Published: 2022 (Sept 2025) | Contact: Theo Lorenc
What do we want to know?
Electronic prescribing systems (‘e-prescribing’) and digital Clinical Decision Support (CDS) tools have been widely implemented as a way to support clinicians in making prescribing decisions, and to reduce prescriber errors and improve patient safety. However, there is debate about their value and impact. Newer-generation tools using artificial intelligence and machine learning aim to address some of the limitations of older systems.
Who wants to know?
This independent research report was commissioned by the National Institute for Health Research Policy Research Programme for the Department of Health and Social Care. The findings may be of interest to clinicians, healthcare policymakers and patients.
What did we find?
Existing evidence on e-prescribing systems shows some impact on clinician behaviour, but it is unclear whether they lead to better outcomes for patients. E-prescribing systems may face a range of problems in practice, including ‘alert fatigue’, increased prescriber error, and issues with partial or inconsistent implementation. More generally, the design and rollout of CDS systems has often not been based on engagement with users or detailed analysis of clinical pathways and workflows.
Most e-prescribing tools use interruptive alerts to warn prescribers of potential risks and flag possible errors. There is scope to think about the possibilities of e-prescribing tools more broadly. This could draw on evidence on the use of CDS for other clinical tasks (e.g. diagnosis, triage, monitoring) and on work on data visualisation in other fields. There is also scope for e-prescribing tools to draw on a broader range of patient data, although there may be practical and ethical challenges. On a theoretical level, it may be helpful to shift away from the focus on alerts towards a broader question of how to mobilise and prioritise data for clinical decision-making.
Patients at high risk of harm from prescriber errors often have multiple health conditions and complex needs, and may be receiving multiple treatments at the same time. Caring for these patients poses specific challenges. Existing tools may not be well-equipped to handle these challenges, partly because they rely on a linear concept of clinical decision-making which does not accurately represent practice.
What are the conclusions?
There is considerable scope for e-prescribing tools to draw on a wider range of data and to present it in more useful ways. There is a large body of data and theory in other fields, clinical and non-clinical, which could inform the development of new tools. However, it is as yet unclear whether these tools represent the most promising solutions to real-world problems, or are likely to overcome the well-known issues with the implementation of existing tools. Future development work should focus on incorporating a realistic understanding of clinical practice, patients’ needs and the functioning of healthcare organisations, rather than being driven mainly by technical advances in data analysis.
How did we get these results?
This project was a pragmatic, non-systematic review with a focus on building theory and identifying areas for future development, and did not use any formally defined methodology.
