The best way to know the effectiveness of a drug or helmet is to get a representative sample of the population, and then randomly assign them to the test group or the control group. Such an experiment allows one to reliably estimate effectiveness subject to a statistical margin of error. But we don’t know who will be involved in a crash, and if we did, it would be unethical to tell them whether to wear a helmet. Instead, researchers collect data after the fact.
In 1989, Thompson et al. obtained data from Seattle hospitals for two groups of cyclists who went to the hospital after a crash. Only 7% of the first group wore a helmet, and they all had head injuries. But 24% of the second group wore helmets, and none of them had head injuries. Assuming that both groups were the same except for the type of injury they experienced, these results imply that helmets reduced head injuries by 75%. Thompson et al. realized that the two groups were different, ran regression analysis on the data to isolate the effects of helmets and found that helmets were even more effective: 85%.
That study led the researchers to start saying two things that have almost become mantras among many public safety advocates: “Helmets reduce head injuries by 85%” and “The most important thing you can do to be safe on a bike is wear a helmet.”
In the last 24 years, similar studies have found that helmets reduce head injuries, but to a less extent than in the Seattle Study. A comprehensive synthesis of all studies in 2001 estimated 53-63% effectiveness, but because helmets increase neck injuries, the net effectiveness is 41--0%. Studies in the last decade have estimated that helmets only prevent 20-40% of potential head injuries, so the most recent synthesis of all studies ever published finds the helmets reduce head injuries by 30--50% and total injuries by 10-20%, when you include the increased neck injuries. But we still hear the refrain “helmets reduce injuries up to 85%!”