Estimating Healthy Volunteer Bias Impact on Mortality in Observational Cohorts - a Projection for the German National Cohort

Ulrich O. Mueller, University of Marburg
Wolfgang Hoffmann, Universität Greifswald

Surveys participants are healthier and live longer than non-participants. Since health status influences all human activities, this healthy volunteer bias may compromise generalizability of survey findings to non-participants and the population of reference. Relevant are long-term prospects of the bias: will it grow, stabilize, diminish? Using data on subjective health in 13.226 respondents aged 20-69 in the German General Social Survey ALLBUS 2004-2012, we estimate the healthy volunteer bias impact on projected mortality and survival of the study protocol population of the German National Cohort (GNC) with n=200.000 aged 20-69, recruited 2014-2018. We show that paradoxically the healthy volunteer bias may increase with participation rates, but will have diminishing effects on generalizability of survey findings. Even with persistent survival differentials, any healthy volunteer bias in recruitment will leave only moderate traces in the mortality-follow-up. Empirically, survival differentials due to the healthy volunteer bias do not persist over long times anyway.