The full version of the lawnmower study that I blogged about last week is available online here. The full study includes much greater detail than the media reports – but also the same problems that I noted last week. Specifically:
1. The study says, “During the 15-year period, the annual number of lawn mower–related injuries to children remained relatively consistent.” No statistical test for this claim is mentioned, but the following graph is provided:
This actually looks like a mild decline to me. Moreover, the raw numbers don’t take rising population into account. Using population figures from WISQARS, plus my eyeballed estimates from the graph above (the study doesn’t provide the precise numbers), I generated lawnmower injuries per 100,000 in the 0-20 age group. (The rates did not exactly match the study’s for the two years in which they were given; I’m guessing they used different population estimates.)
Here, the decline looks even more pronounced. And indeed, every statistical test I performed indicated a statistically significant decrease in the rate of lawnmower injuries over the period in question, always with a confidence level of 95% or better.
2. Like the media reports, the study itself concludes, “The relative consistency of the number of lawn mower–related injuries to children during the 15-year study period is evidence that current prevention strategies are inadequate.” As I said before, this is a complete non sequitur. Even if the number or rate of injuries were constant over time (which does not seem to be the case), that wouldn’t mean it was unacceptably high. The rate could be consistently high, or consistently low, or consistently moderate. To know which, we would need some reasonable (i.e., non-zero) standard for comparison.
The study also says, “Injuries that are related to lawn mowers are an important cause of pediatric morbidity,” but it offers no comparisons with other sources of injury to support this claim. Again using data from the WISQARS site, I found that lawnmower injuries constituted less than one-tenth of one percent of all non-fatal injuries for the designated age group, for every year from 2000 to 2004 (the only years for which I had complete data). Is one-tenth of one percent worth making a big deal about?