Consequences of Declining Survey Response Rates for Smoking Prevalence Estimates