Identifying Optimum Conditions for Using Internet Surveys in Health Research - A Literature Review

Author: Yaojen Chang, MPH

Primary Advisor: Kathy A. Johnson-Throop, PhD (co-author)

Committee Members: James P. Turley, PhD, RN (co-author); Ruth Sellers (co-author)

Masters thesis, The University of Texas School of Health Information Sciences at Houston.

What is known about when to use an Internet survey as the method of choice in health research? What circumstances determine when an Internet survey can give the best possible results in health research studies? Internet surveys are sometimes clearly the best survey method to use for a particular study. There are also studies for which this method is unsuitable as well as those in which no particular survey type is indicated as being the one best method to use. Based on a review of the literature, certain conditions indicate the appropriateness or inappropriateness of developing an Internet survey for a particular study. These will serve as guidelines to help determine the optimum conditions for when to use Internet surveys in health research. Study features, survey layout design, and study populations best suitable for Internet surveys are identified. Studies suitable for Internet surveys are those which require quick and complete responses, rapid validation and editing, a large number of Internet users, and access to instant online help. Survey layout designs adopting multimedia and branching help give validity to many studies. A simple survey layout design can require the minimum in computer software and hardware. Studies targeting Internet user populations, widely dispersed populations, or a large number of difficult-to-access populations are also suitable for Internet surveys. Studies that are not suitable for Internet surveys through publicly accessible websites include those requiring accurate estimates of response rates, controls for participant characteristics, and follow-up surveys or interventions. Survey layout designs that might cause participants to misinterpret the questions or those that are incompatible with the target-computing environment are unsuitable for Internet surveys as well. Studies requiring a representative sample of the general population or those that reach out to populations with varied computer literacy across the target population are also unsuitable for Internet surveys. Studies to evaluate the advantages of Internet surveys over conventional surveys have shown equivocal results in some cases when evaluating response rate, selection bias, and cost saving. Internet surveys can increase the response rate for sensitive data but not for socio-culturally neutral topics. Although selection bias may occur in both Internet and paper surveys, it is more likely to be accentuated in Internet surveys because of the volunteer effect inherent in that type survey. In terms of cost saving, Internet surveys tend to be more expensive than paper surveys at low sample sizes. However, after a certain critical sample size is reached, the incremental cost is often negligible, whereas paper surveys have definite incremental costs with increasing sample size. This article will propose guidelines to identify studies suitable or unsuitable for Internet surveys and to point out those conditions which seem to give equivocal results. The guidelines are based on assessments and findings of past studies.