394 CHAPTER 5 1.2. Integration of screening tools in clinical decision support systems The lack of clarity of some STOPP/START criteria makes their conversion into algorithms for use in CDSSs challenging [29]. In Chapter 2.3, we described that multiple consensus rounds were necessary to make the current version of STOPP/ START (version 2) suitable for software systems [30]. Therefore, in the process of future guideline development, the collaboration of medical experts and experts in medical informatics could be valuable to avoid ambiguous wording in formulating guideline recommendations, possibly easing the integration of such recommendations in software systems. The introduction of electronic health systems provides opportunities to integrate CPG recommendations as screening tools to identify patients at risk of potentially inappropriate prescribing and assist healthcare professionals in clinical decisionmaking, thereby reducing unintentional guideline deviations [24,25]. Although the software cannot replace decision-making by healthcare professionals and patients, it can effectively translate knowledge acquired by CPGs, education and clinical experience into actions (the ‘knowing-doing’ gap) [31–33]. In addition, the screening on potentially inappropriate prescribing may be used to select patients who may benefit from a full medication review considering the number and type of signals. A disadvantage of implementing the investigated tools as clinical decision support in electronic health systems is that they will likely result in false-positive signals in approximately 60% of cases, posing a risk of alert fatigue. In contrast, an observational study on regular drug safety alerts, including drug-drug interactions, overdosing and double medications, found that 91% of signals were overridden by prescribers [34]. Thus, a predictive value of the ADR trigger tool and STOPP/START signals at approximately 40% may be acceptable. However, notably, the acceptance of STOPP/START signals by a pharmacotherapy team was performed by physician-pharmacist pairs specialised in medication optimisation in older people, and the actual implementation of STOPP/STARTbased recommended actions was lower (Chapter 3.2). Therefore, it is important to consider the context in which such signals will be presented, with special attention to the healthcare setting and the healthcare professional receiving such signals. For instance, a recent study investigated the impact of CDSS-assisted alerts to discontinue benzodiazepines in patients ≥ 65 years when integrated into primary care EHRs. Prescribers ignored or overrode over 99% of alerts. The authors concluded that a CDSS-generated signal to alert for benzodiazepine use is insufficient to create behaviour change in clinicians [35]. Thus, better support in handling these signals may be more successful in turning detection into action.
RkJQdWJsaXNoZXIy MTk4NDMw