Comparison of Diabetes Recorded in Primary Care with Hospital Admission Data in England

IntroductionIn UK prospective studies, record linkage to national cause-specific hospital admission databases provides unique opportunities for epidemiological investigation, since long-term virtually complete follow-up is possible. However, hospital records would not necessarily detect all newly diagnosed cases of diabetes, which is generally diagnosed and managed in primary care.\r\nObjectives and ApproachThe Clinical Practice Research Datalink (CPRD) is a linked database of primary care records, which in 2015 had research-usable data for some 11 million people (7% of the UK population). In 2013 we linked 102,076 (8%) Million Women Study participants in England to CPRD records for the period 1 January 1990–31 December 2012. Among the linked women who had a first record of diabetes in CPRD, we examined the cumulative probability of diabetes being mentioned in national hospital admission data, by time since the first record of diabetes in CPRD.\r\nResultsAfter recruitment into the Million Women Study, 1765 of the linked women without prior diabetes had a first record of diabetes in CPRD. The median year that diabetes was first recorded in CPRD was 2011 (IQR 2007-2012). Seven years after diabetes was first recorded in CPRD, 69.5% (95% CI 67.1% - 71.9%) also had a record of diabetes in hospital admission data. The median time between first mention of diabetes in CPRD and in hospital records was 3.7 (95% CI 3.4-4.1) years.\r\nConclusion / ImplicationsDiabetes first recorded in routinely collected hospital admissions data in England identifies the majority of cases of diabetes recorded in primary care, after a lag of a few years. Hospital admission data can, therefore, be used for epidemiological research into diabetes.