Do differences in reported expenditures between household scanner data and expenditure surveys matter in health policy research?