Thread by @rnishimura

Michelle Cavazos (RTI) talks about nonresponse bias in Address-Based Samples with an in-person nonresponse follow-up #AAPOR

RQ's: Impact of NRFU on respondents' demographics and on study outcomes.

Sample design: stratified sample by response propensity, oversampling strata with lower propensities.

Choice+ methodology: higher incentive for responded answering survey on web rather than on paper questionnaire. Used 5 mails.

Subsampling if household with unknown eligibility after the Choice+ recruitment for the NRFU.

Response rates:
Choice+: 34.5%
In-person: 39.4%
Total: 59.7%

Demographic characteristics:
Sex: not many differences

Age: In-person NRFU helped bringing young adults (18-24), from3.7 % in Choice+ only to 5.8% with NRFU compared to a 10% in the population

Race: In-person NRFU brought more diverse sample

Education: In-Person NRFU brought more lower educated respondents to be more align with the general population.

Study outcomes:
Current tobacco use: In-person significantly higher rates if current tobacco use than Choice+; all respondents also significantly higher rates than Choice+ only

Health care provider outcomes: significantly lower proportion of in-person respondents saw a health care provided in past 12 months than Choice+

Conclusion: In-person NRFU brought respondents less likely to respond to standard Choice+ recruitment (younger adults, less educated, people of color).)
There is nonresponse bias in most of important study outcomes when comparing all respondents with Choice+ only respondents.

@threadreaderapp compile unroll thread