Abstract The National Immunization Survey-Flu (NIS-Flu) monitors influenza vaccination among children in the United States, and the accuracy of NIS-Flu estimates is important for evaluating influenza vaccination programs. Total survey error (TSE) provides a framework for assessing survey accuracy. NIS-Flu data from the 2015–16 through 2017–18 influenza seasons were examined to assess components of nonsampling error: noncoverage of households by the sampling frame; household nonresponse; and measurement error resulting from parental reporting of vaccination status. We estimated the distributional parameters of each source of error from related surveys and auxiliary data and employed simulation to estimate bias. We estimated bias in end-of-season estimates of influenza vaccination rates and quantified the extent to which each source of survey error contributed to total bias. Overall point estimates (and 95 percent interval estimates) of total estimated bias were 8.0 (1.3, 14.6), 8.9 (2.6, 15.2), and 6.1 (−0.5, 12.8) percentage points for the 2015–16, 2016–17, and 2017–18 influenza seasons, respectively. Measurement error resulting from parental recall of children’s vaccination status was the largest contributor to bias; the average estimates of measurement error were 7.0, 8.2, and 5.5 percentage points for the three seasons, respectively. Errors due to noncoverage and nonresponse were relatively small (averaging 1 percentage point). Estimates of total bias were larger for Hispanic and non-Hispanic Black children and for the youngest children aged 6–23 months. Findings suggest NIS-Flu influenza vaccination rate estimates may be biased upward, primarily due to measurement error in the form of overreporting influenza vaccinations by parental respondents, with relatively little error due to noncoverage or nonresponse.
Davis et al. (2026) studied this question.