One hundred and fourteen features of general practice prescribing systems that support quality and safety were defined, with a focus on promoting better use of medicines. All features were rated by a multidisciplinary expert panel as being likely to have medium to high positive impact on safety, quality of care and/or usefulness to the clinician or the patient. Ten "aspirational" features were identified; they require significant system changes before they can be implemented, for example availability and use of national clinical terminologies and unique patient identifiers. Many of these are essential "building blocks" for e-health, and are necessary for sharing patient data, for example a shared medication list.
The strengths of our study include the comprehensive approach and the focus on safety and quality, particularly in relation to medicines use. The features were developed and rated drawing on multiple sources, including the international literature and experts with different backgrounds. An expert panel reached consensus on the expected impact of the features not only on patient safety and quality of care, but also on whether they were likely to be useful to clinicians and to patients.
The scope of our features list has some limitations. The features were developed for the Australian general practice setting, although we believe most of the features are relevant to e-prescribing in other settings. Non-clinical functions of software were largely excluded, for example billing and practice management. The features are high level statements rather than detailed specifications; if they are used as the basis for specifications, factors such as usability and the quality of knowledge bases used for decision support also need to be considered.
Our results endorse the importance of many of the features that others have identified in different settings as being important for safety and quality in e-prescribing. We took a similar approach to Avery et al. in the UK and Bell et al. in the United States, with a common goal in all three studies being to support patient safety. Our study focussed on the quality use of medicines in general practice in Australia; however, our list of 114 features includes related aspects of prescribing systems that we believe are important, for example security and back-up of clinical data, transparency and support for privacy legislation. Avery et al. produced 55 clinical statements related to medicines management errors and safety considerations for general practice computer systems in the UK, and their importance was rated. Bell et al. developed 60 recommendations for e-prescribing systems' ability to meet the goals of improving patient safety and health outcomes, helping patients manage costs, maintaining patient privacy and promoting clinicians' acceptance of electronic prescribing. The recommendations were rated by level of impact on each of these four dimensions.
While the rating of our features cannot be compared directly with those of Avery et al. and Bell al (as different scales were used), there was considerable overlap in the areas rated as important. In Avery et al's study the key themes around which consensus was reached on importance included computerised alerts; the need to minimise spurious alerts; making it difficult to override critically important alerts; having audit trails of such overrides; effective computer-user interface; and the need to be able to run safety reports. For the 60 recommendations made by Bell et al, 52 were rated in the clearly positive range on the 'patient safety and health outcomes' dimension, including all 19 features related to patient identification, medication selection and alerts and warnings for prescribing.
The features list produced in this study could be used by policymakers and professional bodies to develop certification criteria or standards for prescribing software systems, and as guidance for software vendors. It could also be used by clinicians to assist them in assessing the suitability of an e-prescribing system when selecting or changing systems.
Several lines of follow on work suggest themselves. Detailed specifications or criteria can be developed for incorporating (or testing) the features in e-prescribing systems. (Following this study we assessed seven e-prescribing systems in Australia to find out whether the features had been implemented in these systems; the results will be reported in the near future). It would be valuable to observe the actual impact of each feature when implemented. We would emphasise the importance of monitoring and evaluating the effects of e-prescribing systems when they are introduced or changed, being especially vigilant for unintended effects. Finally, new functionality will become possible as technologies and systems evolve; any list of recommended software features requires ongoing review.