• Patient safety
• Human error
• Checklists

There is no doubt about Dr Levitt's genuine concern for patient safety.1 His experience, like that of others, must indeed have led him to hospital staff he'd rather do without. One can understand the seduction of sanctioning non-compliant doctors2 or getting rid of the deficient practitioners—the system's bad apples—altogether,3 as also proposed by Levitt. In 1925, German and British psychologists were convinced they had cracked the safety problem in exactly this way. Their statistical analysis of five decades had led them to accident-prone workers; misfits whose personal characteristics predisposed them to making errors and having accidents.4 Their data told the same stories flagged by Levitt: if only a small percentage of people is responsible for a large percentage of accidents, then removing those bad apples will make the system drastically safer.

It didn't work. The reason was a major statistical flaw in the argument. For the accident-prone thesis (or bad apple theory) to work, the probability of error and accident must be equal across every worker or doctor. Of course it …