Random digit dialing: the potential effect on sample characteristics of the conversion of nonresidential telephone numbers.

This study examined the changes over time in the residential status of telephone numbers and the characteristics of households with telephone numbers that changed from nonresidential to residential. The authors determined the status in 1987 of all phone numbers (n = 9,107) that had been found to be nonresidential in six case-control studies conducted between 1979 and 1986 in Washington State. A telephone interview to obtain information on household characteristics was completed for 1,333 of the 1,901 phone numbers that had become residential by 1987. The interviews revealed that households with phone numbers that had previously been nonresidential differed from the general population with respect to household income, the age of household members, and the education of the head of the household. Data for the same area showed that the proportion of all phone numbers that were residential changed differentially according to the prefix (which roughly defines geographic area) during a 12-month period between 1987 and 1988, with some prefixes showing increases in residential phone numbers of more than 20%, while other prefixes decreased or stayed the same over the same period. We conclude that methods of random digit dialing that use two-stage designs or exclude previously dialed numbers may result in biased sampling unless changes over time are considered.