OBJECTIVES: The use of administrative claims data to estimate the effects of health related outcomes for specific diseases, treatments and interventions is surging. To avoid selection bias when comparing effects between groups and to mimic randomization, matching or weighting techniques are applied. Aim of this study was to assess the use of matching and weighting techniques in studies using German administrative claims data.

METHODS: A systematic literature review via PubMed was conducted to assess the application of matching and weighting techniques in studies based on German administrative healthcare data published until May 2017. Relevant studies were identified via keywords linking the terms “matching” or “weighting” with various synonyms for claims data and Germany. Titles and abstract were screened by two independent researchers. The studies were stratified by type of used data, category of study objective (e.g. burden of disease) and applied methodology (e.g. propensity score matching or weighting).

RESULTS: In total, n=363 studies were identified of which n=114 met the inclusion criteria. The most frequent study objectives included cost analyses followed by burden of disease assessments and studies on healthcare resource utilization. Direct matching approaches based on variables such as age and gender were used in almost two thirds of the studies, followed by matching on the propensity score, which was applied in roughly one third of the analyses. Weighting techniques such as inverse probability of treatment weighting or newer approaches such as entropy weighting were rarely incorporated. Claims data from health insurances were the most prominent data source followed by administrative outpatient data.

CONCLUSIONS: To balance treatment and control groups in German claims data, most researchers rely on matching methods, especially direct matching and propensity score matching. The use of weighting techniques and relatively new approaches is rare in the German context but should be considered in the future.