Posted by Nolan Miller on Sep 14, 2011
Filed Under (Health Care)
The New York Times ran a story the other day on a recent study in the journal Pediatrics that “suggests that children may be safer when their grandparents are driving than when their parents are at the wheel.” Both the article and the study are flawed. The study is bad for statistical reasons I’ll get into in a minute, and perhaps it is better to say that the popular reading of the article makes too much of what the results actually show. The New York Times story is flawed because there is no evidence that any thought was given to the possible merits of the study, merely summarizing without any attempt to do actual journalism. This is all-too-often true of the coverage of health-related studies in the popular press. As most people are unlikely to read the original study or critique its methods, reporting results without providing critical analysis runs the risk of giving people wrong, and potentially dangerous, advice.
So, what results are reported in the study? The study finds that grandparents were the drivers in 9.5 percent of vehicle accidents involving children, but only 6.5 percent of the injuries. Thus, children in crashes in which their grandparents were the drivers were less likely to be injured than children in crashes in which their parents were. The result remains true even once you control for known risk factors for injuries such as use of proper restraints, the size of the vehicles involved and measures of severity of the crash. In fact, it gets a bit stronger, so that children in grandparent-driven crashes are 50% as likely to be injured as children in parent-driven crashes. The Times quotes one of the study’s authors as saying “But there’s something about the crashes of grandparents that we were unable to measure that was protective. It would be great to figure out what this is, because it could be protecting a lot of kids.”
OK. So what’s the problem? Well, the study fails to account for the fact that far more miles are driven by parents than by grandparents. So, while it may be true that a child in an accident where a grandparent is the driver are less likely to be injured than a child in an accident where the parent is the driver, grandparents seem to have so many more accidents per mile driven that a child is far more likely to be injured in a mile driven by a grandparent than a parent.
Let’s do some back-of-the-envelope calculations. Some facts take from the paper, which are not ideal, but the best I could find on short notice. “Up to 33% of nonresidential drivers will make trips with children a few days per month, 14% a few days per week, and 5% almost every day. Of such nonresidential adult drivers, 42% were the child’s grandparents.” A quick look at the original source suggests that 4% of grandparents report driving children “almost every day” and 14% report driving kids “a few days a week.” Now, I think it is reasonable to assume that a child drives with a parent every day (especially since the data in the study are limited to kids who are driven in cars and have accidents). It is a bit of a leap, but if 4% of grandparents drive kids almost every day and we round up to account for more occasional grandparent drivers, then maybe the number of trips kids take where their parents drive is something like 20 times as large as the number of trips which grandparents drive (since 100% of parents drive their kids every day). If the trips are of equal length, then that suggests 20 parent-driven miles for ever grandparent-driven mile. My guess is that parents tend to make longer trips, so if anything it is more parent miles per grandparent mile, but lets go with 20:1.
Now, the study found 1143 accidents involving grandparent drivers and 10716 accidents involving parents. But, to account for the fact that parents drive more often, you should weight these numbers, basically by multiplying the 1143 by 20 to get a number that corresponds to what would happen if grandparents drove as many miles as parents. If we do that, we get 22860, meaning that a child is twice as likely to have an accident in a mile driven by a grandparent than in a mile driven by a parent. Going on, the study found 113 injuries when grandparents were the drivers and 1189 injuries when parents drove. Again multiplying the grandparent number by 20, you get 2260, meaning that children are also twice as likely to be injured in a mile driven by a grandparent than by a parent.
Given the data, in order to find that the risk of accident and injury to a child in a mile driven by a parent and grandparent were approximately the same, you’d need the child to be driven by a parent for no more than 10 times as many miles as by a grandparent. This seems unlikely to me.
So, my reading of the data is almost exactly opposite those of the paper. Now, to be fair, the paper usually states its conclusion as ‘if you are going to have an accident, it is better to have one with a grandparent behind the wheel.’ This is accurate, although I believe the authors are overstepping the data when they go on to suggest there is something protective about grandparent driving. (In fact, the protective thing is that it doesn’t occur all that often!) The more relevant question is ‘should I send my child to school with a parent driver or a grandparent driver?’ The study does not directly investigate this question (and the New York Times does not investigate anything in its write-up). My suspicion and back of the envelope calculations suggest that the answer to this question is clearly in favor of the parents.
Note: the views expressed in this blog should not be interpreted as an indictment of any particular grandparents, especially my son’s grandparents, who are welcome to drive him around whenever they like.