Are we setting about improving the safety of computerised prescribing in the right way? A workshop report.

Background: Prescribing errors are common and costly. Technology should enable safer prescribing. The two main current methods of doing so are computer initiated clinical support software (CDSS) and the user initiated information retrieval (IR) systems. However, despite the near universal availability of computerised prescribing support in the UK, errors continue.

Objective: To evaluate the experience of UK primary health care professionals using CDSS and to consolidate current technical opinion and literature in this area with the aim of creating useful hypotheses for guiding future academic investigation and industrial development.

Study Design: The study was a synthesis, drawing together a literature review and views from experts in the field to explore from a qualitative perspective where and how CDSS and IR could be used to improve prescribing safety in primary care. We conducted a literature review, held a workshop to explore issues in practice and had a follow-up expert panel meeting to confirm the findings. The workshop was recorded, transcribed verbatim and analysed thematically.

Participants And Setting: The study involved primary care practitioners, system developers, information suppliers and academics.

Outcomes: Although CDSS is incorporated into primary care electronic patient record systems there does not appear to be an associated marked reduction in prescribing errors. Clinicians are frustrated with current systems, and are concerned these may have a negative impact on patients. There is an unhelpful signal-noise ratio with too many clinically irrelevant alerts and insufficient recognition of the potential downsides of over alerting - possibly making compliance less likely, having a negative impact on the doctor-patient relationship and overloading clinicians. A preferred way forward would be alerts based on quantitative risk assessment of interaction at the level of the preparations being prescribed, rather than theoretical possibilities of interactions between classes of drugs.

Conclusion: Prescribing errors remain a major source of unnecessary morbidity and mortality and current systems do not appear to have significantly reduced this problem; nor has the extensive literature about how to reduce unnecessary alerts been taken into account. We need a new and more rational basis for the selection and presentation of alerts that would help, not hinder, the clinician's performance.