All trusts must have electronic prescribing systems by 2005, the NHS information strategy stipulates. One trust has been successfully using such a package for some time, writes Peter Mitchell

The NHS's drive towards electronic prescribing has plenty of good reasons behind it, but none more compelling than the horrifying number of prescribing errors made in hospitals using manual systems.

It is estimated that there are errors in 7 per cent of all hospital prescriptions worldwide. The largest study yet, at the University of California, showed that deaths due to accidental poisoning doubled to 7,391 people in the 10 years to 1993, and are rising faster than any other cause of death apart from AIDS. Workloads, illegible prescriptions and drugs with similar names are the most common reasons given for mistakes.

More worrying is that the report suggests many errors go undetected and even those detected are not always reported.

In fact, we only know the US figures because there is a great deal of consumer-driven - and lawyer driven - research over there. No-one in UK hospitals is prepared to talk about it on the record. But the general consensus is that the problem is just as big here as in the US (see box).

Part of the problem is the sheer number of drugs stocked by hospitals. NHS trusts hold fewer than the typical US hospital, but nevertheless some large acute trusts here hold between 3,000 and 4,000 different drugs.

Nurses check all drugs with at least one other person before administering, but mistakes still occur.

Some types of error can never be excluded, but electronic prescribing, with its potential for automated checking, can sift out some of the worst bloomers.

One trust, University Hospital Birmingham, has been working with an electronic prescribing package for some time, and has now collected some useful data on its effectiveness.

For example, the hospital's renal unit has disallowed attempted prescriptions 58 times in 11 months.

These are instances where the application has flagged a drug-drug or drug-patient interaction or a contra-indication. Had the drugs been given, they would have had a serious adverse effect.

The package was developed by Birmingham University's Wolfson computer laboratory during the mid1990s. HBOC bought the marketing rights in 1998 after a year-long search for a prescribing package, and has rebadged it as Pathways Prescribing Manager. According to the company (now McKessonHBOC), it was one of the few products available at the time that was not merely a pharmacy ordering system.

The Wolfson computer laboratory is still heavily involved with the product, however, and has been contracted by McKesson HBOC for some of the ongoing development work.

Wolfson laboratory director Margaret Peters says the software was developed partly to address costs, but with an over-riding aim of improving patient safety. Other priorities were to ensure the efficient use of staff and resources, and assist the clinical management of patients.

'The application was never designed to replace doctors but to support them with pertinent and timely information, ' she says. 'We believe strongly that we should not be using computers for diagnostic purposes or to try to represent human reasoning. Besides the fact the NHS simply cannot afford such systems , doctors are perfectly good at arriving at diagnosis.'

But what they are not always consistently good at is following care protocols, she says. Added to this is the fact that leading medical opinion on what therapies are most appropriate is not static, and doctors find it hard to keep up.

The application is designed around traditional methods, with screens identical to familiar paper-based drug charts: this was intended to help gain acceptance from clinicians. For example, there is an equivalent of the customary 'yellow sticky', so that users can attach electronic messages to existing prescriptions.

The package is rule-based, with rules that can be tailored to local conditions, preferences and specialties. It is linked to a drugs database - the hospital stocks 1,000 different drug types - and to the patient administration system, but also receives other input such as lab and x-ray results. It continually reviews patient data, proposes investigation protocols and assesses drug therapies for given diseases and conditions in the light of this data.

Hand-held terminals allow doctors and nurses to enter prescriptions and administration data at the bedside.

The Symbol-made terminals - running Windows NT client software - are radio-linked to the Dell host server and application software, so all input is evaluated in real time against the patient and clinical data. The system can also be accessed via PCs in the ward offices, drug rooms and hospital pharmacies.

There are alarms, flagged on all screens, advising staff of abnormal test results. Alerts identifying less serious irregularities are only seen when the relevant patient record is accessed. Forty per cent of alarms are acknowledged within one minute and 95 per cent within the hour.

To prevent drugs being wrongly prescribed and ensure patient safety, the system highlights drug interactions, dosage and frequency errors, patient allergies and contraindications. The software will block any prescription that it considers dangerous for a particular patient.

Where a contra-indication would not be too severe, warnings can be overridden, but doctors must first enter their password to acknowledge that they are taking full responsibility.

In some instances, only doctors with appropriate clinical authority can override the system. But a full audit trail is produced and this shows that in up to 85 per cent of cases the prescriber has heeded the warning and changed the drug, the frequency or the dosage.

This ability to countermand the system is seen as an important asset.

Dr Nick Richards, the hospital's clinical director of renal services, says it is important that clinical freedom is not being compromised and that in most cases the doctor has the final decision.

Prescribing Manager has been on the renal ward now for two years and is fully entrenched and widely accepted by all staff. A recent audit showed that every doctor and nurse is using it, with an average log-in rate of once an hour. A snapshot of one week's activity showed 2,353 log-ins over seven days, with 114 different users: eight consultants, 21 other doctors, 56 nurses and 29 others (mostly pharmacists).

Dr Richards says many treatments are now protocol-driven, with doctors being guided by the system to preferred therapies. The system has helped make significant improvements in clinical practice, he says. 'Changing rules can change clinical practice dramatically and ensure that treatments are appropriate to the patient and clinical practice.'

As an example, a rule was developed to limit the use of antibiotics.

Infection is a particular problem in patients with renal failure, so the use of antibiotics is high. But some patients remain on antibiotics for longer than necessary - some never come off them. So in September 1998, a rule was introduced to terminate intravenous antibiotics after three days and oral courses after five days after a 24-hour warning. Doctors do have the option to continue the therapy, but in most cases this is not necessary.

The result was a 34 and 22 per cent cut in intravenous and oral prescriptions respectively. Spending on antibiotics is down 25 per cent - a saving of£15,000 a year. And there has been no discernible change in outcomes, so patient care is not being compromised.

Similarly, Tramadol (a narcotic analgesic similar to morphine) was being administered on average 125 times a week, but in some weeks was reaching 200 and clearly on the increase. The unit feared that it was being over-prescribed, and revised the automatic rules: now it is administered only 50 times a week, says Dr Richards.

An internal satisfaction survey found that both doctors and nurses believed that enhanced legibility of scripts and availability of data had improved patient safety. Most doctors value the guidance and warnings, while nurses appreciate the availability of lab data and the fact that drugs are allocated to the appropriate ward round.

Nurses also appreciate not having to trek to the pharmacy with drug charts, being able to administer drugs at the right time, and not having discharged patients kept waiting for drugs. They feel better able to manage discharges, mostly because pharmacy staff are now able to dispense take home drugs from legible scripts.

'It is too easy to make mistakes with hand-written prescriptions, especially with medicines having similar names, ' says Dr Lindsey Harris, McKesson HBOC's EPR programme manager and a former paediatric registrar. She says it is impossible to remember all the known interactions between drugs and their contra-indications.

'With new drugs becoming available daily, doctors are finding it increasingly difficult to maintain drug knowledge, especially given all of the other pressures.'

The drugs don't work: prescribing errors In December, a pilot study investigating medical errors - the first attempt to measure the problem in Britain - showed that one in 14 patients suffered some kind of adverse event such as diagnostic error, operation mistake or drug reaction.

Provisional research figures suggested that a further 280,000 people suffer from nonfatal drug-prescribing errors, overdoses and infections. The victims typically spend six extra days recovering in hospital - costing the NHS in England about£730m.

The study is led by Charles Vincent, head of the clinical risk unit at University College London. The early results come from two London hospitals: in one, 32 out of 480 patients in four different departments were victims of hospital mistakes.

Medical error is the third most frequent cause of death in Britain after cancer and heart disease, killing up to 40,000 people a year.

The Department of Health is planning a three-year national study of 20 hospitals and 10,000 medical records to establish how medical errors occur and how to prevent them.

Another study, of junior doctors at 20 hospitals in the north of England, found that 46 per cent admitted they had given a wrong drug or the wrong dose at least once in the period under investigation.