Differences in realized access to healthcare among newly arrived refugees in Germany: results from a natural quasi-experiment

BACKGROUND: Germany has a statutory health insurance (SHI) that covers nearly the entire population and most of the health services provided. Newly arrived refugees whose asylum claim is still being processed are initially excluded from the SHI. Instead, their entitlements are restricted and parallel access models have been implemented. We assessed differences in realized access of healthcare services between these access models. METHODS: In Germany's largest federal state, North Rhine-Westphalia, two different access models have been implemented in the 396 municipalities: the healthcare voucher (HcV) model and the electronic health card (eHC) model. As refugees are quasi-randomly assigned to municipalities, we were able to realize a natural quasi-experiment including all newly assigned refugees from six municipalities (three for each model) in 2016 and 2017. Using claims data, we compared the standardized incidence rates (SIR) of specialist services use, emergency services use, and hospitalization due to ambulatory care sensitive conditions (ACSC) between both models. We indirectly standardized utilization patterns first for age and then for the sex. RESULTS: SIRs of emergency use were higher in municipalities with HcV (ranging from 1.41 to 2.63) compared to emergency rates in municipalities with eHC (ranging from 1.40 to 1.71) and differed significantly from the expected rates derived from official health reporting. SIRs of emergency and specialist use in municipalities with eHC converged with the expected rates over time. There were no significant differences in standardized hospitalization rates for ACSC. CONCLUSION: The results suggest that the eHC model is slightly better able to provide refugees with SHI-like access to specialist services and goes along with lower utilization of emergency services compared to the HcV model. No difference between the models was found for hospitalizations due to ACSC. Results might be slightly biased due to incompletely documented service use and due to (self-) selection on the level of municipalities with municipalities interested in facilitating access showing more interest in joining the project.