Determinants of Survey Completion In Online Autism Research

Background: As familiarity and use of the internet increase, health researchers have found the online environment to be a viable mechanism for data collection. Online research is attractive because of its superior inclusiveness and low cost compared to center-based research. As with any mode of research, however, internet-mediated research (IMR) has its own particular issues with regards to bias and non-response. Non-response must be systematically addressed since the sample becomes more biased (and less generalizable) as non-response rate increases. In order to best utilize IMR for ASD research, patterns of non-response need to be understood and overcome.

Objectives: To examine the child, family, geographic, and survey-related factors associated with non-response in the largest autism IMR platform.

Methods: Data for this study were collected from parents of children aged 2 to 17 years (M = 9.04 years) with an autism spectrum disorder (ASD) who were engaged in a U.S-based online research initiative: the Interactive Autism Network (IAN). Parent (age, education, # of total children, # of affected children, family structure, and rurality) and child (diagnosis, age, gender, and race) factors were attained through questionnaires filled out upon IAN registration. Survey-related factors, including form type, and time since registration in IAN, were captured for two surveys, one focused on access to care and one exploring child vaccination history. Logistic regression models were used to estimate the association between non-response to survey and other factors.

Results: A total of 21,535 survey instances were examined, of which 4,963 or 23% were completed by the parent. In the final multivariate model, increasing child age, greater number of affected children, and increasing duration since registration with IAN were risk factors for non-response (all p<.001). Higher parental education, two-parent household, Caucasian race, and rurality were protective (all p<.01). In addition, parents were more likely to respond to surveys focusing on their affected child rather than unaffected siblings (p=.001).

Conclusions: To our knowledge, this is the first study to examine predictors of online survey non-response in autism research. Compared to previous IMR studies, the response rate in IAN was higher which may be due to the multifaceted nature of IAN. In contrast to typical one-time online surveys, IAN is an active environment that engages parents not only through surveys and additional research opportunities, but through its IAN Community e-newsletter, discussion forums, and research reports. In regards to predictors of survey non-response, data from this study demonstrates both family and survey-related variables play an important role. Since all subjects were already a part of the IAN research project, the differential response rates reflects issues surrounding retention rather than recruitment. Researchers should look towards developing tailored strategies that promote representative research through active enrollment and continued engagement of subjects in online autism research.