Measuring alcohol consumption--should the 'graduated frequency' approach become the norm in survey research?