Driverless vehicles haven't yet taken to Canadian roads, but they've already rolled out in some other countries. Proponents say the technology will mean fewer accidents, while others have raised concerns about safety, liability and public acceptance, among other issues.
A new study by researchers at Sunnybrook Health Sciences Centre and the University of Toronto provides data on the potential of automated vehicles to improve road safety in various adoption scenarios. The work, published in Jama Surgery , is led by Armaan Malhotra, a U of T neurosurgery resident, and Avery Nathens, a chief of surgery at Sunnybrook and a professor of surgery in U of T's Temerty Faculty of Medicine.
Malhotra recently spoke with writer Erin Howe about the study findings.
What prompted you to look at this question around fully automated vehicles?
One of the most common things Professor Nathens, a trauma surgeon and the senior author on this paper, and I see frequently in our work is motor-vehicle collision-related injuries. This can include severe traumatic brain injury, spinal cord injuries, spine fractures, and chest and abdomen injuries. Often, they're life threatening or severely life-altering. And many of them are preventable.
Motor-vehicle collisions frequently involve human error. In high-velocity collisions on highways, there's frequently intoxication. One thing people who champion fully automated vehicles talk about is that, in a perfect world, self-driving vehicles could eliminate some of these problems. Theoretically, if the safeguards are good and the adoption is high enough, some say that autonomous vehicles are one of the largest possible interventions to reduce mortality and morbidity from traffic injuries.
Before we began this project, there hadn't been much study into this issue through a health-care lens.
Tell me about your findings.
To build our models, we used publicly available U.S. National Highway Traffic Safety Administration data from 2009 to 2023. It captured how many injuries there were each year, the number of miles traveled by vehicles and other metrics.
We also used public data from automated vehicle ride sharing company Waymo that shows an 80 per cent relative reduction in injuries when they compare fully automated vehicles to human drivers. And we tested a more conservative 50 per cent injury reduction in our modelling.
It's important to note that in forecasting like this, there's always potential error because of variables we can't perfectly account for - for example, how quickly these vehicles will be allowed on the roads or what proportion of vehicles will be fully automated.
So, we created multiple scenarios for these parameters and used our baseline estimates to predict how many actual injuries would be avoided. We wanted to provide data public health officials could use to have conversations about this technology.
In the most optimistic scenario, there were over one million injuries avoided over a 10-year period. But to reach that milestone, there would have to be an aggressive rollout of these vehicles, which would require a monumental effort.
Did anything surprise you?
One thing that surprised me was just how many injuries occur in the U.S. each year, and how sensitive the numbers are to the different adoption rates we tested.
There's tremendous potential for these tools to prevent injury and reduce the burden on health-care systems and public health across huge populations, but we will need appropriate regulation and oversight.
How excited are you for wider adoption of automated vehicles?
Right now, these vehicles are rolling out across the U.S. Some companies are expanding and exploring international cities, and may one day come to Toronto.
We have a healthy level of skepticism, and want to make sure further adoption of the technology is data-driven. We've noticed lot of people making statements that these vehicles will completely eliminate injuries or reduce fatalities. We just want to follow the data.
We'll also need to track the automated vehicle companies and be able to hold them accountable. And there will be a need to track whether fully automated vehicles were involved when crashes happen.
That said, we'd be very excited to see a major reduction or end to motor vehicle collisions.
What other questions might need to be answered in the future?
A lot of the data we used for our model came from automated vehicles driving in big cities. As companies scale toward service on long stretches of highway or to more rural areas, we'll need to extrapolate the data to places like Sudbury, Ont.
This modelling is preliminary and we'll need updated models as more data becomes available with automated vehicle rollouts. It will be important to have ongoing forecasting when updated safety data emerges.
