The study of secondary patient data, particularly represented by claims data, has increased in recent years. The strength of this approach involves easy access to data that have been generated for administrative purposes. By contrast, collection of primary data for research is time-consuming and may therefore appear outdated. Both administrative data and data collected prospectively in clinical care can address similar research questions concerning effectiveness and safety of treatments. Therefore, why should we invest the precious time of rheumatologists to generate primary patient data? This article will outline some features of primary patient data collection illustrated by the German biologics register RABBIT (Rheumatoid arthritis: observation of biologic therapy). RABBIT is a long-term observational cohort study that was initiated more than 15 years ago. We will discuss as quality indicators: (i) study design, (ii) type of documentation, standardisation of (iii) clinical and (iv) safety data, (v) monitoring of the longitudinal follow-up, (vi) losses to follow-up as well as (vii) the possibilities to link the data base. The impact of these features on interpretation and validity of results is illustrated using recent publications. We conclude that high quality and completeness of data prospectively-collected offers many advantages over large quantities of non-standardised data collected in an unsupervised manner. We expect the enthusiasm about the use of secondary patient data to decline with more awareness of their methodological limitations while studies with primary patient data like RABBIT will maintain and broaden their impact on daily clinical practice.