Design and implementation of data collection and linkage of electronic health records in a large UK cluster-randomised trial of asthma management (ARRISA-UK)

Electronic health record data holds great potential for conducting large, efficient randomised controlled trials. Despite progress towards greater availability of linked NHS datasets, the use of routine clinical data remains challenging for trialists. In this paper we describe the design, adaptations and implementation of methods for data collection and linkage in ARRISA-UK: a cluster-randomised controlled trial of a complex asthma management intervention involving 275 primary care practices across England, Wales and Scotland. Our methods were designed to build a dataset of linked primary care and secondary care data for approximately 10,000 'at-risk' asthma patients to measure the trial's primary outcome (asthma crisis events comprising respiratory-related hospital admissions, emergency department attendances and/or death for 'at-risk' asthma patients) and secondary clinical outcomes including the impact of the intervention on ∼180,000 asthma patients at participating practices. A high level of practice attrition (33%) was observed due to data extraction delays and technical barriers, patient identification errors, and concerns about the processing of patient identifiable data for the purpose of record linkage. We highlight the technical achievements, barriers and lessons learned from ARRISA-UK and propose recommendations to facilitate future data-enabled trials, including greater resourcing in recognition of their complex nature, improved systems of support and training in primary care, and the need to maintain and improve clinician and public trust in research data use for long term sustainability.