Response Rates and Nonresponse Bias Among 20 Years of Pediatrician Surveys

Our objectives were to examine the following: physician survey response rates across a 20-year period; the impact of a token incentive on response rates; whether survey nonresponse bias is present and if it is associated with response rate; and the impact of a token incentive on nonresponse bias.

We utilized data from 68 American Academy of Pediatrics (AAP) pediatrician surveys from 2000 to 2019 and an AAP administrative database, which included information for both respondents and non-respondents (target sample). Linear regression examined response rates over time. To assess nonresponse bias, a t-test or Wilcoxon rank test compared respondents and the target sample by age, gender, and US Census region. Linear regression or Spearman correlation examined the association of response rate and nonresponse bias. Interrupted time series analyses tested the introduction of a $2 token incentive on both response rates and nonresponse bias.

Overall mean survey response rate was 56.2%. Response rates declined across survey years (β = -0.58, P < 0.001). The $2 incentive generated an 8.7% response rate increase (P < 0.001). The respondent groups had more female pediatricians than the target samples (62.5% vs 60.1%, P < 0.001). Age nonresponse bias was associated with lower response rates (β = -0.47, P < 0.001). The $2 incentive was associated with nonresponse bias shifts toward older, away from female, and away from Northeastern respondents.

Our study demonstrates an overall decline in pediatrician survey response rates and a meaningful impact of a token incentive on response rates and nonresponse bias, underscoring the importance of measuring nonresponse bias whenever possible.