Pitfalls in the use of ‘accident’ victims as comparison groups