Multicentre study of the quality of a large administrative data set and implications for comparing death rates

Background: The aim was to compare the completeness and accuracy of the English Hospital Episode Statistics (HES) with a 'gold standard' data set for a sample of hospitals and to determine the effect of data quality on comparisons of hospital death rates. Methods: A multicentre audit of data quality was undertaken, based on a sample of all elective abdominal aortic aneurysm (AAA) repairs performed in England. All elective AAA repairs in nine collaborating hospital trusts were included over a 2-year interval. Cases were identified from HES, local databases, hospital administration systems and theatre records. The main outcome measures were the numbers of cases and deaths according to HES compared with case-note review. The recording of co-morbidities and the effect of data accuracy on mortality analyses and risk adjustment were quantified. Results: A total of 1102 elective AAA repairs were identified from HES data. Of 962 procedures with case-note review, 827 (86.0 per cent, 95 per cent confidence interval 84.0 to 88.0 per cent) were confirmed as elective AAA repair. The survival status with HES was 99.8 per cent accurate on comparison with the Office for National Statistics death registry. There was no significant difference in mortality assessment between the HES data and the 'gold standard' data set (5.3 versus 5.0 per cent; P = 0.753). Smaller hospitals were more affected by data inaccuracies than larger hospitals. Conclusion: This study confirmed that HES data can be used effectively to compare mortality between hospitals. Administrative data will be used increasingly for assessing performance and clinicians should accept responsibility to improve coding.