The Road to Hell Winds On: The High Administrative Burden of Maintaining Linked National Health Data.

Abstract

Data linkage studies generate opportunities to answer important health questions, but are also notoriously time consuming, with delays accessing data lasting years. Previously we described our experience of accessing and linking five national datasets for our study 'LAUNCHES QI'. However, the delays and challenges we encountered did not end once data were received - maintaining access and research amendments also required significant time and resources. To outline the costs, time, and workload associated with LAUNCHES, including set-up, annual requirements, and amendments. We also describe the experience of re-using the linked dataset for a second study, CHAMPION, which we wrongly assumed would be simpler. We highlight new delays and challenges, offer advice to researchers, and suggest improvements to the system. LAUNCHES set-up involved 39 documents and data access took 2.5 years. Subsequent amendments/renewals added 75 more documents. Repurposing the data for CHAMPION required restarting the entire process from scratch. It took nearly two years to gain initial approvals and another 18 months to receive the datasets, with 44 documents submitted for set-up and 92 more for amendments and annual processes.New challenges and delays included changes of data controller, data extract errors, and the repetitive amendment process. With growing recognition of the challenges in using health data for research and our own increasing experience navigating the system, we had hoped for positive change. Instead, there are more forms, and processes have become harder, not easier. We believe simple changes could streamline the system, yet with each new study, the document set grows, and the forms get longer. Research with linked datasets is incredibly painful, and the system discourages amendments, stifling innovation. Whilst we've realised the value of linked data, delays to research are significant, and researchers may give up without simpler processes to link and update data.