Pros and cons of using anonymised linked routine data to improve efficiency of randomised controlled trials in healthcare: experience in primary and emergency care

Background The use of anonymised routine linked data in designing and conducting randomised controlled trials (RCTs) has great potential. Sample sizes can be large, inclusion rates high and follow up periods prolonged, while the disruption to participants’ usual routines may be minimised. However, challenges and limitations in using routine linked data in RCTs remain. Aims To describe, in primary and emergency settings, challenges and opportunities associated with designing and conducting RCTs using anonymised linked routine data to identify study participants and gather outcomes. Methods In each of these trials we have used routine linked data as a key part of the research study design: PRISMATIC (a stepped wedge trial of predictive risk stratification in primary care) utilised linked data outcomes related to emergency admissions to hospital, GP activity and outpatient appointments. Outcomes were included for 230,000 people registered to participating GP practices in the Swansea area SAFER 2: a cluster randomised trial of referral to falls services by ambulance paramedics included linked data outcomes related to subsequent emergency episodes for 4,655 patients across three UK regions TIME: feasibility trial of Take Home Naloxone randomised by city; routine linked data used to identify population for inclusion in follow up and outcomes Regulatory processes - ethics, research and information governance permissions - have caused delay in each trial; inclusion rates have been much higher than is usual in RCTs (outcomes for >80% of eligible patients); large trials have been achievable at reasonable cost (each trial