Factors associated with health survey response among young employees: a register-based study using online, mailed and telephone interview data collection methods.

Background: Declining response rates are a common challenge to epidemiological research. Response rates further are particularly low among young people. We thus aimed to identify factors associated with health survey response among young employees using different data collection methods.

Methods: We included fully register-based data to identify key socioeconomic, workplace and health-related factors associated with response to a health survey collected via online and mailed questionnaires. Additionally, telephone interviews were conducted for those who had not responded via online or to the mailed survey. The survey data collection was done in autumn 2017 among young employees of the City of Helsinki, Finland (18-39 years, target population n = 11,459).

Results: The overall response to the survey was 51.5% (n = 5898). The overall findings suggest that differences in the distributions of socioeconomic, workplace and health-related factors between respondents in the online or mailed surveys, or telephone interviews, are relatively minor. Telephone interview respondents were of lower socioeconomic position, which helped improve representativeness of the entire cohort. Despite the general broad representativeness of the data, some socioeconomic and health-related factors contributed to response. Thus, non-respondents were more often men, manual workers, from the lowest income quartile, had part-time jobs, and had more long sickness absence spells. In turn, job contract (permanent or temporary) and employment sector did not affect survey response.

Conclusions: Despite a general representativeness of data of the target population, socioeconomically more disadvantaged and those with long sickness absence, are slightly overrepresented among non-respondents. This suggests that when studying the associations between social factors and health, the associations can be weaker than if complete data were available representing all socioeconomic groups.