Effects of repeated callbacks on response rate and nonresponse bias: results from a 17-state pharmacy survey

Although surveys are the most common method of data collection in research in the social & behavioral pharmaceutical sciences, little evidence exists on pharmacists' and pharmacies' patterns of telephone survey participation and the impact of repeated callbacks on minimizing nonresponse bias.

To (1) describe final disposition of the survey sample, (2) describe the effect of repeated callbacks on response rate, and (3) examine whether hard-to-contact cases differ from those that are relatively easier to contact in terms of general pharmacy characteristics and characteristics specific to survey topics.

This study used a cross-sectional descriptive survey design that gathered information from key informants of 2558 community pharmacies in 17 states. Telephone interviews were conducted using a Computer Assisted Telephone Interview system. Final disposition of the sample units and the final response rate were reported. General pharmacy characteristics and characteristics specific to immunization delivery (ID) and emergency preparedness and response (EPR) activities were compared across different ease-of-contact levels.

A total of 1707 pharmacies (66.7%) completed the interview; 757 (29.6%) were eligible cases that were not interviewed; 8 (0.3%) and 86 (3.4%) were cases of unknown eligibility and ineligibility, respectively. The final response rate was 69.1%. The number of calls to pharmacies was most productive on the first 9 calls and leveled off after the 14th call. When comparing characteristics among different ease-of-contact levels, 2 out of 3 characteristics related to ID activities were different. No differences were found in general pharmacy characteristics and characteristics specific to EPR activities, however.

The greater number of callbacks resulted in higher response rates due to greater number of completed interviews. The increase in response rate did not rule out nonresponse bias, however. Hard-to-contact cases differed from relatively easier to contact cases in terms of characteristics related to the survey topic, but not general characteristics.