Proposed new requirements for testing and reporting performance results of arrhythmia detection algorithms

Several arrhythmia performance measures as specified in the current AAMI recommended practice document do not adequately reflect actual clinical experience in real-time patient monitoring. Additional reporting requirements that are more clinical relevant are thus needed: 1) Due to the large number of QRS complexes that need to be analyzed, even an algorithm with high specificity will generate a large number of false positives, which is not directly reflected by the reported false positive rate. A new recommendation is to report the actual number of false positives. 2) Due to the low PVC prevalence in most monitored patients, the high positive predictive values (PPV) reported using databases with high PVC prevalence are not clinically relevant. A proposed recommendation is to report PPVs at much lower PVC rates. 3) Arrhythmia performance measures as specified in the recommended practice do not include performance bounds. A new proposal is to use the bootstrap method with sample replacement to generate the mean values and 95% confident intervals for all the reported arrhythmia performance measures. Conclusions: Additional performance measures are proposed to further improve the clinical relevance of the reported results. These measures should be considered for inclusion as part of the standard reporting.