Telephone sampling bias in surveying disability.