Institute of Epidemiology & Health Care


Using electronic health records to investigate blood biomarkers and onset of cardiovascular diseases: example of differential white cell count

Author A.D. Shah
Author H. Hemingway
Author O. Nicholas
Author A.D. Hingorani
Abstract BACKGROUND: Electronic health records are invaluable for studying clinically recorded biomarkers and outcomes, but conversion of raw datasets to a research-ready format and replication of analyses can be challenging. The differential white cell count is a common blood test which may reflect inflammation and cardiovascular risk, but previous epidemiological studies have been small or in selected populations. AIM: To develop methods to assist in using electronic health record databases for research, and apply them to an exemplar study investigating differential white cell counts and onset of cardiovascular diseases. METHODS: The data source was CALIBER: linked electronic health records from primary care, hospitalisation, acute coronary syndrome registry and death registry. Software was developed to assist selection of relevant diagnostic codes, data manipulation, and multiple imputation of missing data. Individuals in CALIBER without prior cardiovascular disease were followed up for a range of specific initial cardiovascular presentations. Survival models were used to investigate the associations of type 2 diabetes and differential white cell counts with initial cardiovascular presentations. The association of total white cell count with mortality was investigated in CALIBER and replicated in the New Zealand PREDICT cohort. RESULTS: Add-on software packages for the R statistical system were developed and published in an open access repository. Type 2 diabetes was associated with higher risk of coronary disease and ischaemic stroke but lower risk of abdominal aortic aneurysm and subarachnoid haemorrhage. Among 775231 individuals with a record of differential leukocyte count followed up for median 3.8 years, 54980 experienced an initial presentation of cardiovascular disease. High neutrophil counts were strongly associated with heart failure and myocardial infarction, but not angina. Low eosinophil counts were associated with heart failure and unheralded coronary death. Total white cell count showed a 'J' shaped association with mortality in both CALIBER and PREDICT. CONCLUSIONS: White cell subtypes are differentially associated with specific initial cardiovascular presentations. A range of tools were developed to assist researchers using CALIBER and other electronic health record data sources.