Uber vs Drunk Driving: One City’s Experience

Emergency department visits for motor vehicle-related trauma and convictions for driving under the influence in Houston declined significantly after the ridesharing service Uber debuted there, researchers found.

Compared with the 2007 calendar year, trauma center activations for vehicle accidents in the 12 months ending in November 2019 were 20.1% lower (1,527 vs 1,911), according to Christopher R. Conner, MD, PhD, of the University of Texas Health Science Center in Houston, and colleagues.

For the entire two periods before and after Uber’s December 2013 rollout, the rates of motor vehicle trauma visits during peak hours fell 23.8% (0.21 vs 0.26 per hour), the researchers reported in JAMA Surgery. Injuries became less severe as well, though mortality rates did not change.

Court records for persons convicted of driving under the influence or driving while intoxicated showed a similar pattern. From a mean of 22.5 convictions per day (SD 10.9) during 2007-2013, the rate declined in the post-Uber period to 19.0 per day (SD 10.3) through December 2018, with the biggest percentage decreases — 15% to 18% — on Fridays and Saturdays, and especially at night.

Other data also pointed to ridesharing as a causal factor. Looking solely at data on Uber rides (the researchers only had indirect data on Uber’s chief competitor, Lyft), the team found a dramatic reduction in motor vehicle trauma related to rideshare volume (incident rate ratio 0.33, 95% CI 0.17-0.67).

As well, trauma victims’ mean age rose significantly after Uber started service (39.4 vs 37.2, P<0.001), as would be expected from the relatively younger demographic for which ridesharing is popular. Traumas involving people younger than 30 declined nearly 40%.

And most convictions in the pre-Uber period involved arrests in central Houston, whereas convictions there post-Uber were predominantly connected to arrests in the city’s outer areas. Conner and colleagues explained that Uber volume was greatest in the central core, and the decline in convictions there occurred even the population there increased over time.

“We conclude that the decrease in [motor vehicle crash] traumas during both peak alcohol consumption hours and after introducing ridesharing services suggests individuals choose the ridesharing as a safe alternative to impaired driving,” the researchers wrote.

Previous research on the topic yielded mixed results, Conner’s group noted. A few studies had actually suggested an increase in motor vehicle crashes associated with ridesharing. But data for this earlier work was mostly at the level of months, “lacking the detail to control daily fluctuations in rideshare use, diurnal traffic patterns, and frequency of convictions for impaired driving,” Connor and colleagues wrote.

An accompanying commentary by two Pakistani researchers applauded the study, which they said would help in evaluating the potential for new products and services to improve public health without being designed for that purpose.

“Ultimately, it is not only the public health interventions deployed by policy makers and law officials, but also products developed in the free market in response to consumer needs, that become most effective,” wrote Adil Haider, MD, MPH, and Hamna Shahbaz, MD, both of Aga Khan University in Karachi. “This notion reverberates beyond drunk driving, in that ridesharing services also have the immense potential for reducing traffic fatalities associated with sleep deprivation and medical conditions while increasing mobility for people with physical and cognitive limitations.”

But the study by Conner’s group was not without its own limitations, the pair cautioned — chiefly that it only included trauma data from level 1 centers and that it didn’t separate injuries to passengers versus drivers.

Another was that the Houston researchers couldn’t obtain data directly from Lyft, relying instead on secondary data provided by Google on Lyft usage, with a temporal resolution of 1 month. Uber, on the other hand, cooperated with Conner’s group, providing detailed data on more than 24 million rides in the Houston area.

The study also included records on 23,491 motor vehicle trauma evaluations at Houston’s two level 1 trauma centers from January 2007 to November 2019, and court records on 93,742 convictions (including guilty pleas and sentences to probation) for drunk driving offenses through December 2018 (Conner and colleagues considered records for 2019 potentially incomplete.)