Non-response bias in the Norwegian Counties Public Health Survey: Insights from linkage to register data

The Norwegian Counties Public Health Surveys (NCPHS) aim to capture critical aspects of the population's health and well-being. However, selective non-response can introduce bias, potentially compromising survey representativeness. This study assesses non-response patterns in NCPHS using registry data on health-related benefits. NCPHS comprises cross-sectional web-based surveys assessing various health metrics in Norway's adult population. This study uses data from NCPHS conducted in Hordaland County (April 10 to May 17 2018). Health data, including sickness benefits, disability pension and work assessment allowance (WAA), were linked from national registries. From 34,925 invited residents, the response rate was 45.3%, with higher rates among females (50.0%), older individuals (61.0% for ages 60-69 years) and those with higher education (55.9%). Sickness benefit was not associated with participation rates (adjusted risk ratio (RR) 0.96 I, 95% confidence interval (CI): 0.90-1.02). Disability pension, particularly for mental and behavioural disorders, was associated with lower response rates (adjusted RR 0.81, 95% CI: 0.72-0.92). WAA had no overall significant effect, but WWA due to psychological conditions was related to lower response propensity (adjusted RR 0.85, 95% CI: 0.73-0.99). These findings suggest that health-related non-response associated with these benefits has a limited impact on overall population estimates but may introduce bias in specific subgroups, depending on the research question. Though the degree selection into a survey creates bias depend on the research question, knowing that there is relatively little selection by factors such as sickness benefit, WAA and disability pension is important as 22-23% of the population aged 18-67 receive one of these benefits at any time.