The hazards of drawing conclusions from before-and-after studies of system-level interventions: knowing may not be enough.