Bastiaan Sallevelt

116 CHAPTER 2.3 categorized per STOPP/START criterion. The experts were asked to review all the assigned codes as well as the interpretation of the criteria by the preparation panel. A teleconference meeting was organized to discuss the suggested modifications by the expert panel and to reach consensus. During this meeting suggestions to in- and exclude certain ATC-codes (e.g. specifying DMARDs, anticholinergics, high potency opioids) and ICPC/ICD-codes were discussed per STOPP/START criterion, based on clinical guidelines, scientific literature and the (clinical) expertise of the panelists. Third round During the teleconference meeting, discussion between the panelists elucidated the ambiguity of some criteria leading to different interpretations of STOPP/START recommendations and consequent choices regarding the codes (both ICD/ICPC and ATC) to be included in the algorithms. To improve the inter-rater reliability, a set of basic principles for coding the algorithms (Table 1) was deemed necessary. Table 1. Coding principles defined during the third round. 1 We intend to follow the original criteria as closely as possible. If criteria require additional specification in order to be encoded, this is conducted without essentially altering the content of the criterion. 2 We assume the availability of recent laboratory values or measurements and prioritise these values over ICD or ICPC codes. If condition (1) is not satisfied, condition (2) will be evaluated for availability. 3 a. If medication is specified as a class where an exact specification of the included medications within this class (i.e.) is mentioned, only those drugs are included (ATC 5 level). b. If medication is specified as a class on ATC 3 or 4 level, where no or some examples (e.g.) are mentioned, the most important medications within this class are included according to expert consensus. 4 Some medical conditions can contain several underlying diagnoses that are not specifically mentioned. Therefore, the most common and/or most important diagnoses will be included based on consensus within the expert panel. 5 In order to minimize false positive triggers in the practical application of our algorithms, we will add optional conditions to the criteria incorporating common (lack of) indications for certain medications and diseases (that are not actually present in the original criteria).