Primary Care

Computer aided prescribing leaves holes in the safety net

BMJ 2004; 328 doi: (Published 13 May 2004) Cite this as: BMJ 2004;328:1172
  1. R E Ferner (r.e.ferner{at}
  1. 1West Midlands Centre for Adverse Drug Reaction Reporting, City Hospital, Birmingham B18 7QH

    Patients die from poor prescribing. As with so much else, poor communication is a major culprit. Amoxil (amoxicillin) is misread as Daonil (glibenclamide) because of bad handwriting; 10U is interpreted as 100 [units] because of inappropriate abbreviation; patients are overdosed with a standard release drug when a modified release formulation was intended but not specified.1 The prescribing process is complex, and opportunities for error abound. Patients may be given drugs they are allergic to, or which are contraindicated or have already been prescribed under another name; one drug may interact with another; the dosage, or duration, or formulation, or route may be wrong: in short, anything that can go wrong in prescribing will go wrong.

    Computers can help. They reduce medication error rates by as much as 60% simply by ensuring that prescriptions are legible, complete, and in a standard format.2 That is encouraging, but patients still die from the remaining errors. The NHS Information Authority requires that systems used in general practice “shall cross check prescriptions for known sensitivities, interactions and active ingredient duplications in the patient record. An appropriate warning to the prescriber shall be given.”3

    But GP prescribers put their trust in these systems at their patients' peril. Fernando and colleagues tested four computer prescribing systems.4 One failed to meet the NHS requirements; others failed to warn of potentially serious prescribing errors, especially where drugs were contraindicated. Contraindications account for about 4% of adverse drug events in general practice.3

    The systems could be improved. They might list every contraindication to a drug whenever it was prescribed. That change would trap more errors but risk overwhelming the user with alerts: primary care physicians ignore alerts from nagging computers.5 Relevance is the key. Prescribers need not be reminded constantly that etoricoxib is contraindicated in inflammatory bowel disease, that nalidixic acid should be withheld from patients with epilepsy or porphyria, or that hyoscine-N-butylbromide should be avoided in patients with myasthenia gravis. Yet timely and relevant warnings will prevent disaster. Hospital systems already exist that link patient history, laboratory results, and prescribing data and that present a hierarchy of warnings to inform, advise, and occasionally forbid the prescriber to continue.6

    No human activity is error-free, and we have recognised belatedly that prescribing is complex and prone to error. We need to make it safer—which means increasing the chances that important errors will be avoided or caught by checks before they are translated into harm. We can and should ensure that prescribers—who now include nurses and pharmacists—learn to use medicines safely. Practical examinations in the core skills of therapeutics should help.1 That will still leave patients vulnerable to prescribers' human failings. Computers can improve communications by generating a legible and complete prescription. But Fernando and colleagues show that several widely used systems fail to detect known prescribing errors. Those who walk the therapeutic tightrope in general practice will want the assurance of a safety net that will catch important errors before they harm patients, an assurance that current systems cannot provide.


    • Competing interests REF is a member of the Medicines Management Working Group of the NHS Information Authority


    View Abstract