Bastiaan Sallevelt

128 CHAPTER 2.3 clinical judgement, as no inter-criterion priority is predefined when multiple criteria are relevant to one patient. Application of the algorithms to real patients should reveal whether false positive triggers remain an issue, potentially causing alert fatigue [29], despite the addition of optional excluding conditions to minimize this. Therefore, actual validation of the complete set of algorithms together in one patient, preferably in a clinical trial setting, will be an important next step. Finally, we would like to emphasize that STOPP/START criteria are developed as a screening tool for potentially inappropriate prescribing, not an absolute guiding principle. Clinical judgement determining the applicability of the criteria for individual patients will remain indispensable. Our algorithms should be utilized as an extension of this principle. Focus for future research As concluded previously by Anrys et al. [15], many criteria within STOPP/START version 2 lack sufficient explicitness for translation into coded algorithms. By setting rules for universal coding and using multiple rounds of consensus and validation, we have attempted to overcome this problem. Unfortunately, this led to a simplification of certain criteria, as some parts are just not convertible into codes. For the development of STOPP/START version 3 or other sets of explicit criteria, we advise the developers to be as clear and unequivocal as possible. This includes mentioning clear cut-off values or numbers instead of ‘hyperkalemia’ or ‘recurrent episodes’ and avoid ambiguous wordings such as ‘first-line’, ‘long-term’, ‘radiological evidence’ and ‘continuing provoking risk factors’. With the growing digitalisation of medical practice, future guidelines and explicit screening tools should complement and facilitate the possibility for software applications.