Effects of alternative follow-up procedures on mail survey response rates.