- Email: [email protected]
Fully automatic rhythm analysis during chest compression pauses
Resuscitation 89 (2015) 25–30
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Clinical Paper
Fully automatic rhythm analysis during chest compression pauses夽 U. Ayala a,∗ , U. Irusta a , J. Ruiz a , S. Ruiz de Gauna a , D. González-Otero a , E. Alonso a , J. Kramer-Johansen b , H. Naas b , T. Eftestøl c a
Communications Engineering Department, University of the Basque Country UPV/EHU, Alameda Urquijo S/N, 48013 Bilbao, Spain National Advisory Unit for Prehospital Emergency Medicine (NAKOS) and Department of Anaesthesiology, Oslo University Hospital and University of Oslo, P.O. Box 4956 Nydalen, N-0424 Oslo, Norway c Department of Electrical Engineering and Computer Science, Faculty of Science and Technology, University of Stavanger, 4036 Stavanger, Norway b
a r t i c l e
i n f o
Article history: Received 18 July 2014 Received in revised form 7 November 2014 Accepted 18 November 2014 Keywords: Cardiopulmonary resuscitation (CPR) Chest compression Cardiac arrest Automated external defibrillator (AED) Transthoracic impedance
a b s t r a c t Aim: Chest compression artefacts impede a reliable rhythm analysis during cardiopulmonary resuscitation (CPR). These artefacts are not present during ventilations in 30:2 CPR. The aim of this study is to prove that a fully automatic method for rhythm analysis during ventilation pauses in 30:2 CPR is reliable an accurate. Methods: For this study 1414 min of 30:2 CPR from 135 out-of-hospital cardiac arrest cases were analysed. The data contained 1942 pauses in compressions longer than 3.5 s. An automatic pause detector identified the pauses using the transthoracic impedance, and a shock advice algorithm (SAA) diagnosed the rhythm during the detected pauses. The SAA analysed 3-s of the ECG during each pause for an accurate shock/no-shock decision. Results: The sensitivity and PPV of the pause detector were 93.5% and 97.3%, respectively. The sensitivity and specificity of the SAA in the detected pauses were 93.8% (90% low CI, 90.0%) and 95.9% (90% low CI, 94.7%), respectively. Using the method, shocks would have been advanced in 97% of occasions. For patients in nonshockable rhythms, rhythm reassessment pauses would be avoided in 95.2% (95% CI, 91.6–98.8) of occasions, thus increasing the overall chest compression fraction (CCF). Conclusion: An automatic method could be used to safely analyse the rhythm during ventilation pauses. This would contribute to an early detection of refibrillation, and to increase CCF in patients with nonshockable rhythms. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Early cardiopulmonary resuscitation (CPR) and early defibrillation are key factors for the survival of out-of-hospital cardiac arrest (OHCA) patients.1 The use of automated external defibrillators (AED) may shorten time to defibrillation. Current CPR guidelines recommend a 30:2 compression to ventilation (CV) ratio before tracheal intubation, with emphasis on delivering minimally interrupted high-quality chest compressions.2 Chest compressions should be interrupted only to assess the rhythm, to deliver the shock or to ventilate the patient, taking less than 5 s for the two rescue breaths.2 Interruptions in chest compressions are frequent during OHCA.3 These interruptions have a detrimental effect on shock success
夽 A Spanish translated version of the summary of this article appears as Appendix in the final online version at http://dx.doi.org/10.1016/j.resuscitation.2014.11.022. ∗ Corresponding author. E-mail address: [email protected] (U. Ayala). http://dx.doi.org/10.1016/j.resuscitation.2014.11.022 0300-9572/© 2014 Elsevier Ireland Ltd. All rights reserved.
and patient survival.4,5 More importantly, pre-, post- and perishock pauses all have a negative effect on shock success.6–8 For instance, a 5 s increase in pre-shock or peri-shock pause duration may decrease the odds of survival by as much as 18% or 14%, respectively.8 Post-shock pauses are shortened by immediately resuming compressions after shock delivery. Strategies to shorten pre-shock pauses involve compressions during defibrillator charging,9,10 simplifying voice prompts,11 or immediately triggering rhythm analysis at the end of chest compressions.12 However, interruptions for rhythm analysis are necessary because chest compression artefacts may confound the diagnosis of the shock advice algorithm (SAA) in current AEDs.13 These interruptions contribute to the pre-shock pause, which lasts between 5.2 and 28.4 s 14 and may compromise the survival of the patient.6 Currently chest compressions are interrupted every 2 min to reassess the rhythm.2 An automatic and reliable method to analyse the rhythm during CPR would shorten or even eliminate the need for these interruptions. For patients with nonshockable rhythms uninterrupted CPR could be prolonged beyond the recommended
26
U. Ayala et al. / Resuscitation 89 (2015) 25–30
2-min cycles,2,15 thus increasing the chest compression fraction (CCF). In addition, ventricular fibrillation (VF) could be detected earlier and shocks advanced. All these actions may have a significant positive impact on outcome.16,17 Over the years, considerable effort has been dedicated to develop methods for rhythm analysis during CPR.13 Several approaches have been studied such as adaptive artefact filters 18–20 or new SAAs designed to analyse either the artefacted 21,22 or the filtered ECG.23 When tested on short strips (< 20 s) of OHCA data from defibrillators, only a recently published SAA23 produced sensitivity and specificity results above 90% and 95%, the performance goals recommended by the American Heart Association (AHA).24 Nevertheless, the reported sensitivities and specificities were well below those obtained for SAAs on artefactfree ECG.25 Furthermore, the algorithms were not tested in long strips (minutes) of defibrillator data, i.e. when the method is used to continuously assess the rhythm during CPR. This latter type of testing allows the statistical evaluation of the impact of using the method on the delivery of CPR,26 as suggested by the 2010 consensus on resuscitation science.27 Basic Life Support (BLS) guidelines recommend 30:2 CV-ratio CPR. At the standard rate of 100 compressions per minute, pauses for two ventilations occur approximately every 20 s. Furthermore, in BLS scenarios these pauses normally last longer than 5 s.28 Recently, Ruiz et al 29 proposed doing rhythm analysis during these pauses because the ECG would not have chest compression artefacts. This would allow an accurate rhythm assessment every 20 s during CPR. The objective of this study is to prove the reliability of a fully automatic method to analyse the rhythm during ventilation pauses, and to evaluate the method’s impact on the delivery of CPR.
2. Materials and methods 2.1. Data collection Data from 370 OHCA cases from adult patients were collected between July 2012 and June 2013 in the city of Oslo (Norway). From the 370 cases 135 qualified for inclusion in this study, as explained at the end of this section. In all cases, CPR was performed by Advanced Life Support (ALS) responders. Lifepak 12/15 defibrillators were used (Physio-Control, Redmond, WA, USA), and ECG and thoracic impedance (TI) acquired from the defibrillation pads were stored. Within each case, the initial rhythm and each subsequent change in rhythm were annotated by consensus between a clinical researcher (HN) under the supervision of an experienced anaesthesiologist (JKJ), and a biomedical engineer (UI), all specialised in resuscitation. The following rhythm annotations were used: VF and ventricular tachycardia in the shockable category, and asystole (ASY) and organised (ORG) rhythm in the nonshockable category. Intermediate rhythms such as fine VF were annotated as undecided. These are the rhythms for which, according to the AHA statement,24 the benefits of defibrillation are unclear. Chest compressions were automatically detected by the Code-Stat 9.0 review software (Physio-Control). Periods longer than 1.5 s between consecutive chest compressions were defined as a pause.30 The chest compression marks in the vicinity of the pauses were visually inspected, and manually corrected when necessary based on the TI and ECG signals. The resulting annotated pauses were used as gold standard for the automatic detection of pauses. The criterion to include a case in this study was that the case contained at least 4 min of CPR delivered with 30:2 CV-ratio. Frequently the CPR pattern changed within a case, for instance from 30:2 CPR to continuous chest compressions after intubation. Consequently, for each case intervals in which 30:2 CPR was administered were identified. These 30:2 CPR intervals had to have a duration above
1 min and at least two pauses every minute of CPR. The case was included in the study when the aggregate duration of its 30:2 CPR intervals was more than 4 min. In this study only the 30:2 CPR intervals within each case were analysed. The top panels in Fig 1 show a 2-min interval with six pauses for ventilation. 2.2. Fully automatic rhythm analysis method The reliability of an automatic pause detector based on the TI was evaluated. The detector was used in combination with the automated rhythm analysis of a SAA capable of analysing the rhythm using only a few seconds of the ECG. If a pause was detected and its duration was long enough for a rhythm analysis, the SAA was launched for a shock/no-shock decision. 2.2.1. Shock advice algorithm The SAA is an AHA compliant algorithm that analyses 3 s of the ECG for an accurate shock/no-shock diagnosis,29 a detailed description of the algorithm is available.31 Briefly, the SAA is composed of an asystole detector, a QRS presence detector (nonshockable rhythms) and a final shock/no-shock algorithm based on the regularity, spectral distribution and heart-rate of the rhythm. When tested on an AHA compliant database the SAA presented a sensitivity of 98.5% for shockable rhythms, and a specificity of 99.1% and 100% for ORG and ASY rhythms, respectively.29 The SAA is a modified Matlab version (MathWorks Inc., Natick, MA) of the one incorporated in the Reanibex R-series defibrillators (Bexen Cardio, Ermua, Spain). 2.2.2. Pause detection and rhythm analysis The SAA used in this study analyses 3 s of the ECG to make a decision. Consequently, for this study pauses longer than 3.5 s were defined as diagnosable-pauses, i.e. pauses long enough for the SAA to diagnose the rhythm. The initial 0.5 s of the pause were discarded to avoid transients and artefact residuals in the pause. Diagnosablepauses were automatically detected following these steps (see Fig 1 for a visual description): 1 Pause identification: The envelope of the TI signal was obtained by applying standard signal processing techniques 32 (step 1 of Fig 1). The TI-envelope is large during compressions and vanishes during the pauses. Pauses were automatically detected in the TI-envelope using an adaptive threshold. Adaptivity is necessary because the amplitude of the TI-envelope may vary substantially within each case. 2 Pause onset/offset detection: A chest compression detector was applied in the vicinity of the detected pause to accurately determine the onset/offset of the pause (step 2 of Fig 1). 3 Rhythm analysis: If the pause was longer than 3.5 s, the SAA was automatically launched and the rhythm was analysed 0.5 s after the detected pause onset (step 3 of Fig 1). 2.3. Performance evaluation The method was evaluated in terms of: (1) accuracy of the pause detector, (2) accuracy of the SAA in the detected pauses and (3) the potential therapeutic benefits of using the method. 2.3.1. Reliability of the pause detector Detected diagnosable-pauses were compared to the diagnosable-pauses obtained from the gold standard. Pause detection sensitivity was defined as the proportion of correctly detected diagnosable-pauses, and the positive predictive value (PPV) as the proportion of detected diagnosable-pauses that were true diagnosable-pauses. Sensitivity and PPV values were calculated for the whole dataset and for each case. The pause onset
U. Ayala et al. / Resuscitation 89 (2015) 25–30
27
Fig. 1. Processing steps for automatic pause detection and rhythm analysis. The figure shows on top the ECG and impedance during a 30:2 CPR interval of 2 min duration. In this 30:2 CPR interval there were 6 pauses for two ventilations (slow impedance fluctuations during the pause). In the bottom panels the processing steps are highlighted around a diagnosable-pause. In step 1 the envelope of the impedance (thick grey line) reflects chest compression activity, so pauses are detected by adaptively identifying when this envelope vanishes. In step 2 chest compressions (labelled with dots) are detected in the vicinity of the pause to accurately mark pause onset/offset. In step 3 the SAA is launched 0.5 s after pause onset, and the rhythm is diagnosed as nonshockable (NS) analysing a 3 s ECG segment.
detection accuracy was determined by comparing the detected onset time with that of the last compression annotation before the pause. 2.3.2. Accuracy of the rhythm analysis The SAA was launched in each automatically detected pause, and its diagnosis was compared to the rhythm annotation in the pause. The rhythm analysis accuracy was measured in terms of the proportion of correct shock (sensitivity) and no-shock (specificity) diagnoses. 2.3.3. Potential therapeutic benefits There are two potential benefits of using this method. For patients in shockable rhythms refibrillation could be detected earlier, and the shock advanced. For patients in nonshockable rhythms CPR would not be stopped after 2 min for rhythm assessment, but would rather continue uninterrupted until a change in rhythm
occurs. To measure these potential benefits, the hypothetical effect on CPR delivery of following the automated diagnoses given by the method was quantified. The results depend on when the rescuer would have stopped CPR using the feedback from the method. Two stop-CPR conditions were analysed: stop-CPR after the first shock diagnosis (regular), and stop-CPR after two consecutive shock diagnoses (conservative). For patients in shockable rhythms the proportion of cases in which the shock would have been advanced was analysed. For this evaluation segments were selected within the analysis intervals meeting these conditions: the rhythm was shockable, the interval ended in a defibrillation attempt, and it had at least two diagnosable-pauses. For patients with nonshockable rhythms how the method would prolong CPR cycles beyond the currently recommended 2 min was quantified. Kaplan–Meier survival curves were used to estimate the probability of prolonging CPR cycles as a function of their duration.26 Segments of at least 2 min in which the
28
U. Ayala et al. / Resuscitation 89 (2015) 25–30
rhythm was nonshockable were selected, and for each segment the time at which the stop-CPR condition occurred determined the duration of the CPR cycle.
Table 1 Positive predictive value of the pause detection method, and sensitivity/specificity (90% one-sided lower CI in parenthesis) of the SAA in the detected diagnosablepauses. Rhythm type
2.4. Statistical analysis The statistical distributions of the database’s descriptive data did not pass the one-sample Kolmogorov–Smirnov normality test and are presented as median and interquartile range (IQR). The pause detector’s sensitivity and PPV per case are presented as mean, their 95% confidence intervals (CI) were calculated using a bootstrap procedure with 1000 replications and the basic bootstrap interval. Consecutive rhythm analysis diagnoses within an interval are correlated, so sensitivity and specificity and their 90% one-sided lower CI were adjusted for clustering within the intervals using generalised estimating equations within a logistic regression model.33 The calculations were done using the Geepack library 34 in the R statistical software. For nonshockable segments, Kaplan–Meier survival curves were computed, and their 95% CI were calculated using Greenwood’s variance. Segments in which the stop-CPR condition did not occur were regarded as censored observations.
Shockable Nonshockable AS ORG
Detected pauses n
PPV (%)
SAA Se/Sp (%)
291 1599 1032 567
97.9 97.2 97.0 97.7
93.8 (90.0) 95.9 (94.7) 96.2 (94.7) 95.5 (93.2)
3.2. Reliability of the pause detector The overall sensitivity and PPV of the pause detector were 93.5% and 97.3%, respectively. The mean sensitivity and PPV per case were 93.5% (95% CI, 92.0–95.0) and 96.6% (95% CI, 95.7–97.5), respectively. The median error in pause onset identification was 0.0 s (5–95 percentile, −0.1 to 0.6). Panel a of Fig 2 shows an example of an accurate pause detection, while panels b and c expose the challenges of pause detection using the TI.
3.3. Accuracy of the rhythm analysis 3. Results 3.1. Database summary There were a total of 229 30:2 CPR intervals within the 135 cases, comprising 1414 min. The median duration of the intervals was 4.8 min (IQR, 3.0–7.5), and a median of 9 min (IQR, 5.9–12.4) were analysed per case. There were 1942 diagnosable-pauses in the analysed data, with 11 (IQR, 6–18) pauses per case. The median duration of the diagnosable-pauses was 4.7 s (IQR, 4.0–6.5).
The SAA was launched in the 1917 diagnosable-pauses that were automatically detected, in 27 of those the rhythm was annotated as undecided and were therefore discarded for the evaluation of the sensitivity and specificity. The sensitivity and specificity results are shown in Table 1. The overall sensitivity and specificity were 93.8% (90% one-sided lower CI, 90.0%) and 95.9% (90% one-sided lower CI, 94.7%), respectively. Fig 3 shows three examples of rhythm analysis. In panels a and b the rhythm was correctly diagnosed, while in panel c an erroneous shock was advised due to electrode/cord movement noise in the ECG.
Fig. 2. Examples of correct and missed chest compression pause detection. In (b) noise in the impedance caused the detected pause to be too short (<3.5 s) to launch an analysis. In (c) very low fluctuations in TI caused by chest compressions were not detected and pause onset was not accurately determined resulting in a false positive pause detection (pause in an artefacted ECG segment).
Fig. 3. Examples of correct and erroneous rhythm analysis during ventilation pauses. The diagnosis given by the SAA is shown within the pause. In (c) the first pause was annotated as nonshockable (NS) but was erroneously classified as shockable (S) by the SAA.
U. Ayala et al. / Resuscitation 89 (2015) 25–30
Fig. 4. Kaplan–Meier survival curves with 95% confidence intervals for nonshockable cases. The curves represent the probability of prolonging the CPR cycle as a function of the duration of the cycle. The cut-off point at 2 min gives the probability of increasing CCF compared to current guidelines.
3.4. Potential therapeutic benefits A total of 35 segments with at least two diagnosablepauses before defibrillation, and 168 segments with nonshockable rhythms were identified. The median duration of the shockable and nonshockable segments were 2.6 min (IQR, 2.0–3.3) and 4.9 min (IQR, 3.6–7.5), respectively. With the regular criterion for stopCPR all shocks would be advanced, 91.4% (32/35) of cases to the first diagnosable-pause and the remaining 8.6% to the second diagnosable-pause. As shown in Fig 4, the 2-min CPR cycles would be prolonged in 86.3% (81.1–91.5) of nonshockable cases. With the conservative criterion 97% (34/35) of shocks would be advanced, and the 2-min CPR cycles would be prolonged in 97.6% (95.3–99.9) of the nonshockable cases. Moreover, uninterrupted CPR cycles of at least 4 min would occur in 95.2% (91.6–98.8) of nonshockable cases, substantially increasing the CCF. 4. Discussion In the past, methods for rhythm analysis during CPR have followed or have combined two main lines of attack: the removal of the chest compression artefacts,18–20 and special purpose rhythm analysis methods marginally affected by the artefacts.21–23 Generally, the specificity of these methods is below the 95% value recommended by the AHA.13 Furthermore, when used to continuously monitor the rhythm during CPR their performance degrades substantially.26 This study demonstrates the reliability of a fully automatic method that avoids chest compression artefacts altogether by analysing the rhythm during the ventilation pauses in 30:2 CV-ratio CPR (a video in the supplementary material shows a real-time demonstration of the method). This method would eliminate the need to stop CPR for rhythm analysis, and could therefore serve to advance the shock when patients refibrillate and to prolong the CPR cycle in patients with nonshockable rhythms. The method could be used in any circumstance in which pauses in chest compressions are frequent and regular. However, it was primarily conceived to be used in a Basic Life Support (BLS) scenario, in which the 30:2 CV-ratio is recommended.2 The method was therefore designed to work with the only data available in an AED scenario, the ECG and the TI signals recorded through the defibrillation pads. TI has been effectively used to detect chest compressions,35,36 ventilations,37 pulse,38 and to monitor CPR quality parameters like compression rate or CCF.36 However,
29
TI based chest compression detectors fail to accurately determine short pauses in chest compressions.36 Our method successfully overcomes this limitation by combining an adaptive TI-envelope detector to identify potential pauses with a compression detector to accurately determine pause onset and offset. This resulted in a very reliable detector that identified more than 93% of pauses, with very few false positives, the PPV was above 97%. Avoiding false positives is paramount for the success of the method because it guarantees rhythm analysis is done on an ECG free of chest compression artefacts. Moreover, our testing conditions were very challenging because pauses were very short (median duration 4.7 s). Indeed, pauses for two ventilations in CPR performed by laypersons may last around 7 s (IQR, 6–9),28 in those circumstances the accuracy and reliability of the detector would further improve. Accurate pause onset detection is very important in this application. This is a challenging task because compression induced fluctuations in TI change a great deal within an case, and even adaptive compression detectors fail to accurately determine pause onset and offset.36 However, applying a compression detector only locally around detected pauses results in a very accurate pause onset detection with a median delay of 0 s. Consequently, pauses as short as 3.5 s could be safely used to diagnose the rhythm, using a SAA capable of accurately analysing the ECG in such short ECG intervals. The resulting sensitivity and specificity were above AHA performance goals, and the accuracy for the fully automatic method was very similar to the one obtained in a previous study in which the analysis pauses were manually annotated.29 In this study the method was also used to analyse long OHCA segments, in which the rhythm was monitored during ventilation pauses. For a conservative stop-CPR criterion defibrillation would have been advanced in 97% of cases. Earlier recognition of recurrent VF could have a beneficial impact on outcome given the high oxygen demands of recurrent VF,17 and that time in recurrent VF is associated with worse outcomes.39 Furthermore, considering that ventilation pauses occur approximately every 20 s in 30:2 CPR, the coarseness of VF could be monitored to optimise defibrillation time.40 Our results show that a conservative stop-CPR criterion is needed to safely increase CCF during nonshockable rhythms. Uninterrupted CPR cycles could be prolonged beyond 4 min in over 95% of cases, avoiding many rhythm reassessment pauses. This would substantially increase CCF in patients with nonshockable rhythms, which could increase the likelihood of ROSC.16 Decisions that may improve survival such as an increased duration of the CPR cycles during asystole,15 or treatment recommendations tailored to the type of nonshockable rhythm could be incorporated to the AED. In summary, adding a rhythm monitor function during CPR could turn current AEDs into intelligent devices that guide the rescuer to follow optimal treatment decisions. In any case, feedback to the rescuer would only be given when action is needed, such as to stop CPR to deliver the shock when refibrillation is detected. 4.1. Limitations In our dataset CPR was performed by highly trained ALS responders, which resulted in very short pauses for two ventilations. The method is primarily intended for a BLS scenario with lay responders. In these cases the expected pause duration would be longer. This should most likely increase the accuracy of the method since the testing conditions would be more favourable. 5. Conclusions This study proved that a fully automatic method could be used to safely analyse the rhythm during ventilation pauses in 30:2 CV-ratio CPR. This would contribute to an early detection of
30
U. Ayala et al. / Resuscitation 89 (2015) 25–30
refibrillation and to an increase in CCF in patients with nonshockable rhythms. Conflict of interest statement Authors U. Irusta, J. Ruiz and S. Ruiz de Gauna have received research support from Bexen Cardio (Ermua, Spain) for studies on AED shock advice algorithms. Acknowledgements This work received financial support from the Ministerio de Ciencia e Innovación of Spain through the projects TEC2012-31928 and TEC2012-31144, from the University of the Basque Country (UPV/EHU) through the unit UFI11/16 and from the Programa de Formación de Personal Investigador del Departamento de Educación, Universidades e Investigación del Gobierno Vasco through the grants BFI-2010-174, BFI-2011-166 and BFI-2010-235. Appendix A. Supplementary Data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.resuscitation.2014.11.022. References 1. Nolan J, Soar J, Eikeland H. The chain of survival. Resuscitation 2006;71(3):270–1. 2. Koster RW, Baubin MA, Bossaert LL, et al. European Resuscitation Council Guidelines for Resuscitation 2010 Section 2. Adult basic life support and use of automated external defibrillators. Resuscitation 2010;81(10):1277–92. 3. Kramer-Johansen J, Myklebust H, Wik L, et al. Quality of out-of-hospital cardiopulmonary resuscitation with real time automated feedback: a prospective interventional study. Resuscitation 2006;71:283–92. 4. Berg R, Sanders A, Kern K, et al. Adverse hemodynamic effects of interrupting chest compressions for rescue breathing during cardiopulmonary resuscitation for ventricular fibrillation cardiac arrest. Circulation 2001;104(20):2465–70. 5. Eftestøl T, Sunde K, Steen P. Effects of interrupting precordial compressions on the calculated probability of defibrillation success during out-of-hospital cardiac arrest. Circulation 2002;105(19):2270–3. 6. Edelson DP, Abella BS, Kramer-Johansen J, et al. Effects of compression depth and pre-shock pauses predict defibrillation failure during cardiac arrest. Resuscitation 2006;71:137–45. 7. Mader T, Paquette A, Salcido D, Nathanson B, Menegazzi J. The effect of the preshock pause on coronary perfusion pressure decay and rescue shock outcome in porcine ventricular fibrillation. Prehosp Emerg Care 2009;13(4):487–94. 8. Cheskes S, Schmicker R, Christenson J, et al. Perishock pause: an independent predictor of survival from out-of-hospital shockable cardiac arrest. Circulation 2011;124(1):58–66. 9. Jost D, Degrange H, Verret C, et al. DEFI 2005: a randomized controlled trial of the effect of automated external defibrillator cardiopulmonary resuscitation protocol on outcome from out-of-hospital cardiac arrest. Circulation 2010;121(14):1614–22. 10. Cheskes S, Common MR, Byers PA, Zhan C, Morrison LJ. Compressions during defibrillator charging shortens shock pause duration and improves chest compression fraction during shockable out of hospital cardiac arrest. Resuscitation 2014;85(8):1007–11. 11. Rhee JE, Kim T, Kim K, Choi S. Is there any room for shortening hands-off time further when using an AED? Resuscitation 2009;80(2):231–7. 12. Didon JP, Krasteva V, Ménétré S, Stoyanov T, Jekova I. Shock advisory system with minimal delay triggering after end of chest compressions: accuracy and gained hands-off time. Resuscitation 2011;82(Suppl. 2):S8–15. 13. Ruiz de Gauna S, Irusta U, Ruiz J, Ayala U, Aramendi E, Eftestøl T. Rhythm analysis during cardiopulmonary resuscitation: past, present, and future. BioMed Res Int 2014;2014. Article ID 386010. 14. Snyder D, Morgan C. Wide variation in cardiopulmonary resuscitation interruption intervals among commercially available automated external defibrillators may affect survival despite high defibrillation efficacy. Crit Care Med 2004;32(9 Suppl.):S421–4. 15. Nordseth T, Edelson DP, Bergum D, et al. Optimal loop duration during the provision of in-hospital advanced life support (ALS) to patients with an initial non-shockable rhythm. Resuscitation 2014;85(1):75–81.
16. Vaillancourt C, Everson-Stewart S, Christenson J, et al. The impact of increased chest compression fraction on return of spontaneous circulation for out-ofhospital cardiac arrest patients not in ventricular fibrillation. Resuscitation 2011;82(12):1501–7. 17. Hoogendijk MG, Schumacher CA, Belterman CN, et al. Ventricular fibrillation hampers the restoration of creatine-phosphate levels during simulated cardiopulmonary resuscitations. Europace 2012;14(10):1518–23. 18. Eilevstjønn J, Eftestøl T, Aase SO, Myklebust H, Husøy JH, Steen PA. Feasibility of shock advice analysis during CPR through removal of CPR artefacts from the human ECG. Resuscitation 2004;61(2):131–41. 19. Werther T, Klotz A, Granegger M, et al. Strong corruption of electrocardiograms caused by cardiopulmonary resuscitation reduces efficiency of two-channel methods for removing motion artefacts in non-shockable rhythms. Resuscitation 2009;80(11):1301–7. 20. Aramendi E, Ayala U, Irusta U, Alonso E, Eftestøl T, Kramer-Johansen J. Suppression of the cardiopulmonary resuscitation artefacts using the instantaneous chest compression rate extracted from the thoracic impedance. Resuscitation 2012;83(6):692–8. 21. Li Y, Bisera J, Geheb F, Tang W, Weil MH. Identifying potentially shockable rhythms without interrupting cardiopulmonary resuscitation. Crit Care Med 2008;36(1):198–203. 22. Krasteva V, Jekova I, Dotsinsky I, Didon JP. Shock advisory system for heart rhythm analysis during cardiopulmonary resuscitation using a single ECG input of automated external defibrillators. Ann Biomed Eng 2010;38(4):1326–36. 23. Ayala U, Irusta U, Ruiz J, et al. A reliable method for rhythm analysis during cardiopulmonary resuscitation. BioMed Res Int 2014;2014. Article ID 872470. 24. Kerber RE, Becker LB, Bourland JD, et al. Automatic external defibrillators for public access defibrillation: recommendations for specifying and reporting arrhythmia analysis algorithm performance, incorporating new waveforms, and enhancing safety: a statement for health professionals from the American heart association task force on automatic external defibrillation, subcommittee on AED safety and efficacy. Circulation 1997;95(6):1677–82. 25. Atkins DL, Scott WA, Blaufox AD, et al. Sensitivity and specificity of an automated external defibrillator algorithm designed for pediatric patients. Resuscitation 2008;76(2):168–74. 26. Ruiz J, Ayala U, Ruiz de Gauna S, et al. Direct evaluation of the effect of filtering the chest compression artifacts on the uninterrupted cardiopulmonary resuscitation time. Am J Emerg Med 2013;31(6):910–5. 27. Sayre MR, Koster RW, Botha M, et al. Part 5: Adult basic life support: 2010 International consensus on cardiopulmonary resuscitation and emergency cardiovascular care science with treatment recommendations. Circulation 2010;16(S2):S298–324. 28. Beesems SG, Wijmans L, Tijssen JGP, Koster RW. Duration of ventilations during cardiopulmonary resuscitation by lay rescuers and first responders: relationship between delivering chest compressions and outcomes. Circulation 2013;127(15):1585–90. 29. Ruiz J, Ayala U, Ruiz de Gauna S, et al. Feasibility of automated rhythm assessment in chest compression pauses during cardiopulmonary resuscitation. Resuscitation 2013;84(9):1223–8. 30. Kramer-Johansen J, Edelson DP, Losert H, Köhler K, Abella BS. Uniform reporting of measured quality of cardiopulmonary resuscitation (CPR). Resuscitation 2007;74(3):406–17. 31. Irusta U, Ruiz J, Aramendi E, Ruiz de Gauna S, Ayala U, Alonso E. A high-temporal resolution algorithm to discriminate shockable from nonshockable rhythms in adults and children. Resuscitation 2012;83(9):1090–7. 32. González-Otero D, de Gauna SR, Ruiz J, Ayala U, Alonso E. Automatic detection of chest compression pauses using the transthoracic impedance signal. In: Computing in Cardiology (CinC), 2012, IEEE. 2012. p. 21–4. 33. Sternberg MR, Hadgu A. A GEE approach to estimating sensitivity and specificity and coverage properties of the confidence intervals. Stat Med 2001;20(9–10):1529–39. 34. Halekoh U, Højsgaard S, Yan J. The R package geepack for generalized estimating equations. J Stat Softw 2006;15(2):1–11. 35. Stecher FS, Olsen JA, Stickney RE, Wik L. Transthoracic impedance used to evaluate performance of cardiopulmonary resuscitation during out of hospital cardiac arrest. Resuscitation 2008;79(3):432–7. 36. Ayala U, Eftestøl T, Alonso E, et al. Automatic detection of chest compressions for the assessment of CPR-quality parameters. Resuscitation 2014;85(7):957–63. 37. Risdal M, Aase SO, Stavland M, Eftestøl T. Impedance-based ventilation detection during cardiopulmonary resuscitation. IEEE Trans Biomed Eng 2007;54(12):2237–45. 38. Ruiz J, Alonso E, Aramendi E, et al. Reliable extraction of the circulation component in the thoracic impedance measured by defibrillation pads. Resuscitation 2013;84(10):1345–52. 39. Berdowski J, ten Haaf M, Tijssen JGP, Chapman FW, Koster RW. Time in recurrent ventricular fibrillation and survival after out-of-hospital cardiac arrest. Circulation 2010;122(11):1101–8. 40. Ristagno G, Li Y, Fumagalli F, Finzi A, Quan W. Amplitude spectrum area to guide resuscitation: a retrospective analysis during out-of-hospital cardiopulmonary resuscitation in 609 patients with ventricular fibrillation cardiac arrest. Resuscitation 2013;84(12):1697–703.
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Clinical Paper
Fully automatic rhythm analysis during chest compression pauses夽 U. Ayala a,∗ , U. Irusta a , J. Ruiz a , S. Ruiz de Gauna a , D. González-Otero a , E. Alonso a , J. Kramer-Johansen b , H. Naas b , T. Eftestøl c a
Communications Engineering Department, University of the Basque Country UPV/EHU, Alameda Urquijo S/N, 48013 Bilbao, Spain National Advisory Unit for Prehospital Emergency Medicine (NAKOS) and Department of Anaesthesiology, Oslo University Hospital and University of Oslo, P.O. Box 4956 Nydalen, N-0424 Oslo, Norway c Department of Electrical Engineering and Computer Science, Faculty of Science and Technology, University of Stavanger, 4036 Stavanger, Norway b
a r t i c l e
i n f o
Article history: Received 18 July 2014 Received in revised form 7 November 2014 Accepted 18 November 2014 Keywords: Cardiopulmonary resuscitation (CPR) Chest compression Cardiac arrest Automated external defibrillator (AED) Transthoracic impedance
a b s t r a c t Aim: Chest compression artefacts impede a reliable rhythm analysis during cardiopulmonary resuscitation (CPR). These artefacts are not present during ventilations in 30:2 CPR. The aim of this study is to prove that a fully automatic method for rhythm analysis during ventilation pauses in 30:2 CPR is reliable an accurate. Methods: For this study 1414 min of 30:2 CPR from 135 out-of-hospital cardiac arrest cases were analysed. The data contained 1942 pauses in compressions longer than 3.5 s. An automatic pause detector identified the pauses using the transthoracic impedance, and a shock advice algorithm (SAA) diagnosed the rhythm during the detected pauses. The SAA analysed 3-s of the ECG during each pause for an accurate shock/no-shock decision. Results: The sensitivity and PPV of the pause detector were 93.5% and 97.3%, respectively. The sensitivity and specificity of the SAA in the detected pauses were 93.8% (90% low CI, 90.0%) and 95.9% (90% low CI, 94.7%), respectively. Using the method, shocks would have been advanced in 97% of occasions. For patients in nonshockable rhythms, rhythm reassessment pauses would be avoided in 95.2% (95% CI, 91.6–98.8) of occasions, thus increasing the overall chest compression fraction (CCF). Conclusion: An automatic method could be used to safely analyse the rhythm during ventilation pauses. This would contribute to an early detection of refibrillation, and to increase CCF in patients with nonshockable rhythms. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Early cardiopulmonary resuscitation (CPR) and early defibrillation are key factors for the survival of out-of-hospital cardiac arrest (OHCA) patients.1 The use of automated external defibrillators (AED) may shorten time to defibrillation. Current CPR guidelines recommend a 30:2 compression to ventilation (CV) ratio before tracheal intubation, with emphasis on delivering minimally interrupted high-quality chest compressions.2 Chest compressions should be interrupted only to assess the rhythm, to deliver the shock or to ventilate the patient, taking less than 5 s for the two rescue breaths.2 Interruptions in chest compressions are frequent during OHCA.3 These interruptions have a detrimental effect on shock success
夽 A Spanish translated version of the summary of this article appears as Appendix in the final online version at http://dx.doi.org/10.1016/j.resuscitation.2014.11.022. ∗ Corresponding author. E-mail address: [email protected] (U. Ayala). http://dx.doi.org/10.1016/j.resuscitation.2014.11.022 0300-9572/© 2014 Elsevier Ireland Ltd. All rights reserved.
and patient survival.4,5 More importantly, pre-, post- and perishock pauses all have a negative effect on shock success.6–8 For instance, a 5 s increase in pre-shock or peri-shock pause duration may decrease the odds of survival by as much as 18% or 14%, respectively.8 Post-shock pauses are shortened by immediately resuming compressions after shock delivery. Strategies to shorten pre-shock pauses involve compressions during defibrillator charging,9,10 simplifying voice prompts,11 or immediately triggering rhythm analysis at the end of chest compressions.12 However, interruptions for rhythm analysis are necessary because chest compression artefacts may confound the diagnosis of the shock advice algorithm (SAA) in current AEDs.13 These interruptions contribute to the pre-shock pause, which lasts between 5.2 and 28.4 s 14 and may compromise the survival of the patient.6 Currently chest compressions are interrupted every 2 min to reassess the rhythm.2 An automatic and reliable method to analyse the rhythm during CPR would shorten or even eliminate the need for these interruptions. For patients with nonshockable rhythms uninterrupted CPR could be prolonged beyond the recommended
26
U. Ayala et al. / Resuscitation 89 (2015) 25–30
2-min cycles,2,15 thus increasing the chest compression fraction (CCF). In addition, ventricular fibrillation (VF) could be detected earlier and shocks advanced. All these actions may have a significant positive impact on outcome.16,17 Over the years, considerable effort has been dedicated to develop methods for rhythm analysis during CPR.13 Several approaches have been studied such as adaptive artefact filters 18–20 or new SAAs designed to analyse either the artefacted 21,22 or the filtered ECG.23 When tested on short strips (< 20 s) of OHCA data from defibrillators, only a recently published SAA23 produced sensitivity and specificity results above 90% and 95%, the performance goals recommended by the American Heart Association (AHA).24 Nevertheless, the reported sensitivities and specificities were well below those obtained for SAAs on artefactfree ECG.25 Furthermore, the algorithms were not tested in long strips (minutes) of defibrillator data, i.e. when the method is used to continuously assess the rhythm during CPR. This latter type of testing allows the statistical evaluation of the impact of using the method on the delivery of CPR,26 as suggested by the 2010 consensus on resuscitation science.27 Basic Life Support (BLS) guidelines recommend 30:2 CV-ratio CPR. At the standard rate of 100 compressions per minute, pauses for two ventilations occur approximately every 20 s. Furthermore, in BLS scenarios these pauses normally last longer than 5 s.28 Recently, Ruiz et al 29 proposed doing rhythm analysis during these pauses because the ECG would not have chest compression artefacts. This would allow an accurate rhythm assessment every 20 s during CPR. The objective of this study is to prove the reliability of a fully automatic method to analyse the rhythm during ventilation pauses, and to evaluate the method’s impact on the delivery of CPR.
2. Materials and methods 2.1. Data collection Data from 370 OHCA cases from adult patients were collected between July 2012 and June 2013 in the city of Oslo (Norway). From the 370 cases 135 qualified for inclusion in this study, as explained at the end of this section. In all cases, CPR was performed by Advanced Life Support (ALS) responders. Lifepak 12/15 defibrillators were used (Physio-Control, Redmond, WA, USA), and ECG and thoracic impedance (TI) acquired from the defibrillation pads were stored. Within each case, the initial rhythm and each subsequent change in rhythm were annotated by consensus between a clinical researcher (HN) under the supervision of an experienced anaesthesiologist (JKJ), and a biomedical engineer (UI), all specialised in resuscitation. The following rhythm annotations were used: VF and ventricular tachycardia in the shockable category, and asystole (ASY) and organised (ORG) rhythm in the nonshockable category. Intermediate rhythms such as fine VF were annotated as undecided. These are the rhythms for which, according to the AHA statement,24 the benefits of defibrillation are unclear. Chest compressions were automatically detected by the Code-Stat 9.0 review software (Physio-Control). Periods longer than 1.5 s between consecutive chest compressions were defined as a pause.30 The chest compression marks in the vicinity of the pauses were visually inspected, and manually corrected when necessary based on the TI and ECG signals. The resulting annotated pauses were used as gold standard for the automatic detection of pauses. The criterion to include a case in this study was that the case contained at least 4 min of CPR delivered with 30:2 CV-ratio. Frequently the CPR pattern changed within a case, for instance from 30:2 CPR to continuous chest compressions after intubation. Consequently, for each case intervals in which 30:2 CPR was administered were identified. These 30:2 CPR intervals had to have a duration above
1 min and at least two pauses every minute of CPR. The case was included in the study when the aggregate duration of its 30:2 CPR intervals was more than 4 min. In this study only the 30:2 CPR intervals within each case were analysed. The top panels in Fig 1 show a 2-min interval with six pauses for ventilation. 2.2. Fully automatic rhythm analysis method The reliability of an automatic pause detector based on the TI was evaluated. The detector was used in combination with the automated rhythm analysis of a SAA capable of analysing the rhythm using only a few seconds of the ECG. If a pause was detected and its duration was long enough for a rhythm analysis, the SAA was launched for a shock/no-shock decision. 2.2.1. Shock advice algorithm The SAA is an AHA compliant algorithm that analyses 3 s of the ECG for an accurate shock/no-shock diagnosis,29 a detailed description of the algorithm is available.31 Briefly, the SAA is composed of an asystole detector, a QRS presence detector (nonshockable rhythms) and a final shock/no-shock algorithm based on the regularity, spectral distribution and heart-rate of the rhythm. When tested on an AHA compliant database the SAA presented a sensitivity of 98.5% for shockable rhythms, and a specificity of 99.1% and 100% for ORG and ASY rhythms, respectively.29 The SAA is a modified Matlab version (MathWorks Inc., Natick, MA) of the one incorporated in the Reanibex R-series defibrillators (Bexen Cardio, Ermua, Spain). 2.2.2. Pause detection and rhythm analysis The SAA used in this study analyses 3 s of the ECG to make a decision. Consequently, for this study pauses longer than 3.5 s were defined as diagnosable-pauses, i.e. pauses long enough for the SAA to diagnose the rhythm. The initial 0.5 s of the pause were discarded to avoid transients and artefact residuals in the pause. Diagnosablepauses were automatically detected following these steps (see Fig 1 for a visual description): 1 Pause identification: The envelope of the TI signal was obtained by applying standard signal processing techniques 32 (step 1 of Fig 1). The TI-envelope is large during compressions and vanishes during the pauses. Pauses were automatically detected in the TI-envelope using an adaptive threshold. Adaptivity is necessary because the amplitude of the TI-envelope may vary substantially within each case. 2 Pause onset/offset detection: A chest compression detector was applied in the vicinity of the detected pause to accurately determine the onset/offset of the pause (step 2 of Fig 1). 3 Rhythm analysis: If the pause was longer than 3.5 s, the SAA was automatically launched and the rhythm was analysed 0.5 s after the detected pause onset (step 3 of Fig 1). 2.3. Performance evaluation The method was evaluated in terms of: (1) accuracy of the pause detector, (2) accuracy of the SAA in the detected pauses and (3) the potential therapeutic benefits of using the method. 2.3.1. Reliability of the pause detector Detected diagnosable-pauses were compared to the diagnosable-pauses obtained from the gold standard. Pause detection sensitivity was defined as the proportion of correctly detected diagnosable-pauses, and the positive predictive value (PPV) as the proportion of detected diagnosable-pauses that were true diagnosable-pauses. Sensitivity and PPV values were calculated for the whole dataset and for each case. The pause onset
U. Ayala et al. / Resuscitation 89 (2015) 25–30
27
Fig. 1. Processing steps for automatic pause detection and rhythm analysis. The figure shows on top the ECG and impedance during a 30:2 CPR interval of 2 min duration. In this 30:2 CPR interval there were 6 pauses for two ventilations (slow impedance fluctuations during the pause). In the bottom panels the processing steps are highlighted around a diagnosable-pause. In step 1 the envelope of the impedance (thick grey line) reflects chest compression activity, so pauses are detected by adaptively identifying when this envelope vanishes. In step 2 chest compressions (labelled with dots) are detected in the vicinity of the pause to accurately mark pause onset/offset. In step 3 the SAA is launched 0.5 s after pause onset, and the rhythm is diagnosed as nonshockable (NS) analysing a 3 s ECG segment.
detection accuracy was determined by comparing the detected onset time with that of the last compression annotation before the pause. 2.3.2. Accuracy of the rhythm analysis The SAA was launched in each automatically detected pause, and its diagnosis was compared to the rhythm annotation in the pause. The rhythm analysis accuracy was measured in terms of the proportion of correct shock (sensitivity) and no-shock (specificity) diagnoses. 2.3.3. Potential therapeutic benefits There are two potential benefits of using this method. For patients in shockable rhythms refibrillation could be detected earlier, and the shock advanced. For patients in nonshockable rhythms CPR would not be stopped after 2 min for rhythm assessment, but would rather continue uninterrupted until a change in rhythm
occurs. To measure these potential benefits, the hypothetical effect on CPR delivery of following the automated diagnoses given by the method was quantified. The results depend on when the rescuer would have stopped CPR using the feedback from the method. Two stop-CPR conditions were analysed: stop-CPR after the first shock diagnosis (regular), and stop-CPR after two consecutive shock diagnoses (conservative). For patients in shockable rhythms the proportion of cases in which the shock would have been advanced was analysed. For this evaluation segments were selected within the analysis intervals meeting these conditions: the rhythm was shockable, the interval ended in a defibrillation attempt, and it had at least two diagnosable-pauses. For patients with nonshockable rhythms how the method would prolong CPR cycles beyond the currently recommended 2 min was quantified. Kaplan–Meier survival curves were used to estimate the probability of prolonging CPR cycles as a function of their duration.26 Segments of at least 2 min in which the
28
U. Ayala et al. / Resuscitation 89 (2015) 25–30
rhythm was nonshockable were selected, and for each segment the time at which the stop-CPR condition occurred determined the duration of the CPR cycle.
Table 1 Positive predictive value of the pause detection method, and sensitivity/specificity (90% one-sided lower CI in parenthesis) of the SAA in the detected diagnosablepauses. Rhythm type
2.4. Statistical analysis The statistical distributions of the database’s descriptive data did not pass the one-sample Kolmogorov–Smirnov normality test and are presented as median and interquartile range (IQR). The pause detector’s sensitivity and PPV per case are presented as mean, their 95% confidence intervals (CI) were calculated using a bootstrap procedure with 1000 replications and the basic bootstrap interval. Consecutive rhythm analysis diagnoses within an interval are correlated, so sensitivity and specificity and their 90% one-sided lower CI were adjusted for clustering within the intervals using generalised estimating equations within a logistic regression model.33 The calculations were done using the Geepack library 34 in the R statistical software. For nonshockable segments, Kaplan–Meier survival curves were computed, and their 95% CI were calculated using Greenwood’s variance. Segments in which the stop-CPR condition did not occur were regarded as censored observations.
Shockable Nonshockable AS ORG
Detected pauses n
PPV (%)
SAA Se/Sp (%)
291 1599 1032 567
97.9 97.2 97.0 97.7
93.8 (90.0) 95.9 (94.7) 96.2 (94.7) 95.5 (93.2)
3.2. Reliability of the pause detector The overall sensitivity and PPV of the pause detector were 93.5% and 97.3%, respectively. The mean sensitivity and PPV per case were 93.5% (95% CI, 92.0–95.0) and 96.6% (95% CI, 95.7–97.5), respectively. The median error in pause onset identification was 0.0 s (5–95 percentile, −0.1 to 0.6). Panel a of Fig 2 shows an example of an accurate pause detection, while panels b and c expose the challenges of pause detection using the TI.
3.3. Accuracy of the rhythm analysis 3. Results 3.1. Database summary There were a total of 229 30:2 CPR intervals within the 135 cases, comprising 1414 min. The median duration of the intervals was 4.8 min (IQR, 3.0–7.5), and a median of 9 min (IQR, 5.9–12.4) were analysed per case. There were 1942 diagnosable-pauses in the analysed data, with 11 (IQR, 6–18) pauses per case. The median duration of the diagnosable-pauses was 4.7 s (IQR, 4.0–6.5).
The SAA was launched in the 1917 diagnosable-pauses that were automatically detected, in 27 of those the rhythm was annotated as undecided and were therefore discarded for the evaluation of the sensitivity and specificity. The sensitivity and specificity results are shown in Table 1. The overall sensitivity and specificity were 93.8% (90% one-sided lower CI, 90.0%) and 95.9% (90% one-sided lower CI, 94.7%), respectively. Fig 3 shows three examples of rhythm analysis. In panels a and b the rhythm was correctly diagnosed, while in panel c an erroneous shock was advised due to electrode/cord movement noise in the ECG.
Fig. 2. Examples of correct and missed chest compression pause detection. In (b) noise in the impedance caused the detected pause to be too short (<3.5 s) to launch an analysis. In (c) very low fluctuations in TI caused by chest compressions were not detected and pause onset was not accurately determined resulting in a false positive pause detection (pause in an artefacted ECG segment).
Fig. 3. Examples of correct and erroneous rhythm analysis during ventilation pauses. The diagnosis given by the SAA is shown within the pause. In (c) the first pause was annotated as nonshockable (NS) but was erroneously classified as shockable (S) by the SAA.
U. Ayala et al. / Resuscitation 89 (2015) 25–30
Fig. 4. Kaplan–Meier survival curves with 95% confidence intervals for nonshockable cases. The curves represent the probability of prolonging the CPR cycle as a function of the duration of the cycle. The cut-off point at 2 min gives the probability of increasing CCF compared to current guidelines.
3.4. Potential therapeutic benefits A total of 35 segments with at least two diagnosablepauses before defibrillation, and 168 segments with nonshockable rhythms were identified. The median duration of the shockable and nonshockable segments were 2.6 min (IQR, 2.0–3.3) and 4.9 min (IQR, 3.6–7.5), respectively. With the regular criterion for stopCPR all shocks would be advanced, 91.4% (32/35) of cases to the first diagnosable-pause and the remaining 8.6% to the second diagnosable-pause. As shown in Fig 4, the 2-min CPR cycles would be prolonged in 86.3% (81.1–91.5) of nonshockable cases. With the conservative criterion 97% (34/35) of shocks would be advanced, and the 2-min CPR cycles would be prolonged in 97.6% (95.3–99.9) of the nonshockable cases. Moreover, uninterrupted CPR cycles of at least 4 min would occur in 95.2% (91.6–98.8) of nonshockable cases, substantially increasing the CCF. 4. Discussion In the past, methods for rhythm analysis during CPR have followed or have combined two main lines of attack: the removal of the chest compression artefacts,18–20 and special purpose rhythm analysis methods marginally affected by the artefacts.21–23 Generally, the specificity of these methods is below the 95% value recommended by the AHA.13 Furthermore, when used to continuously monitor the rhythm during CPR their performance degrades substantially.26 This study demonstrates the reliability of a fully automatic method that avoids chest compression artefacts altogether by analysing the rhythm during the ventilation pauses in 30:2 CV-ratio CPR (a video in the supplementary material shows a real-time demonstration of the method). This method would eliminate the need to stop CPR for rhythm analysis, and could therefore serve to advance the shock when patients refibrillate and to prolong the CPR cycle in patients with nonshockable rhythms. The method could be used in any circumstance in which pauses in chest compressions are frequent and regular. However, it was primarily conceived to be used in a Basic Life Support (BLS) scenario, in which the 30:2 CV-ratio is recommended.2 The method was therefore designed to work with the only data available in an AED scenario, the ECG and the TI signals recorded through the defibrillation pads. TI has been effectively used to detect chest compressions,35,36 ventilations,37 pulse,38 and to monitor CPR quality parameters like compression rate or CCF.36 However,
29
TI based chest compression detectors fail to accurately determine short pauses in chest compressions.36 Our method successfully overcomes this limitation by combining an adaptive TI-envelope detector to identify potential pauses with a compression detector to accurately determine pause onset and offset. This resulted in a very reliable detector that identified more than 93% of pauses, with very few false positives, the PPV was above 97%. Avoiding false positives is paramount for the success of the method because it guarantees rhythm analysis is done on an ECG free of chest compression artefacts. Moreover, our testing conditions were very challenging because pauses were very short (median duration 4.7 s). Indeed, pauses for two ventilations in CPR performed by laypersons may last around 7 s (IQR, 6–9),28 in those circumstances the accuracy and reliability of the detector would further improve. Accurate pause onset detection is very important in this application. This is a challenging task because compression induced fluctuations in TI change a great deal within an case, and even adaptive compression detectors fail to accurately determine pause onset and offset.36 However, applying a compression detector only locally around detected pauses results in a very accurate pause onset detection with a median delay of 0 s. Consequently, pauses as short as 3.5 s could be safely used to diagnose the rhythm, using a SAA capable of accurately analysing the ECG in such short ECG intervals. The resulting sensitivity and specificity were above AHA performance goals, and the accuracy for the fully automatic method was very similar to the one obtained in a previous study in which the analysis pauses were manually annotated.29 In this study the method was also used to analyse long OHCA segments, in which the rhythm was monitored during ventilation pauses. For a conservative stop-CPR criterion defibrillation would have been advanced in 97% of cases. Earlier recognition of recurrent VF could have a beneficial impact on outcome given the high oxygen demands of recurrent VF,17 and that time in recurrent VF is associated with worse outcomes.39 Furthermore, considering that ventilation pauses occur approximately every 20 s in 30:2 CPR, the coarseness of VF could be monitored to optimise defibrillation time.40 Our results show that a conservative stop-CPR criterion is needed to safely increase CCF during nonshockable rhythms. Uninterrupted CPR cycles could be prolonged beyond 4 min in over 95% of cases, avoiding many rhythm reassessment pauses. This would substantially increase CCF in patients with nonshockable rhythms, which could increase the likelihood of ROSC.16 Decisions that may improve survival such as an increased duration of the CPR cycles during asystole,15 or treatment recommendations tailored to the type of nonshockable rhythm could be incorporated to the AED. In summary, adding a rhythm monitor function during CPR could turn current AEDs into intelligent devices that guide the rescuer to follow optimal treatment decisions. In any case, feedback to the rescuer would only be given when action is needed, such as to stop CPR to deliver the shock when refibrillation is detected. 4.1. Limitations In our dataset CPR was performed by highly trained ALS responders, which resulted in very short pauses for two ventilations. The method is primarily intended for a BLS scenario with lay responders. In these cases the expected pause duration would be longer. This should most likely increase the accuracy of the method since the testing conditions would be more favourable. 5. Conclusions This study proved that a fully automatic method could be used to safely analyse the rhythm during ventilation pauses in 30:2 CV-ratio CPR. This would contribute to an early detection of
30
U. Ayala et al. / Resuscitation 89 (2015) 25–30
refibrillation and to an increase in CCF in patients with nonshockable rhythms. Conflict of interest statement Authors U. Irusta, J. Ruiz and S. Ruiz de Gauna have received research support from Bexen Cardio (Ermua, Spain) for studies on AED shock advice algorithms. Acknowledgements This work received financial support from the Ministerio de Ciencia e Innovación of Spain through the projects TEC2012-31928 and TEC2012-31144, from the University of the Basque Country (UPV/EHU) through the unit UFI11/16 and from the Programa de Formación de Personal Investigador del Departamento de Educación, Universidades e Investigación del Gobierno Vasco through the grants BFI-2010-174, BFI-2011-166 and BFI-2010-235. Appendix A. Supplementary Data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.resuscitation.2014.11.022. References 1. Nolan J, Soar J, Eikeland H. The chain of survival. Resuscitation 2006;71(3):270–1. 2. Koster RW, Baubin MA, Bossaert LL, et al. European Resuscitation Council Guidelines for Resuscitation 2010 Section 2. Adult basic life support and use of automated external defibrillators. Resuscitation 2010;81(10):1277–92. 3. Kramer-Johansen J, Myklebust H, Wik L, et al. Quality of out-of-hospital cardiopulmonary resuscitation with real time automated feedback: a prospective interventional study. Resuscitation 2006;71:283–92. 4. Berg R, Sanders A, Kern K, et al. Adverse hemodynamic effects of interrupting chest compressions for rescue breathing during cardiopulmonary resuscitation for ventricular fibrillation cardiac arrest. Circulation 2001;104(20):2465–70. 5. Eftestøl T, Sunde K, Steen P. Effects of interrupting precordial compressions on the calculated probability of defibrillation success during out-of-hospital cardiac arrest. Circulation 2002;105(19):2270–3. 6. Edelson DP, Abella BS, Kramer-Johansen J, et al. Effects of compression depth and pre-shock pauses predict defibrillation failure during cardiac arrest. Resuscitation 2006;71:137–45. 7. Mader T, Paquette A, Salcido D, Nathanson B, Menegazzi J. The effect of the preshock pause on coronary perfusion pressure decay and rescue shock outcome in porcine ventricular fibrillation. Prehosp Emerg Care 2009;13(4):487–94. 8. Cheskes S, Schmicker R, Christenson J, et al. Perishock pause: an independent predictor of survival from out-of-hospital shockable cardiac arrest. Circulation 2011;124(1):58–66. 9. Jost D, Degrange H, Verret C, et al. DEFI 2005: a randomized controlled trial of the effect of automated external defibrillator cardiopulmonary resuscitation protocol on outcome from out-of-hospital cardiac arrest. Circulation 2010;121(14):1614–22. 10. Cheskes S, Common MR, Byers PA, Zhan C, Morrison LJ. Compressions during defibrillator charging shortens shock pause duration and improves chest compression fraction during shockable out of hospital cardiac arrest. Resuscitation 2014;85(8):1007–11. 11. Rhee JE, Kim T, Kim K, Choi S. Is there any room for shortening hands-off time further when using an AED? Resuscitation 2009;80(2):231–7. 12. Didon JP, Krasteva V, Ménétré S, Stoyanov T, Jekova I. Shock advisory system with minimal delay triggering after end of chest compressions: accuracy and gained hands-off time. Resuscitation 2011;82(Suppl. 2):S8–15. 13. Ruiz de Gauna S, Irusta U, Ruiz J, Ayala U, Aramendi E, Eftestøl T. Rhythm analysis during cardiopulmonary resuscitation: past, present, and future. BioMed Res Int 2014;2014. Article ID 386010. 14. Snyder D, Morgan C. Wide variation in cardiopulmonary resuscitation interruption intervals among commercially available automated external defibrillators may affect survival despite high defibrillation efficacy. Crit Care Med 2004;32(9 Suppl.):S421–4. 15. Nordseth T, Edelson DP, Bergum D, et al. Optimal loop duration during the provision of in-hospital advanced life support (ALS) to patients with an initial non-shockable rhythm. Resuscitation 2014;85(1):75–81.
16. Vaillancourt C, Everson-Stewart S, Christenson J, et al. The impact of increased chest compression fraction on return of spontaneous circulation for out-ofhospital cardiac arrest patients not in ventricular fibrillation. Resuscitation 2011;82(12):1501–7. 17. Hoogendijk MG, Schumacher CA, Belterman CN, et al. Ventricular fibrillation hampers the restoration of creatine-phosphate levels during simulated cardiopulmonary resuscitations. Europace 2012;14(10):1518–23. 18. Eilevstjønn J, Eftestøl T, Aase SO, Myklebust H, Husøy JH, Steen PA. Feasibility of shock advice analysis during CPR through removal of CPR artefacts from the human ECG. Resuscitation 2004;61(2):131–41. 19. Werther T, Klotz A, Granegger M, et al. Strong corruption of electrocardiograms caused by cardiopulmonary resuscitation reduces efficiency of two-channel methods for removing motion artefacts in non-shockable rhythms. Resuscitation 2009;80(11):1301–7. 20. Aramendi E, Ayala U, Irusta U, Alonso E, Eftestøl T, Kramer-Johansen J. Suppression of the cardiopulmonary resuscitation artefacts using the instantaneous chest compression rate extracted from the thoracic impedance. Resuscitation 2012;83(6):692–8. 21. Li Y, Bisera J, Geheb F, Tang W, Weil MH. Identifying potentially shockable rhythms without interrupting cardiopulmonary resuscitation. Crit Care Med 2008;36(1):198–203. 22. Krasteva V, Jekova I, Dotsinsky I, Didon JP. Shock advisory system for heart rhythm analysis during cardiopulmonary resuscitation using a single ECG input of automated external defibrillators. Ann Biomed Eng 2010;38(4):1326–36. 23. Ayala U, Irusta U, Ruiz J, et al. A reliable method for rhythm analysis during cardiopulmonary resuscitation. BioMed Res Int 2014;2014. Article ID 872470. 24. Kerber RE, Becker LB, Bourland JD, et al. Automatic external defibrillators for public access defibrillation: recommendations for specifying and reporting arrhythmia analysis algorithm performance, incorporating new waveforms, and enhancing safety: a statement for health professionals from the American heart association task force on automatic external defibrillation, subcommittee on AED safety and efficacy. Circulation 1997;95(6):1677–82. 25. Atkins DL, Scott WA, Blaufox AD, et al. Sensitivity and specificity of an automated external defibrillator algorithm designed for pediatric patients. Resuscitation 2008;76(2):168–74. 26. Ruiz J, Ayala U, Ruiz de Gauna S, et al. Direct evaluation of the effect of filtering the chest compression artifacts on the uninterrupted cardiopulmonary resuscitation time. Am J Emerg Med 2013;31(6):910–5. 27. Sayre MR, Koster RW, Botha M, et al. Part 5: Adult basic life support: 2010 International consensus on cardiopulmonary resuscitation and emergency cardiovascular care science with treatment recommendations. Circulation 2010;16(S2):S298–324. 28. Beesems SG, Wijmans L, Tijssen JGP, Koster RW. Duration of ventilations during cardiopulmonary resuscitation by lay rescuers and first responders: relationship between delivering chest compressions and outcomes. Circulation 2013;127(15):1585–90. 29. Ruiz J, Ayala U, Ruiz de Gauna S, et al. Feasibility of automated rhythm assessment in chest compression pauses during cardiopulmonary resuscitation. Resuscitation 2013;84(9):1223–8. 30. Kramer-Johansen J, Edelson DP, Losert H, Köhler K, Abella BS. Uniform reporting of measured quality of cardiopulmonary resuscitation (CPR). Resuscitation 2007;74(3):406–17. 31. Irusta U, Ruiz J, Aramendi E, Ruiz de Gauna S, Ayala U, Alonso E. A high-temporal resolution algorithm to discriminate shockable from nonshockable rhythms in adults and children. Resuscitation 2012;83(9):1090–7. 32. González-Otero D, de Gauna SR, Ruiz J, Ayala U, Alonso E. Automatic detection of chest compression pauses using the transthoracic impedance signal. In: Computing in Cardiology (CinC), 2012, IEEE. 2012. p. 21–4. 33. Sternberg MR, Hadgu A. A GEE approach to estimating sensitivity and specificity and coverage properties of the confidence intervals. Stat Med 2001;20(9–10):1529–39. 34. Halekoh U, Højsgaard S, Yan J. The R package geepack for generalized estimating equations. J Stat Softw 2006;15(2):1–11. 35. Stecher FS, Olsen JA, Stickney RE, Wik L. Transthoracic impedance used to evaluate performance of cardiopulmonary resuscitation during out of hospital cardiac arrest. Resuscitation 2008;79(3):432–7. 36. Ayala U, Eftestøl T, Alonso E, et al. Automatic detection of chest compressions for the assessment of CPR-quality parameters. Resuscitation 2014;85(7):957–63. 37. Risdal M, Aase SO, Stavland M, Eftestøl T. Impedance-based ventilation detection during cardiopulmonary resuscitation. IEEE Trans Biomed Eng 2007;54(12):2237–45. 38. Ruiz J, Alonso E, Aramendi E, et al. Reliable extraction of the circulation component in the thoracic impedance measured by defibrillation pads. Resuscitation 2013;84(10):1345–52. 39. Berdowski J, ten Haaf M, Tijssen JGP, Chapman FW, Koster RW. Time in recurrent ventricular fibrillation and survival after out-of-hospital cardiac arrest. Circulation 2010;122(11):1101–8. 40. Ristagno G, Li Y, Fumagalli F, Finzi A, Quan W. Amplitude spectrum area to guide resuscitation: a retrospective analysis during out-of-hospital cardiopulmonary resuscitation in 609 patients with ventricular fibrillation cardiac arrest. Resuscitation 2013;84(12):1697–703.