Extended recruitment efforts minimize attrition but not necessarily bias

J Clin Epidemiol. 2009 Mar;62(3):252-60. doi: 10.1016/j.jclinepi.2008.06.010. Epub 2008 Oct 1.

Abstract

Objectives: There has been a debate about the effect of extended recruitment efforts on attrition and bias. The aims of the present study are (1) to investigate the effectiveness of extensive multimode recruitment procedures; (2) to study their effect on attrition and bias; and (3) to determine the potential predictors of attrition.

Study design and setting: We used data from the longitudinal population-based study of health in Pomerania.

Results: Using multimode recruitment methods, we reached a follow-up response proportion of 83.6%. In-person contacts at home turned out to be an effective recruitment tool. Sociodemographic and health characteristics of late respondents and converted nonrespondents were most distinct from early respondents but not necessarily indicative of nonrespondents. Analyzing attrition bias, extended recruitment efforts produced an effect only for sociodemographic characteristics but not for health-related indicators. The strongest predictors for attrition from the regression model were late recruitment at baseline, unemployment, low educational level, female sex, and smoking habit.

Conclusion: Extended recruitment efforts appeared justified in terms of response maximization. However, enhanced response proportions may not necessarily minimize bias. In our analysis, aiming for a high-response proportion in terms of health-related indicators had no effect, because late respondents did not differ from early respondents.

Publication types

  • Multicenter Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Aged
  • Aged, 80 and over
  • Bias
  • Female
  • Germany / epidemiology
  • Health Status
  • Health Surveys
  • Humans
  • Logistic Models
  • Longitudinal Studies
  • Male
  • Middle Aged
  • Odds Ratio
  • Patient Compliance / statistics & numerical data*
  • Patient Selection
  • Refusal to Participate / statistics & numerical data*
  • Risk Factors