| Summary: | Previous studies in the UK have established that minority ethnic groups as a whole experience more ill-health and onset of morbidity at younger ages or at lower levels of risk than the ‘White British’ population. Since the Race Relations Act of 1968, the official collection of ethnic group statistics by all government bodies has been mandated as a pre-requisite for identifying and tackling ethnic inequalities. The capture of ethnicity data in routine health records across the UK National Health Service forms part of this initiative. Although the facility to record ethnicity has been available in primary care since 1991 and in secondary care since 1995, until recently, unsystematic recording resulted in poor quality of the initial data, limiting the usefulness of these data for clinical care, commissioning and research. The incentivisation of ethnicity recording in 2006 as part of the Quality and Outcomes Framework has resulted in an improvement of the quality of these data, though their suitability for use in UK-wide population-based research, at the commencement of this PhD, had not yet been explored. The studies reported in this thesis investigated the utility of electronic health records for research into ethnic differences in health and comprised three sub-studies. Firstly, the completeness, usability and generalisability of ethnicity data captured in primary and secondary care databases were assessed. Results showed that in 2012, valid ethnicity was recorded for 78.3% of patients in the Clinical Practice Research Datalink (CPRD), 79.4% of inpatients, and 50% of A&E patients and outpatients in the Hospital Episode Statistics for England (HES). Over 80% of patients with multiple ethnicities recorded had codes which either were identical or fell into the same five high-level ethnic group categorisation. The ethnic breakdown of the CPRD was found to be comparable to that of the combined censuses for England, Wales, Scotland and Northern Ireland, suggesting that studies of ethnic populations within the CPRD can be generalised to the UK population, particularly when using data from 2006 onwards, where completeness and consistency are highest. Secondly, in collaboration with the UK Biobank study, a pragmatic and comprehensive definition of diabetes mellitus for use in electronic health databases was developed. Once applied to the CPRD, the algorithms identified 34,530 individuals with type 1 diabetes and 355,717 individuals with type 2 diabetes. The incidence of type 2 diabetes was almost doubled in South Asian compared with White groups (70.7 vs 42.0 events per 10,000 person years). After adjustment for gender and age group, the risk of type 2 diabetes was over three times higher in the South Asian group compared with White the group (Hazard Ratio 3.27 95%CI 3.19, 3.35). Finally, a prospective cohort study of 860,000 patients registered with the CPRD was undertaken to quantify ethnic differences in the risk of incident coronary heart disease (CHD) and the extent to which this relationship is modified by the presence of type 2 diabetes. The presence of diabetes increased the risk of CHD by 40%, although this reduced to 22% after accounting for age, gender and deprivation (Hazard Ratio 1.22 CI95 1.20, 1.25). The excess risk associated with diabetes was markedly higher for ethnic minority groups, with an adjusted increase of 60% and 75% in South Asian and Black African/Caribbean groups respectively, compared with 28% in the White groups. Adjusted rates of CHD were consistently higher in South Asian groups and lower in Black African/Caribbean groups, with differences more pronounced amongst men than women. Ethnic differences in CHD risk were consistently more pronounced amongst patients without type 2 diabetes than in those with type 2 diabetes. The studies have generated novel results which provide valuable information about the usability and generalisability of ethnicity data available in UK electronic health records. They have replicated findings from non-database studies of the prevalence and incidence of diabetes and extended our knowledge of the patterning of ethnic differences in heart disease outcomes. They represent the first ever use of UK routine electronic health records to answer these questions in relation to ethnicity. Together, the findings reported in this thesis provide a unique insight into the ways in which routinely recorded ethnicity data can be maximised for the purposes of epidemiological research into health inequalities across the UK.
|
|---|
