In 2006 it was reported that almost half of all police forces that were audited by the HM Inspector of Constabulary – HMIC – were found to have errors in their police databases
Sir Michael Bichard’s enquiry into the intelligence failures leading up to the murders of Soham school girls Holly Wells and Jessica Chapman led to a 2004 report that recommended measures to improve the quality and timeliness of data input into the Police National Computer (PNC). HMIC’s audits of Britain’s 51 police forces were subsequently trained through Bichard’s lens.
The “progress report” published yesterday was meant to demonstrate how well the Home Office had responded to Bichard by making police data more reliable. It showed how there was a long way to go before police data could be treated as gospel.
“HMIC has commenced direct communications with 13 forces which are causing varying degrees of concern in relation to their actual performance or their general direction of travel,” said the progress report.
It noted evidence provided by HMIC audits about the timeliness of data input into police computers.
Almost a third of British forces were not meeting tough statutory targets for inputting data about arrests and summons on the computer in time, it said, drawing its data from the completed audits of data quality and related working practices HMIC has done of British police forces.
It also noted that 39 per cent of forces were not inputting records of court proceedings within statutory deadlines.
But it skirted over the other key data concern for Bichard, that of data quality. Error rates of between 15 and 86 per cent were identified in police data in the years before the Soham murders. Data errors are still a problem, as demonstrated by recent string of reports about the Criminal Records Bureau, which draws its data from the PNC.
The most recent PNC audit report published by HMIC, that of Avon and Somerset Constabulary, noted that 22 per cent of records that had already been checked by supervisors still contained an error. The error rate concerned a sample of records input in recent months. Old data, which might contain more errors, is not audited.