Improvements have occurred since study*

*Longer versions of three of these letters appear on the BMJ's website (www.bmj.com)

EDITOR—Nicholl and Turner's attempt to perform a definitive before and after study on regionalised trauma care was beset by logistical problems.1 Firstly, ambulance workers were not empowered to bypass the surrounding hospitals, who in turn were reluctant to be bypassed during the vulnerable period of health service reforms. Secondly, similar systems were compared. The central hospitals in Stoke, Hull, and Preston are all large hospitals with neurosurgical units on site. Thirdly, data were not collected prospectively. The researchers trawled the patients' case records often years after admission. Notes from 1990 were not requested for initial examination until 1993, by which time many had been reduced and put on to microfiche.

Fourthly, the local researcher was not trained on the nationally recognised injury scaling course. There were no intra-observer variability checks to confirm consistent application of scoring methods over the four years. Lastly, significant discrepancies in data accuracy were evident. When the number of direct admissions with severe trauma in 1993 were compared with those counted by the Trauma Research Group at Keele University there was a 25% difference. An outside expert scored the same patients independently and concurred with the Keele findings to within 3%.

Since 1994 we have adopted a strategy to enhance data accuracy. Details on every major trauma patient are checked weekly by a senior clinician and circulated to medical and nursing staff involved in the patient's care. Data shared freely in the clinical domain acts as a two way feedback system to promote accuracy and militate against entry bias in the trauma database. The problem of data validation must be addressed nationally, especially if audit information is to be released to purchasers of health care.

Nicholl and Turner's study represents at best …