The Health Survey (GE) is designed to provide reliable direct estimates on an annual basis. Data collection is based on a combination of Internet monitoring (CAWI) and face-to-face monitoring (CAPI). During the Covid-19 pandemic, CAPI monitoring was partially or completely discontinued, resulting in an abrupt change in the measurement and selection effects on GE scores. Moreover, publishing annual figures on the impact of Covid-19 on health-related topics with a delay of about a year limits the importance of GE. The GE sample size does not allow reliable direct estimates to be published over shorter reference periods.
Both issues are resolved by developing a bivariate structural time series model (STM), which estimates quarterly numbers for a selection of eight key health indicators. The method is compared to two alternative methods. The first alternative method uses a univariate STM where no CAPI loss correction is applied. The second variant is based on a univariate STM with an interference variant, which clearly demonstrates the effect of CAPI loss during closure.