Project Overview +

The web is an attractive option for the implementation of surveys, but it suffers from potential limitations, including non-response. This study's primary goal is to identify ways to reduce non-response, attrition, and missing data for web surveys in parallel with the core research activities of the CECCR Research Projects.

Aims +

This study manipulates ways to increase survey responses through incentives, content of the invitations, design of the opening page, motivational messages to reduce missing data, timing and framing of reminders, and other factors related to the content of the surveys themselves.

Because web data collection involves a low per-unit cost as well as a viable means of implementing complicated designs, results from this study will have major benefits both for the CECCR Research Projects and for the broader health research community.

Participants +

Seven hundred non-respondents of a 12-month follow-up survey to an online weight management program were contacted for this study. All participants were members of the Kaiser Permanente health care organization.

Intervention +

Three hundred people are assigned to be followed up by phone and 400 people assigned to be followed up by mail. The telephone survey is conducted by Group Health Cooperative. Participants are not provided with an advance letter or incentive. The mailed survey is mailed with a Kaiser Permanente return address and returned to GHC for tabulation. The mailed survey includes a cover letter signed by KP regional directors and a $5 incentive.

Findings +

The non-response follow-up survey obtained a response rate of 57.3%, with 58.6% of those assigned to phone follow-up responding, and 55.0% of those assigned to mail completing the survey. The difference in response rates by mode is not statistically significant, and the mail follow-up was substantially cheaper than the phone follow-up.

We also found a significant effect of the method of surveying non-respondents (i.e., phone versus mail), with phone respondents giving more positively (socially desirable) responses than mail respondents.

Many of those who attrit from online RCTs and health interventions do so for reasons unrelated to the intervention, and they can be brought back into the sample using a follow-up by phone or mail.

Mail is a more effective follow-up method than telephone, being both cheaper and having similar measurement properties to the online surveys.

Multiple imputation can be used effectively to produce complete datasets for analysis to reduce the effects of selective attrition.

Cash incentives are an effective method for recruiting and retaining subjects for online health interventions.

Capturing paradata or user metrics in online interventions gives researchers insight into the level of engagement or dosage, and can identify areas for improvement to the design of the intervention.

Conclusion +

This study finds that weight loss outcomes among non-respondents are similar to that of respondents.

We found mail to be a cost-effective method of increasing responses rates while minimizing nonresponse bias and avoiding mode effects associated with reports of sensitive topics on the telephone.