HES data analysis needs more accurate interpretation
Editor--Two other points are important to add to Keogh's review of the surgical practice for congenital heart conditions in Oxford (BMJ 2005; 330:319-20). The Hospital Episode Statistics (HES) data analysis published in BMJ (BMJ 2004; 329; 825-) has two additional pitfalls.
Firstly, the study depends on the 30-day mortality extracted from HES data. This does not reflect the actual accurate outcome of the operation performed. This has been shown by the Linkage of HES Data to the Office for National Statistics (ONS) Mortality Records report, which showed that overall 8.4% of 30-day deaths are not recorded within HES, with this figure ranging amongst the Inquiry centres from 3% to 19%. Missed links between the HES and ONS records mean that these figures will underestimate the true proportion of 30-day deaths which are not recorded within HES [http://www.bristol- inquiry.org.uk/final_report/annex_b/images/ HES_ONSlinkage_R.pdf].
Secondly, in order to validate the model, Dr. Foster's group converted all missing outcomes of the spells into survivals. This was stated in their study as: "we calculated the odds ratios, including all spells with a missing outcome, by coding them as discharged alive". Our analysis of HES data (presented at the Association of Surgeons of Great Britain and Ireland meeting 2004) has shown that between 15-25% of major operations recorded in HES database have missing outcomes. When followed record-by- record, the average mortality rate is 15% in the same spell. This will without doubt over- or under-estimate the mortality figures, and will distort the conclusions and the model as a whole. In their conclusion, Oxford was shown to have a significant higher mortality than the national average in all three epochs - a conclusion that is not evidence-based. HES data analysis needs to be accurately interpreted, as partly shown by Keogh's review, before reaching significant reliable conclusions.
Abdullah S A Jibawi, Surgical Research Registrar. Leighton Research Unit, the Mid Cheshire Hospitals NHS Trust, Crewe, England, CW1 4QJ.
Competing interests: None declared
Competing interests: No competing interests