospital openings between database updates through internet searches and phone calls. However, new facilities are not always well-marked in mapping programs. This problem is especially severe for “freestanding ERs” (i.e., ERs located outside a hospital).
Every year, many ERs close. For example, approximately 60 facilities with ERs closed between January 2008 and December 2010. Each year, EMNet obtains information from several national hospital databases on hospital openings and closures. To keep up-to-date on the most recent ER closures, EMNet researchers also track facility closures between database updates through internet searches and phone calls. However, facility closures do not always appear in Google Maps for some time. Similar problems appear in phone apps that utilize Google or other general internet programs, rather than an academic dataset like NEDI-USA to identify ERs.
Facilities change names for a variety of reasons (e.g., merger, change in ownership). For example, approximately 130 existing ERs changed names between December 2007 and December 2010. Even if the facility’s services or location does not change at the same time, it is important to update this data to avoid confusion – especially when trying to locate an ER. Unfortunately, map programs sometimes still display old hospital names or the names of smaller facilities or services within a hospital.
There are many types of ERs. Searching for an “emergency room” on the web or through a phone application other than findERnow can yield unhelpful information. Common difficulties with these alternatives to findERnow are the identification of animal hospitals, which also have emergency rooms, or duplicate entries for the same medical facility, which can be confusing and misleading.
To summarize, regularly-updated, research-quality data yields the unique NEDI-USA database that underlies EMNet findERnow. As a result, we are confident that EMNet findERnow is the most accurate tool available for the identification of U.S. ERs.