Background: In health surveys, considerable effort and expense are invested to achieve a high response proportion and thereby to reduce selection bias. We investigated the interrelation of recruitment efforts and expense with potential nonresponse bias based on data from a large health survey.
Methods: In a population-based health survey, a stratified sample of 6640 residents of the Augsburg (Germany) region was selected, of whom 4261 attended the main study between October 1999 and April 2001. A short telephone interview yielded additional information on nearly half of the nonparticipants. All recruitment contacts were documented, and expenses were estimated on the basis of unit costs. Different recruitment strategies were modeled retrospectively. We compared their cost savings as well as their influence on the response proportion and on prevalence estimates.
Results: The distribution of total contacting cost per individual was highly skewed with 50% of the total sum spent on 17% of the sample. Late responders showed many similarities with nonresponders; both included a higher percentage of people with impaired health and with greater behavioral health risks. We were able to identify recruitment strategies that may save up to 25% of the recruitment costs without significant shift in the parameter estimates. Data collected in the short nonresponder interview proved to be important to correct for possible nonresponse bias.
Conclusions: In general, prolonged recruitment efforts lead to a larger and more representative sample but at increasing marginal costs. Specific cost-saving recruitment strategies that do not enhance response bias can be suggested. Interviews of nonresponders are also useful.