BPharm MPharm MBA MRPharmS
City Hospitals Sunderland NHS Trust
Electronic prescribing (EP) based on an American system called Meditech is now a reality in all inpatient areas at City Hospitals, Sunderland (UK). Postimplementation, we expected to realise the benefits of a more effectively controlled clinical and financial environment leading to safer, more effective and economic management of medicines. This article deals with the reality of those expectations.
The EP system
A key feature expected of an EP system would be better-quality prescribing, and the system does provide controlled and secure access with clear, legible and complete prescription entry by the physician. The control can be further enhanced to allow some group access to restricted lists (for example, dieticians can access nutritional products lists). In addition, we have used agreed order sets, which are treatment protocols of drugs such as standard surgical prophylaxis or medical regimens for treating conditions such as H pylori eradication. This allows the prescriber to simply produce a combination prescription with the correct organisationally approved regimen of drugs, formulations, doses and frequencies, which leads to enhanced consistency of care. Thus, the system has the tools to deliver better use of medicines, but, in practice, there have been problems – as, unfortunately, it has to interact with humans.
Hence, we have seen numerous examples where the clear, legible and unambiguous order is clearly not “what the doctor ordered”. A simple example is the ratio of “aspirin, enteric” to “aspirin, soluble” changes, dramatically in favour of the latter, when you change the formulation type to dispersible. This is due to “picking” lists presented to the prescriber. To ensure unambiguity, the prescriber must choose the drug, formulation, strength and pack size required to be supplied from this alphabetical list. Other examples we have had in practice include amoxicillin 500mg capsules prescribed as 250mg three times a day, which translates onto an automatic label as “Take half a capsule”. A cautionary warning was initially added, but this did not reduce the occurrence of this error. To prevent these errors, we introduced a “dose step” as a mandatory function that could not be overridden. This allowed multiples of an agreed dose only (in the case of amoxicillin, 500mg), and this is capped at a multiple of 9, preventing an inadvertent addition of an extra zero (ie, 5,000).
This ability to provide what is often termed decision support (but is better described as decision constraint) to control the system users and prevent inappropriate medication use has many advantages. The system can look for duplication, dose range, allergy and interaction checking, and provides restrictions of available routes of administration, with the added advantage of warnings, should a route be unlicensed. Automatic directions could also be provided when a drug is chosen. In use, this can be a problem, as the quality of output depends on the data input and system structure as, often, drug choice errors are identified by inappropriate frequencies (eg, penicillamine and penicillin). In addition, unless carefully managed, decision support can potentially lead to a large number and frequency of warnings (often described as “noise”) and, with time, an over-reliance on the system to stop all errors develops, which leads to less awareness by the system user.
Our major problem in terms of control has been the inability to maintain our formulary, traditionally an important role for pharmacy services in the UK. The system requires that all medicines to be administered by nursing staff are available on the system. Clinically, we would not want a nonspecialist (especially in a short-stay surgical speciality) to alter a specialist medicine started in primary care or by another consultant. This means that all specialist medicines are available for prescription on the system. Although this does not mean that the pharmacy needs to stock the medication, the transparency of the system soon makes it apparent if patients do not receive their medication.
The system allows the pharmacy to provide clinical support information to prescribers, reviewers and those administering medicines, although this needs to be constantly updated and maintained. As everything is recorded by individualised passwords and PINs, there is now a complete, accessible and auditable trail of staff who prescribe, review and administer each medication. Additionally, the coding of the order changes at each stage of the supply and discharge processes, making it possible to determine the status of the order and avoiding the need to ring the dispensary. However, to access these features requires ongoing training of ward-level users and, in practice, a telephone call remains the preferred option.
This accessible audit trail means that we now have improved available information on usage, but not on expected costs. While we have automated the individualised prescribing and administration of medicines, we maintain a traditional UK hybrid system of supply using stock and individualised packs rather than a unit-dose supply system. Thus, we are unable to allocate accurately all the costs to specialty, consultant or prescriber.
The major advantage of the system to pharmacy staff is that work is prioritised, with only newly prescribed items being identified as a priority. With computerised prescription, different staff can access a prescription at multiple locations and it never gets lost. However, the complexities of the system and the perception of users that pharmacy staff are system experts means that clinical and technical staff time on the ward is diverted to solving system problems, although this provides opportunities to correct and clarify other clinical errors.
The two aims of EP in the UK model of healthcare are cost savings and a reduction in medication errors, but this will not occur without modernisation of current working practices and an effective system design that accepts the frailties of the human interface. At Sunderland, the system requires intensive pharmaceutical support mechanism and infrastructure at patient level to maintain error reduction, as users often seek alternative “workarounds” that precipitate different and potentially higher-risk errors. Those who simply expect a computer to provide all the solutions should remember that, “The computer is a fast idiot, it has no imagination; it cannot originate action. It is, and will remain, only a tool to man” (American Library Association 1964).
This article is based on a presentation at the British Pharmaceutical Conference 2005 in Manchester