In order to show that helmets *cause* accidents, they need to create a randomized study where they force employees to flip a coin to decide whether to wear a helmet or not.
Otherwise, here's one plausible scenario: Employees who work in tiny suburbs with small roads and very little traffic feel safer, and this are less likely to wear a helmet. They get in less crashes because their town has fewer and safer drivers. Employees working in the city have more crashes simply because of being in a busy city, so they are more likely to wear helmets.
Right now, you cannot prove that helmets cause crashes and not the above.