Reducing false alarm rates for critical arrhythmias using the arterial blood pressure waveform.
Reducing false alarm rates for critical arrhythmias using the arterial blood pressure waveform.
复制标题
DOI：
10.1016/j.jbi.2008.03.003
复制
发表时间：
2008-06-01
影响因子：
4.5
通讯作者：
Clifford, Gari D.
中科院分区：
文献类型：
--
作者：
Aboukhalil, Anton;Nielsen, Larry;Clifford, Gari D.
背景在过去的二十年里，高误报率（FA）仍然是重症监护室（ICU）中一个重要但尚未解决的问题。高FA率导致主治人员对此类警告的脱敏，伴随着响应时间的相关减慢和对患者的护理质量的有害降低。错误的心律失常警报通常是由于单通道ECG伪影和低电压信号，因此，如果使用来自其他独立信号的信息来形成警报病因学的更稳健假设，则可能会降低FA率。使用一个大型多参数ICU数据库（PhysioNet的MIMIC II数据库）研究了由市售ICU监测系统产生的五类假临界（“红色”或“危及生命”）ECG心律失常警报的频率，即：心搏停止、极度心动过缓、极度心动过速、室性心动过速和室颤/心动过速。本研究未考虑非关键（“黄色”）心律失常报警。使用相关的41，301 h同步ECG和动脉血压（ABP）波形，对MIMIC II数据库中总计447例成人患者记录的5386例严重ECG心律失常报警进行了多位专家评审。然后测试了一种使用来自ABP信号的形态学和时序信息来抑制虚假临界ECG心律失常警报的算法。平均42.7%的严重ECG心律失常警报被发现是假的，五个警报类别中的每一个的FA率在23.1%和90.7%之间。FA抑制算法能够抑制59.7%的假警报，心搏停止的FA减少率高达93.5%，极端心动过缓的FA减少率高达81.0%。极端心动过速（63.7%）和心室相关报警（室颤/心动过速为58.2%，室性心动过速为33.0%）的FA降低率最低。除室性心动过速报警（9.4%）外，真实报警（TA）减少率均为0%。FA抑制算法将错误的严重ECG心律失常警报的发生率从42.7%降低到17.2%，其中同时提供ECG和ABP数据。本算法证明了数据融合在临床环境中减少错误ECG心律失常警报的潜力，但室性心动过速的非零TA减少率表明需要进一步完善抑制策略。为了避免抑制任何真实报警，该算法可用于除室性心动过速以外的所有报警。在这些条件下，FA率将从42.7%降至22.7%。前瞻性临床评价应考虑该算法的实现。多参数生理波形的真实ICU数据库及其相关注释警报的公开可用性对于算法开发人员来说是一种新的有价值的研究资源。C 2008 Elsevier Inc. All rights reserved.
Background Over the past two decades, high false alarm (FA) rates have remained an important yet unresolved concern in the Intensive Care Unit (ICU). High FA rates lead to desensitization of the attending staff to such warnings, with associated slowing in response times and detrimental decreases in the quality of care for the patient. False arrhythmia alarms are commonly due to single channel ECG artifacts and low voltage signals, and therefore it is likely that the FA rates may be reduced if information from other independent signals is used to form a more robust hypothesis of the alarm's etiology.Methods. A large multi-parameter ICU database (PhysioNet's MIMIC II database) was used to investigate the frequency of five categories of false critical ("red" or "life-threatening") ECG arrhythmia alarms produced by a commercial ICU monitoring system, namely: asystole, extreme bradycardia, extreme tachycardia, ventricular tachycardia and ventricular fibrillation/tachycardia. Non-critical ("yellow") arrhythmia alarms were not considered in this study. Multiple expert reviews of 5386 critical ECG arrhythmia alarms from a total of 447 adult patient records in the MIMIC II database were made using the associated 41,301 h of simultaneous ECG and arterial blood pressure (ABP) waveforms. An algorithm to suppress false critical ECG arrhythmia alarms using morphological and timing information derived from the ABP signal was then tested.Results. An average of 42.7% of the critical ECG arrhythmia alarms were found to be false, with each of the five alarm categories having FA rates between 23.1% and 90.7%. The FA suppression algorithm was able to suppress 59.7% of the false alarms, with FA reduction rates as high as 93.5% for asystole and 81.0% for extreme bradycardia. FA reduction rates were lowest for extreme tachycardia (63.7%) and ventricular- related alarms (58.2% for ventricular fibrillation/tachycardia and 33.0% for ventricular tachycardia). True alarm (TA) reduction rates were all 0%, except for ventricular tachycardia alarms (9.4%).Conclusions. The FA suppression algorithm reduced the incidence of false critical ECG arrhythmia alarms from 42.7% to 17.2%, where simultaneous ECG and ABP data were available. The present algorithm demonstrated the potential of data fusion to reduce false ECG arrhythmia alarms in a clinical setting, but the non-zero TA reduction rate for ventricular tachycardia indicates the need for further refinement of the suppression strategy. To avoid suppressing any true alarms, the algorithm could be implemented for all alarms except ventricular tachycardia. Under these conditions the FA rate would be reduced from 42.7% to 22.7%. This implementation of the algorithm should be considered for prospective clinical evaluation. The public availability of a real-world ICU database of multi-parameter physiologic waveforms, together with their associated annotated alarms is a new and valuable research resource for algorithm developers. C 2008 Elsevier Inc. All rights reserved.