Published in: Health Econ, 2019 Jul; 28(7): 817–29.
By now many of you have used ride-sharing services. They are popular, convenient, and seemingly everywhere.
There has been some research evaluating how ride-sharing services impact a community. Previous studies have evaluated whether Uber was associated with reductions in drunk-driving fatalities (results were mixed) and how Uber complements other public transportation. There have even been reports that cities have investigated how Uber could be integrated into EMS systems.
Some have speculated that our patients may be opting to utilize these services rather than call 9-1-1 for transport to hospitals. However, until now there has not been any published research specifically addressing this question.
The authors of this month’s manuscript sought to estimate the impact of Uber’s entry into markets on EMS call volume. An interesting note about this study is that it was published in an economics journal. It is nice to see EMS data utilized in other disciplines!
The authors employed some interesting methods to acquire the data they used for this study. To identify the dates Uber entered markets, they scoured the Uber Newsroom blog for Uber launches throughout the country. They also searched external publications to identify any additional Uber launches that were not identified from the blog search. The study period was from 2012 to 2015. At the time data was collected, Uber was present in 43 states and Washington, D.C.
Once dates of Uber market entry were obtained, the authors then worked with the National EMS Information System (NEMSIS) to acquire EMS response volumes for the respective cities. We’ve discussed NEMSIS many times before; it is a great resource for national EMS data, and you can find more information at https://nemsis.org. EMS response rates were defined here as the response volume multiplied by 1,000 and divided by the city population.
Now, one of the more interesting parts of this study was who the authors worked with to obtain the NEMSIS data. The agreement NEMSIS has with the states that submit data to it does not allow NEMSIS to release information that identifies specific states, cities, or zip codes. Therefore, the authors actually had to submit the Uber market entry data to NEMSIS. NEMSIS then calculated EMS call volume and response rates based on the authors’ instructions and merged the data into a deidentifed dataset. Cities like New York and San Francisco were excluded because Uber entered these markets so early these data could be identified simply from the entry dates.
The authors built a regression model to estimate the EMS call volume after adjusting for other important variables such as geography and year.
Now let’s talk about what they found. There were 703 cities included in this analysis; however, because some crossed state lines, the authors split them into 723 city-state groups. As expected, throughout the study period there was a rapid increase in Uber availability. The average EMS response rate was 18 per 1,000 residents per quarter, with a minimum of less than one and a maximum of 752 per 1,000 residents.
The authors used this average EMS response rate to calculate the impact of Uber entry into a city. Specifically, they found that when Uber is available in a city, there are at least 1.2 fewer ambulance trips per 1,000 residents per quarter than before Uber entered the market. This roughly translates to a 6.7% drop in EMS call volume.
The authors then tried to determine whether the reduction in EMS response rate was in low-severity or high-severity calls. To do this they used documented lights-and-sirens transports vs. transports with no lights and sirens. Documented lights-and-sirens use during transport may not be the best proxy for patient severity, and this plays out in their results: The results from the analysis of calls that used lights and sirens did not reveal a statistically significant difference compared to the no-lights-and-sirens analysis. So the authors were not able to determine whether Uber availability reduced unnecessary EMS usage. They also did not see any statistically significant differences in response times, transport times, lights-and-sirens utilization, traffic accidents, alcohol-related traffic accidents, patient age, or patient sex based on Uber’s entry into a market.
As with all research, this study had some limitations. Most notably we have only discussed Uber, and there are other ride-sharing services. The authors focused on Uber because adding Lyft would not have affected the overall ride-share availability, since the study period was before Lyft was broadly available, Lyft typically enters cities after Uber, and previous literature focused primarily on Uber. Nonetheless, this is still a limitation.
It is fantastic to see EMS data published in an economics journal! The more disciplines that use our data, the more we can learn and improve our field. This was an interesting study that used national EMS data coupled with a unique data collection method for obtaining Uber market entry dates. There is also a section discussing robustness checks that was out of the scope of this review but is an interesting read into how the authors evaluated threats to their analytical assumptions, the impact of Uber rollouts, and the impact of NEMSIS expansion across the nation.
It is encouraging to see that even though NEMSIS must abide by agreements preventing the release of identifying information, it was willing and able to work with the authors to help them complete this study. This should be encouraging to anyone interested in completing research. NEMSIS is a valuable resource for data, and this manuscript shows it’s willing to provide assistance.
Antonio R. Fernandez, PhD, NRP, FAHA, is research director at the EMS Performance Improvement Center and an assistant professor in the Department of Emergency Medicine at the University of North Carolina–Chapel Hill. He is on the board of advisors of the Prehospital Care Research Forum at UCLA.