Identifying effective prescribing alerts to include an electronic medication management system

Hospitals in Australia are making vast investments in electronic medication management (EMM) systems; a major driver is the potential benefits of these systems to significantly reduce medication errors, particularly when systems incorporate clinical decision support (CDS). Most EMM systems include a number of CDS alert categories to warn prescribers of potential medication errors, such as drug-allergy interactions, drug-drug interactions, exceeding a maximum dose range, and therapeutic duplication. Enabling multiple alert categories may reduce prescribing error rates further than any single alert category. However alert fatigue and high rates of alert override are well-recognised consequences of excessive interruptive alerts. Information is lacking on which prescribing alert category or categories to include in an EMM system to maximise the potential safety benefits of alerts, balanced against the risk of alert fatigue. The overall aim of my research was to fill this evidence-gap by synthesising the literature, consulting with experts and then summarising this information to present evidenced-based guidance to Australian hospitals to assist in identifying effective prescribing alerts to include in hospital EMM systems. The first aim of this thesis was to critically appraise the literature on interruptive medication prescribing alerts in hospital inpatient EMM systems. A systematic review identified twenty-three papers that met all the inclusion criteria. The review revealed that in 53% of studies, alerts improved prescriber behaviour or patient outcomes. The greatest volume of evidence arose from drug-condition interaction, drug-drug interaction and corollary order alerts, with drug-condition alerting having the greatest evidence of positive effect. There was no comparative research evidence indicating that a specific category of alerts is more effective than another, and little is known about the impact on prescribing or patients when alerts from multiple categories were incorporated within the same system. With limited research evidence available to guide alert selection in EMM systems, the second aim of this thesis was to determine the process by which Australian hospitals make decisions about which alerts to include in the EMM systems, and the basis for these decisions. To do this, a standardised, semi-structured telephone survey was conducted on a purposive sample of key stakeholders involved in EMM implementation in Australian hospitals. This survey revealed that the three most frequently implemented alert categories were drug allergy interaction alerts (100% of hospitals surveyed), drug-drug- interaction alerts (100% of hospitals surveyed), and dose range checking alerts (69% of hospitals surveyed). Respondents reported that a high degree of customisation of the vendor out-of-the box functionality was required to improve sensitivity and specificity of alerts and to minimise alert fatigue. Configuration decisions appeared to have been influenced by a perception that alerts change prescriber behaviour and improve patient outcomes and may have been shaped by factors additional to those identified by respondents, such as alert experiences in pharmacy dispensing systems. Few hospitals had undertaken evaluation activities, and stakeholders were not confident that the perceived benefits of alerts are being achieved in their local settings. Despite this, stakeholders favoured optimising existing alerts rather than removing alerts. In summary, this program of research has identified that while there is some evidence of the effectiveness of specific individual categories of alerts, there is little research evidence available to guide selection of the most effective combination of alert categories for inclusion in EMM systems. Yet implementers in Australian hospitals have actively embraced multiple categories of interruptive alerts and believe that there is published evidence which supports their use. Few hospitals have assessed alert effectiveness, and implementers harbour doubts about the likely effectiveness of their alerts locally. The Australian experience offers guidance to new implementers on the most commonly used alerts, and lessons learned on designing and implementing effective interruptive prescribing alerts. There is a significant research gap on which alerts to include and exclude from an EMM system. Ongoing evaluation of the effectiveness of interruptive medication prescribing alerts is required, in particular the cumulative impact when different combinations of alert categories are incorporated within the same system.