The four most common computerised drug information systems in place in the UK fared poorly at producing alerts for a series of 18 potentially problematic scenarios, says a report in this week's British Journal of Medicine.
The researchers, largely primary care doctors, fed dummy patient data and prescriptions into the four systems.
No system provided more than seven alerts for the 18 dangerous scenarios tested. Two systems produced four alerts and the final system warned doctors in only three of the cases.
The names of the systems were withheld to preserve the anonymity of the technology suppliers.
Robin E Ferner, director of the West Midlands Centre for Adverse Drug Reaction Reporting at City Hospital, Birmingham, said the results suggest that doctors "put their trust in these systems at their patients' peril".
Ferner also pointed to findings that show how error rates can be cut by 60% alone by computerised systems, which ensure prescriptions are complete and legible.
He emphasised that bombarding doctors with alerts for every possible problem would not solve the situation automatically.
"That change would trap more errors, but risks overwhelming the user with alerts: primary care physicians ignore alerts from nagging computers."
The article appeared in a special issue of BMJ dedicated to electronic communication and healthcare.