Sick at work: methodological problems with research on workplace presenteeism