Advertisement

Continuous heart rate dynamics preceding in-hospital pulseless electrical activity or asystolic cardiac arrest of respiratory etiology

Open AccessPublished:July 26, 2022DOI:https://doi.org/10.1016/j.resuscitation.2022.07.026

      Abstract

      Introduction

      Respiratory failure is a common cause of pulseless electrical activity (PEA) and asystolic cardiac arrest, but the changes in heart rate (HR) pre-arrest are not well described. We describe HR dynamics prior to in-hospital cardiac arrest (IHCA) among PEA/asystole arrest patients with respiratory etiology.

      Methods

      In this retrospective study, we evaluated 139 patients with 3–24 hours of continuous electrocardiogram data recorded preceding PEA/asystole IHCA from 2010-2017. We identified respiratory failure cases by chart review and evaluated electrocardiogram data to identify patterns of HR changes, sinus bradycardia or sinus arrest, escape rhythms, and development right ventricular strain prior to IHCA.

      Results

      A higher proportion of respiratory cases (58/73, 79 %) fit a model of HR response characterized by tachycardia followed by rapid HR decrease prior to arrest, compared to non-respiratory cases (30/66, 45 %, p < 0.001). Among the 58 respiratory cases fitting this model, 36 (62 %) had abrupt increase in HR occurring 64 (IQR 23–191) minutes prior to arrest, while 22 (38 %) had stable tachycardia until time of HR decrease. Mean peak HR was 123 ± 21 bpm. HR decrease occurred 3.0 (IQR 2.0–7.0) minutes prior to arrest. Sinus arrest occurred during the bradycardic phase in 42/58 of cases; escape rhythms were present in all but 2/42 (5 %) cases. Right ventricular strain ECG pattern, when present, occurred at a median of 2.2 (IQR −0.05–17) minutes prior to onset of HR decrease.

      Conclusion

      IHCAs of respiratory etiology follow a model of HR increase from physiologic compensation to hypoxia, followed by rapid HR decrease prior to arrest.

      Keywords

      Abbreviations:

      IHCA (in hospital cardiac arrest), PEA (pulseless electrical activity), HR (heart rate), ECG (electrocardiogram), bpm (beats per minute), RV (right ventricle), IQR (interquartile range), AV (atrioventricular)

      Introduction

      In-hospital cardiac arrests (IHCA) affect over 292,000 patients in the United States annually, with fewer than 30 % surviving to discharge.
      • Holmberg M.J.
      • Ross C.E.
      • Fitzmaurice G.M.
      • et al.
      Annual Incidence of Adult and Pediatric In-Hospital Cardiac Arrest in the United States.
      Survival is even lower for pulseless electrical activity (PEA) arrest, and our understanding of its pathophysiology is limited, despite PEA and asystole arrest accounting for 37 % and 39 % of cardiac arrest rhythms, respectively.
      • Meaney P.A.
      • Nadkarni V.M.
      • Kern K.B.
      • Indik J.H.
      • Halperin H.R.
      • Berg R.A.
      Rhythms and outcomes of adult in-hospital cardiac arrest.
      Strategies to improve survival after IHCA include prevention and early recognition of clinical deterioration.
      • Morrison L.J.
      • Neumar R.W.
      • Zimmerman J.L.
      • et al.
      Strategies for improving survival after in-hospital cardiac arrest in the United States: 2013 consensus recommendations: A consensus statement from the American heart association.
      To this end, the detection of changes in continuous electrocardiogram (ECG) recording may be a useful addition to monitoring other vital signs and laboratory values.
      • Do D.H.
      • Kuo A.
      • Lee E.S.
      • et al.
      Usefulness of Trends in Continuous Electrocardiographic Telemetry Monitoring to Predict In-Hospital Cardiac Arrest.
      • Attin M.
      • Feld G.
      • Lemus H.
      • et al.
      Electrocardiogram characteristics prior to in-hospital cardiac arrest.
      Respiratory failure is the most common etiology of PEA/asystole IHCA. Prompt recognition and reversal with adequate ventilation/oxygenation can improve the success of resuscitation. We have previously shown that right ventricular (RV) strain pattern on ECG is associated with IHCA of respiratory etiology.
      • Do D.H.
      • Yang J.J.
      • Kuo A.
      • et al.
      Electrocardiographic right ventricular strain precedes hypoxic pulseless electrical activity cardiac arrests: Looking beyond pulmonary embolism.
      These etiologies include processes that cause elevated pulmonary vascular resistance such as pulmonary embolism and hypoxic vasoconstriction, or processes that cause pulmonary vascular bed compression, such as acute respiratory distress syndrome and aspiration.
      • Do D.H.
      • Yang J.J.
      • Kuo A.
      • et al.
      Electrocardiographic right ventricular strain precedes hypoxic pulseless electrical activity cardiac arrests: Looking beyond pulmonary embolism.
      • Netzer N.C.
      • Strohl K.P.
      • Högel J.
      • Gatterer H.
      • Schilz R.
      Right ventricle dimensions and function in response to acute hypoxia in healthy human subjects.
      The progression of these conditions can lead to cardiac arrest. Additionally, developing RV strain pattern on ECG and echocardiogram has been associated with increased mortality in COVID-19 pneumonia.
      • Li Y.
      • Li H.
      • Zhu S.
      • et al.
      Prognostic Value of Right Ventricular Longitudinal Strain in Patients With COVID-19.
      • Barman H.A.
      • Atici A.
      • Sahin I.
      • et al.
      Prognostic value of right ventricular strain pattern on ECG in COVID-19 patients.
      Heart rate (HR) patterns preceding arrests due to respiratory failure, however, have not been well studied in the clinical setting.
      • Skrifvars M.B.
      • Nurmi J.
      • Ikola K.
      • Saarinen K.
      • Castrén M.
      Reduced survival following resuscitation in patients with documented clinically abnormal observations prior to in-hospital cardiac arrest.
      • Bhalala U.S.
      • Bonafide C.P.
      • Coletti C.M.
      • et al.
      Antecedent bradycardia and in-hospital cardiopulmonary arrest mortality in telemetry-monitored patients outside the ICU.
      In this study, we describe HR dynamics prior to IHCA and its associations with timing of respiratory failure and RV strain. We hypothesized that PEA/asystole arrest of respiratory etiology would be associated with a pattern of HR increase in response to hypoxic insult,

      Simon PM, Taha BH, Dempsey JA, Skatrud JB, Iber C, Taha H. Role of vagal feedback from the lung in hypoxic-induced tachycardia in humans.

      • Clowes G.H.
      • Hopkins A.L.
      • Simeone F.A.
      A Comparison of the Physiological Effects of Hypercapnia and Hypoxia in the Production of Cardiac Arrest.
      followed by HR decrease prior to arrest.

      Methods

      This was a retrospective cross-sectional study at the University of California, Los Angeles (UCLA) Ronald Reagan Medical Center and the University of California, San Francisco (UCSF) Medical Center. Telemetry data was obtained by bedside patient monitoring systems (General Electric Healthcare, Waukesha, WI), and pooled on a remote data server via BedmasterEx (Excel Medical Electronics, Jupiter, FL). Signals were sampled at 240 Hz with 12-bit representation. Continuous ECG was obtained using a standard 5 electrode configuration providing 4 ECG leads (I, II, III and precordial lead). This study received approval from the UCLA and UCSF Institutional Review Boards. Due to the nature of this chart review study, patient consent was waived by the Institutional Review Board.
      We evaluated all “code blues” (calls for emergency response team) between April 2010 and August 2014 at UCLA and intensive care unit “code blues” between March 2013 and December 2017 at UCSF. We included IHCA cases (defined as lack of central pulse, apnea, and unresponsiveness) due to PEA/asystole in patients age ≥ 18 years, with telemetry data available for ≥ 3 consecutive hours prior to and including the onset of IHCA. Only the first IHCA for a given patient was included. We excluded patients with a do-not-resuscitate order at time of “code blue,” ventricular-paced rhythm, prior orthotopic heart transplant, left ventricular assist device or extracorporeal membranous oxygenation support at time of arrest, out-of-hospital cardiac arrest leading to current admission, IHCA in a procedural or operating room, IHCA within the first 24 hours of a trauma admission, and ECG data that was incomplete or had excess artifact. Additionally, we excluded cases with atrial fibrillation or flutter prior to arrest, as this would confound analysis of sinus node function.
      We obtained telemetry data for up to 24 hours preceding IHCA and analyzed the ECG data using LabChart Reader (AD Instruments, version 8.1.13). The time of IHCA was determined by ECG review and defined as the onset of asystole or the visualization of chest compression artifact in PEA. The clinical cause of IHCA and approximate time course (e.g., minutes, hours, days) of preceding respiratory event, hypoxia, and hypercarbia were determined by chart review of medical history, resuscitation notes, pre- and post-arrest vital signs, labs, imaging, and autopsy data. Clinical cause was determined by two physicians (DD, AK) independently, both blinded to the results of ECG analysis, with discrepancies resolved by discussion or adjudication by a third physician (NB). The timing of respiratory event was determined from chart review based on timing of events such as aspiration, endotracheal intubation, altered mental status, clinical signs of respiratory distress, hypoxia, and blood gas results. Demographics and IHCA outcome were also obtained through chart review.
      We applied a MatLab script (MathWorks Inc, version 9.8) to graph the 5-minute averaged HR for each included case (Supplemental Figure). Using this graph and review of continuous ECG tracings, we identified the presence and timing of HR increase and decrease, maximum HR, sinus arrest, and escape rhythms. We then classified the cases into five models of HR dynamics:
      • 1.
        Sinus rhythm with HR increase, followed by a rapid decrease (Model 1)
      • 2.
        Sinus tachycardia without HR increase during period of available data, followed by rapid decrease (Model 2)
      • 3.
        Normal sinus rhythm, without HR change during period of available data, followed by rapid decrease (Model 3)
      • 4.
        Gradual decline in sinus rate without inflection points (Model 4)
      • 5.
        Other – none of the above patterns
      In cases of multiple HR fluctuations, HR pattern closest to the time of IHCA was prioritized. Model classifications were made by author RS with additional adjudication by author DD. Additionally, we identified the presence and onset of RV strain, which was defined as gradual delay in RV depolarization in lead V1 with at least one supporting finding of RV ischemia or right axis deviation.
      • Do D.H.
      • Yang J.J.
      • Kuo A.
      • et al.
      Electrocardiographic right ventricular strain precedes hypoxic pulseless electrical activity cardiac arrests: Looking beyond pulmonary embolism.
      Normally distributed variables were reported as mean ± SD. Non-normally distributed variables were reported as median (IQR). Chi-squared test was used to evaluate for differences between categories. Two-sided p-value of < 0.05 was considered statistically significant. Analysis was performed using Stata (StataCorp, version 15.1).

      Results

      We reviewed 336 cases of PEA/asystole IHCA; following exclusions (Fig. 1), 139 cases were included in the analysis. The most common reasons for exclusion were not first arrest (43/196), missing ECG data (44/196), atrial fibrillation around time of arrest (32/196), and ventricular paced rhythm (21/196). There were 73 cases (mean age 59.9 ± 17.3, 52.1 % male, 79.5 % return of spontaneous circulation, 30.1 % survival to discharge) of respiratory etiology and 66 cases of non-respiratory etiology (mean age 58.2 ± 17.2, 50.0 % male, 63.6 % return of spontaneous circulation, and 18.2 % survival to discharge). Specific etiologies are listed in Table 1.
      Figure thumbnail gr1
      Fig. 1Study Flowchart. Abbreviations. PEA: pulseless electrical activity. ECMO: extracorporeal membrane oxygenation. ECG: electrocardiogram. DNR: Do-Not-Resuscitate.
      Table 1Patient characteristics.
      Respiratory, n = 73Non-Respiratory, n = 66
      Age59.9 ± 17.358.2 ± 17.2
      Male, n(%)38 (52.1 %)33 (50.0 %)
      ROSC, n(%)58 (79.5 %)42 (63.6 %)
      Survive to discharge, n(%)22 (30.1 %)12 (18.2 %)
      24 hours of ECG data, n(%)53 (72.6 %)39 (59.1 %)
      Primary respiratory etiology
      Endotracheal tube mechanical complication5 (6.9 %)
      Mucous plug9 (12.3 %)
      Acute aspiration8 (11.0 %)
      Pneumothorax1 (1.4 %)
      Hemoptysis1 (1.4 %)
      Pulmonary embolism5 (6.9 %)
      ARDS9 (12.3 %)
      Pneumonia8 (11.0 %)
      Pulmonary hypertension5 (6.9 %)
      Interstitial lung disease2 (2.7 %)
      Other hypoxic respiratory failure or unknown20 (27.4 %)
      Primary non-respiratory etiology
      Metabolic acidosis24 (36.4 %)
      Cardiogenic shock5 (7.6 %)
      Hemorrhagic shock9 (13.6 %)
      Distributive shock3 (4.6 %)
      Multiorgan failure9 (13.6 %)
      Myocardial infarction2 (3.0 %)
      Cardiac conduction abnormality2 (3.0 %)
      Obstructive shock1 (1.5 %)
      Unknown or Other11 (16.7 %)
      Concurrent Processes
      Hypoxia73 (100 %)25 (37.9 %)
      Hypercarbia52 (71.2 %)17 (25.8 %)
      Metabolic acidosis8 (11.0 %)48 (72.7 %)
      Ab breviations. ROSC: return of spontaneous circulation, ECG: electrocardiogram, ARDS: acute respiratory distress syndrome.

      Models of HR response in IHCA

      A significantly higher proportion of respiratory cases (58/73, 79 %) fit either Model 1 or Model 2 (tachycardia followed by rapid HR decrease), compared to non-respiratory cases (45 %, 30/66, p < 0.001, Table 2). The proportion of Model 3 patterns (sinus rhythm followed by rapid HR decrease) was similar between respiratory and non-respiratory cases, while non-respiratory cases had a higher proportion of Model 4 (gradual decline in HR). HR patterns not fitting any of the above models were further described as: persistent sinus rhythm, sinus tachycardia without bradycardic phase, and high degree heart block at the time of arrest (Table 2). Because of the low frequency of these patterns, no further analysis was performed.
      Table 2Distribution of Models. Model 1: sinus rhythm with HR increase followed by rapid HR decrease, Model 2: sinus tachycardia followed by rapid HR decrease, Model 3: sinus rhythm followed by rapid HR decrease, Model 4: gradual decline in HR.
      ModelRespiratory, n = 73Non-respiratory, n = 66p-value
      Model 1, n(%)36 (49.3 %)13 (19.4 %)p < 0.001
      Model 2, n(%)22 (30.1 %)17 (25.4 %)
      Model 3, n(%)11 (15.1 %)10 (14.9 %)
      Model 4, n(%)1 (1.4 %)11 (16.45)
      Other
      Persistent sinus rhythm, n(%)1 (1.4 %)3 (4.5 %)
      Sinus tachycardia without bradycardic phase, n(%)2 (2.7 %)7 (10.6 %)
      Heart block, n(%)04 (5.97 %)
      Unclassified, n(%)01 (1.5 %)
      Heart rate dynamics in respiratory cases fitting Model 1 or 2 are detailed in Fig. 2. Out of 58 cases fitting Model 1 or 2, 36 (62 %) cases exhibited an abrupt HR increase, occurring at a median of 64 (IQR 23–191) minutes prior to arrest. The mean peak HR was 123 ± 21 bpm. An abrupt decrease in HR occurred at 3.0 (IQR 2.0–7.0) minutes prior to arrest. The rate of HR decrease from the onset of HR decrease to either sinus arrest (when present) or time of PEA was 26 beats/min
      • Meaney P.A.
      • Nadkarni V.M.
      • Kern K.B.
      • Indik J.H.
      • Halperin H.R.
      • Berg R.A.
      Rhythms and outcomes of adult in-hospital cardiac arrest.
      (IQR 14–44). This slope was not calculated in the 10 cases for which the time of HR decrease coincided with the time of sinus arrest. Sinus arrest occurred during the HR decrease phase in 42/58 (72 %) of cases; the first escape rhythm was atrial in 20/42 (47 %) of cases, junctional in 14/42 (33 %), ventricular in 6/42 (14 %), and no escape (ie. asystole) in 2/42 (5 %) of cases (Figs. 2 and 3). The mean HR at the start of the first escape rhythm was 63 ± 19 bpm for atrial, 46 ± 20 bpm for junctional, and 56 ± 20 bpm for ventricular escape rhythms.
      Figure thumbnail gr2
      Fig. 2Proposed comprehensive model of HR dynamics prior to IHCA of respiratory etiology (N = 58). Abbreviations. HR: heart rate, RV: right ventricle.
      Figure thumbnail gr3
      Fig. 3Example of sinus rhythm progressing into sinus arrest with atrial escape (A), junctional escape (B), ventricular escape (C), and no escape (D). All tracings represent Lead II from cases with respiratory etiology of arrest.
      Looking at the progression of escape rhythms, among the 42 respiratory cases fitting Model 1 or 2 with sinus arrest, most cases had only one type of escape rhythm, with 11 (26 %) atrial, 10 (24 %) junctional, and 5 (12 %) ventricular. The next most common progression was atrial to junctional escape (6/42, 14 %), followed by junctional to atrial (2/42, 5 %) and atrial to junctional to ventricular (2/41, 5 %). Among the Model 3 cases, sinus arrest occurred during the HR decrease phase in 7 (64 %) cases, with no escape (i.e., asystole) in 3/7 cases (Supplemental Table).
      To assess hemodynamic response during the escape rhythms, we measured blood pressure at the start of each escape rhythm for cases fitting Model 1 and 2 with available arterial line data. The median systolic blood pressure for atrial escape was 70 mmHg (IQR 35–70, n = 11), for junctional escape was 50 mmHg (IQR 40–55, n = 9), and for ventricular escape was 45 mmHg (IQR 40–75, n = 5). The median pulse pressure for atrial escape was 30 mmHg (IQR 15–40, n = 11), for junctional escape was 10 (IQR 10–20, n = 9), and for ventricular escape was 15 mmHg (IQR 5–20, n = 5).
      Among the 73 respiratory cases, 7 cases had high degree or complete atrioventricular (AV) block, which developed concomitantly with HR slowing and was preceded by PR prolongation in 5/7 cases. Sinus arrest occurred around the time of AV block in 4/7 cases. Among these 7 cases of AV block, 2 cases fit Model 1, 4 cases fit Model 2, and 1 case fit Model 3.

      Associations between HR and timing of respiratory event

      When correlating the models with timing of respiratory events, patients fitting Model 1 had a greater proportion of acute events (onset minutes to hours prior to arrest), while Model 2 tended to have subacute onset (≥24 hours or unclear timing), p < 0.001. Specifically, 25/36 (69 %) of Model 1 cases had onset of respiratory event minutes to hours prior to arrest. Among cases fitting Model 2, 13/22 (59 %) cases had respiratory event occurring ≥ 24 hours prior or had unclear onset of event based on clinical documentation. The remaining 9/22 (41 %) cases had respiratory event onset minutes to hours prior to arrest (Fig. 4).
      Figure thumbnail gr4
      Fig. 4Distribution of respiratory event time course, n (%), prior to arrest for Models 1, 2, and 3 (n = 69). Model 1: sinus rhythm with abrupt HR increase followed by rapid decrease in HR (n = 36), Model 2: persistent sinus tachycardia followed by rapid decrease in HR (n = 22), Model 3: normal sinus rhythm followed by rapid decrease in HR (n = 11).
      Among the 54 respiratory cases in which RV strain could be discerned from available data (i.e., correct lead placement, no/minimal artifact in lead v1), RV strain occurred in 37 (69 %) cases. Among the 42 respiratory cases with discernable RV strain and that fit Model 1 or 2, RV strain was present in 28 (67 %) cases; and 18/28 cases (64 %) had RV strain onset occur during the tachycardia phase. The median time from RV strain onset to cardiac arrest in cases that fit Model 1 or 2 was 5.9 minutes (IQR 1.9–19.6); median time between RV strain onset and HR decrease onset was 2.2 minutes (IQR −0.5–17.0, Fig. 2).
      Model 3 (normal sinus rhythm with rapid HR decrease) had 4/11 (36 %) of cases with an acute respiratory event (due to aspiration or mucus plugging) minutes prior to arrest, and 4/11 (36 %) cases with unclear onset. RV strain was seen in 8/11 (73 %) cases. Chart review of the Model 3 cases revealed that 9/11 (82 %) of cases had concurrent factors that may contribute to a lack of tachycardic response, including intracerebral hemorrhage, hypercarbia, hypothermia, quadriplegia, beta blocker and sedative use. Specifically, regarding beta blocker, non-dihydropyridine calcium channel blocker, and digoxin use, there were 6/36 (17 %) of Model 1 cases that used these medications, 2/22 (9 %) of Model 2 cases, and 2/11 (18 %) of Model 3 cases. Only one respiratory case matched Model 4 (slow HR decline without inflection point), and this case had respiratory event onset ≥ 24 hours prior to arrest.

      Discussion

      In this study, we provide the most detailed evaluation to date of HR dynamics and changes in rhythm leading up to PEA/asystole IHCA from respiratory failure, using continuous electrocardiographic data. Consistent with clinical experience, we found that most PEA/asystole arrests caused by respiratory failure present initially with sinus tachycardia, developing along a similar time frame as worsening respiratory status. RV strain ECG pattern, when present, occurs towards the end of this tachycardic phase. Minutes prior to arrest, patients develop rapid slowing of the sinus rate that is mediated by sinus node suppression. In many cases, the sinus node arrests, which progresses into asystolic arrest when there are no escape rhythms, or which may be followed by escape rhythms. Atrial escape rhythms were the most common, followed by junctional, then ventricular. Vagally-mediated AV block also occurred in a small number of cases concomitantly with sinus rate slowing. Analysis of arterial line tracings, where present, showed that significant hemodynamic compromise had occurred by the time of onset of escape rhythms.
      Two studies have previously evaluated HR changes prior to IHCA. Attin et al. reviewed 10-second ECG strips hourly in 39 patients with structural heart disease and PEA/asystole arrest from all causes. They found that 82 % had HR decrease prior to arrest, mostly occurring in the last hour. Sinus bradycardia occurred in 47 %, junctional rhythm 28 %, atrial fibrillation 19 %, and third degree AV block in 6 %.
      • Attin M.
      • Feld G.
      • Lemus H.
      • et al.
      Electrocardiogram characteristics prior to in-hospital cardiac arrest.
      Bhalala et al. reviewed telemetry HR trends over 10 minutes prior to IHCA for 98 respiratory arrests and compared these with average HR over 8 hours prior to arrest. They found that 54 % of cases had antecedent bradycardia, which was independently associated with death prior to hospital discharge.
      • Bhalala U.S.
      • Bonafide C.P.
      • Coletti C.M.
      • et al.
      Antecedent bradycardia and in-hospital cardiopulmonary arrest mortality in telemetry-monitored patients outside the ICU.
      Our study significantly expands upon prior work by using continuous ECG monitoring over 3–24 hours pre-arrest to systemically evaluate sinus node behavior and progression of escape rhythms, with correlation to etiology, RV strain, and timing of respiratory event.
      Acidemia, hypercarbia, and hypoxia affect cardiovascular physiology differently,
      • Clowes G.H.
      • Hopkins A.L.
      • Simeone F.A.
      A Comparison of the Physiological Effects of Hypercapnia and Hypoxia in the Production of Cardiac Arrest.
      which explains some heterogeneity of the HR patterns observed. In animal experiments, hypoxia induces co-activation of sympathetic and vagal nerve activity, but leads to increased HR as sympathetic activity predominates. However, a background of sympathetic activity can intensify cardiac response to vagal stimulation,
      • Randall D.C.
      • Brown D.R.
      • Mcguirt A.S.
      • Thompson G.W.
      • Armour J.A.
      • Ardell J.L.
      Interactions within the intrinsic cardiac nervous system contribute to chronotropic regulation.
      potentially explaining the sudden decline in HR after prolonged period of sinus tachycardia, as well as development of AV block in some patients. Sympathetic outflow increases cardiac automaticity, predisposing to ectopic atrial, junctional, or ventricular rhythms,
      • Tisdale J.E.
      • Patel R.V.
      • Webb C.R.
      • Borzak S.
      • Zarowitz B.J.
      Proarrhythmic effects of intravenous vasopressors.
      • Peiss C.N.
      • Spurgeon H.A.
      Origin of initial escape beat during graded vagal stimulation.
      which can cause usurpation (often in the form of atrial tachycardia, which may trigger atrial fibrillation) or serve as an escape rhythm during sinus arrest. Although hypercapnia causes sympathetic activation, carbon dioxide locally suppresses cardiac pacemaker cells, resulting in an overall stable HR.
      • Fukuda Y.
      • Sato A.
      • Suzuki A.
      • Trzebski A.
      Autonomic nerve and cardiovascular responses to changing blood oxygen and carbon dioxide levels in the rat.
      Carotid body chemoreceptors respond to hypoxemia and acidemia to reflexively induce hyperventilation, bradycardia, and peripheral vasoconstriction, but ventilation modulates this pathway via pulmonary stretch and airway receptors, such that hypoventilation may cause exaggerated bradycardia.
      • De Burgh D.M.
      • Angell-James J.E.
      • Elsner R.
      Role of carotid-body chemoreceptors and their reflex interactions in bradycardia and cardiac arrest.
      Changes in the balance of these pathways and in sympathetic versus vagal predominance likely explain the progression of HR in the model presented.
      • Yang T.
      • Senturia J.B.
      • Levy M.N.
      Antecedent sympathetic stimulation alters time course of chronotropic response to vagal stimulation in dogs.
      The rapid reversibility of PEA from respiratory failure when identified and treated early also suggests that vagal influences may predominate during the early phases of arrest, rather than metabolic derangements which predominate later in the cardiac arrest course.
      We found that ECG signs of RV strain often present near the end of the tachycardia phase prior to HR slowing. RV strain has been associated with hypoxia and increased pulmonary vascular resistance;
      • Netzer N.C.
      • Strohl K.P.
      • Högel J.
      • Gatterer H.
      • Schilz R.
      Right ventricle dimensions and function in response to acute hypoxia in healthy human subjects.
      • Barman H.A.
      • Atici A.
      • Sahin I.
      • et al.
      Prognostic value of right ventricular strain pattern on ECG in COVID-19 patients.
      thus, it may serve as a surrogate for increased activation of pulmonary stretch receptors
      • Glick G.
      • Wechsler A.S.
      • Epstein S.E.
      • Lewis R.M.
      • Mcgill R.D.
      Reflex Cardiovascular Depression Produced By Stimulation Of Pulmonary Stretch Receptors In The Dog Reflex Cardiovascular Depression Produced by Stimulation of Pulmonary Stretch Receptors in the Dog with the technical assistance of.
      and pulmonary arterial baroceptors, which may also have varying effects on HR as well as systemic vascular tone.
      • Moore J.P.
      • Hainsworth R.
      • Drinkhill M.J.
      Reflexes from pulmonary arterial baroreceptors in dogs: interaction with carotid sinus baroreceptors.
      One hypothesis for why certain cases of respiratory failure did not lead to sinus tachycardia (15 % in this study) was suppression of the sympathetic response to hypoxia by other patient factors including medical conditions or medications, as suggested by a higher proportion of patients with no escape rhythm following sinus arrest (i.e. asystole) in the patients fitting Model 3. Additionally, two respiratory cases lacked the bradycardic phase after a period of sinus tachycardia. This could represent cardiac arrest from acute hemodynamic instability, possibly due to RV failure, which would not allow time for vagal influences to cause bradycardia. However, given the small sample sizes, such associations are difficult to prove.
      Several implications are worth noting. Because bradyarrhythmias commonly occur in PEA/asystole arrests from respiratory failure and are likely vagally-mediated, pacemakers are not warranted in survivors who do not have other pacing indications. Translationally, this work can be applied to early recognition of clinical deterioration from respiratory etiologies. Artificial intelligence has been used to develop prediction algorithms for sepsis
      • Fleuren L.M.
      • Klausch T.L.T.
      • Zwager C.L.
      • et al.
      Machine learning for the prediction of sepsis: a systematic review and meta-analysis of diagnostic test accuracy.
      and refine Early Warning Systems for IHCA based on vital signs abnormalities.
      • Kwon J.M.
      • Lee Y.
      • Lee Y.
      • Lee S.
      • Park J.
      An Algorithm Based on Deep Learning for Predicting In-Hospital Cardiac Arrest.
      The HR dynamics model presented here can be used to guide supervised machine learning to develop algorithms for incorporating telemetry data.
      • Johnson K.W.
      • Torres Soto J.
      • Glicksberg B.S.
      • et al.
      Artificial Intelligence in Cardiology.
      • Yang J.J.
      • Hu X.
      • Boyle N.G.
      • Do D.H.
      Novel Approaches to Risk Stratification of In-Hospital Cardiac Arrest.
      Improved understanding of “signature” findings of specific causes of IHCA can also guide management and appropriate workup for cardiac arrest etiology.

      Limitations

      There are several limitations to this study. First, the classification of etiology of arrest, type of arrest, and the timing of the respiratory event may be confounded by ambiguous clinical documentation, common in the often chaotic setting of IHCA management. This is somewhat alleviated by comprehensive chart review and use of continuous ECG data. Second, the noise and artifact in continuous ECG data may occasionally impact the classification of HR trends and rhythms. Third, we lacked complete data on the timing and use of medications such as inotropes, vasopressors, atropine, and beta blockers/calcium channel blockers that may confound HR patterns and progression of escape rhythms. Pre-arrest use of these medications and medical comorbidities may contribute to the observed inconsistencies between HR dynamics and course of respiratory events. Fourth, excluding atrial fibrillation, which often occurs in the setting of respiratory failure due to enhanced sympathetic tone, limits generalizability but allows more valid interpretation of sinus node function.

      Conclusions

      In this cross-sectional study, we found that most IHCAs of respiratory etiology follow a typical model of HR increase, likely due to physiologic compensation to hypoxia, followed by rapid HR decrease leading to PEA arrest, likely from the vagal effect of hypoxia and sinus node suppression from acidosis. The inciting respiratory event tends to coincide with HR increase, and the ECG signs of RV strain tend to appear during the tachycardic period. Additional studies should validate these patterns in larger patient samples. Understanding HR trends can aid clinical management and the development of prediction models for IHCA.

      CRediT authorship contribution statement

      Rongzi Shan: Investigation, Methodology, Formal analysis, Writing – original draft. Jason Yang: Methodology, Investigation. Alan Kuo: Methodology, Investigation. Randall Lee: Methodology, Investigation. Xiao Hu: Software, Writing – review & editing. Noel Boyle: Supervision, Writing – review & editing. Duc H. Do: Conceptualization, Investigation, Methodology, Supervision, Writing – review & editing.

      Declaration of Competing Interest

      The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: XH is an inventor of US Patent 9,936,923 and 9,600,990, assigned to the University of California, Los Angeles. Other authors have no conflicts of interest to disclose.

      Acknowledgments

      We thank the UCLA CTSI Bioinformatics group for assistance with cohort identification.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Holmberg M.J.
        • Ross C.E.
        • Fitzmaurice G.M.
        • et al.
        Annual Incidence of Adult and Pediatric In-Hospital Cardiac Arrest in the United States.
        Circ Cardiovasc Qual Outcomes. 2019; 12: e005580
        • Meaney P.A.
        • Nadkarni V.M.
        • Kern K.B.
        • Indik J.H.
        • Halperin H.R.
        • Berg R.A.
        Rhythms and outcomes of adult in-hospital cardiac arrest.
        Crit Care Med. 2010; 38: 101-108https://doi.org/10.1097/CCM.0B013E3181B43282
        • Morrison L.J.
        • Neumar R.W.
        • Zimmerman J.L.
        • et al.
        Strategies for improving survival after in-hospital cardiac arrest in the United States: 2013 consensus recommendations: A consensus statement from the American heart association.
        Circulation. 2013; 127: 1538-1563https://doi.org/10.1161/CIR.0b013e31828b2770
        • Do D.H.
        • Kuo A.
        • Lee E.S.
        • et al.
        Usefulness of Trends in Continuous Electrocardiographic Telemetry Monitoring to Predict In-Hospital Cardiac Arrest.
        Am J Cardiol. 2019; 124: 1149https://doi.org/10.1016/J.AMJCARD.2019.06.032
        • Attin M.
        • Feld G.
        • Lemus H.
        • et al.
        Electrocardiogram characteristics prior to in-hospital cardiac arrest.
        J Clin Monit Comput. 2015; 29: 385https://doi.org/10.1007/S10877-014-9616-0
        • Do D.H.
        • Yang J.J.
        • Kuo A.
        • et al.
        Electrocardiographic right ventricular strain precedes hypoxic pulseless electrical activity cardiac arrests: Looking beyond pulmonary embolism.
        Resuscitation. 2020; 151: 127-134https://doi.org/10.1016/J.RESUSCITATION.2020.04.024/ATTACHMENT/13BD5861-080F-43D0-A5AC-3E96F64A89C2/MMC1.DOCX
        • Netzer N.C.
        • Strohl K.P.
        • Högel J.
        • Gatterer H.
        • Schilz R.
        Right ventricle dimensions and function in response to acute hypoxia in healthy human subjects.
        Acta Physiol (Oxf). 2017; 219: 478-485https://doi.org/10.1111/APHA.12740
        • Li Y.
        • Li H.
        • Zhu S.
        • et al.
        Prognostic Value of Right Ventricular Longitudinal Strain in Patients With COVID-19.
        JACC Cardiovasc Imaging. 2020; 13: 2287-2299https://doi.org/10.1016/J.JCMG.2020.04.014
        • Barman H.A.
        • Atici A.
        • Sahin I.
        • et al.
        Prognostic value of right ventricular strain pattern on ECG in COVID-19 patients.
        Am J Emerg Med. 2021; 49: 1-5https://doi.org/10.1016/J.AJEM.2021.05.039
        • Skrifvars M.B.
        • Nurmi J.
        • Ikola K.
        • Saarinen K.
        • Castrén M.
        Reduced survival following resuscitation in patients with documented clinically abnormal observations prior to in-hospital cardiac arrest.
        Resuscitation. 2006; 70: 215-222https://doi.org/10.1016/J.RESUSCITATION.2006.01.002
        • Bhalala U.S.
        • Bonafide C.P.
        • Coletti C.M.
        • et al.
        Antecedent bradycardia and in-hospital cardiopulmonary arrest mortality in telemetry-monitored patients outside the ICU.
        Resuscitation. 2012; 83: 1106-1110https://doi.org/10.1016/j.resuscitation.2012.03.026
      1. Simon PM, Taha BH, Dempsey JA, Skatrud JB, Iber C, Taha H. Role of vagal feedback from the lung in hypoxic-induced tachycardia in humans.

        • Clowes G.H.
        • Hopkins A.L.
        • Simeone F.A.
        A Comparison of the Physiological Effects of Hypercapnia and Hypoxia in the Production of Cardiac Arrest.
        Ann Surg. 1955; 142: 446https://doi.org/10.1097/00000658-195509000-00011
        • Randall D.C.
        • Brown D.R.
        • Mcguirt A.S.
        • Thompson G.W.
        • Armour J.A.
        • Ardell J.L.
        Interactions within the intrinsic cardiac nervous system contribute to chronotropic regulation.
        Am J Physiol Regul Integr Comp Physiol. 2003; 285: 1066-1075https://doi.org/10.1152/ajpregu.00167.2003.-The
        • Tisdale J.E.
        • Patel R.V.
        • Webb C.R.
        • Borzak S.
        • Zarowitz B.J.
        Proarrhythmic effects of intravenous vasopressors.
        Ann Pharmacother. 1995; 29: 269-281https://doi.org/10.1177/106002809502900309
        • Peiss C.N.
        • Spurgeon H.A.
        Origin of initial escape beat during graded vagal stimulation.
        J Electrocardiol. 1975; 8: 25-29https://doi.org/10.1016/S0022-0736(75)80034-3
        • Fukuda Y.
        • Sato A.
        • Suzuki A.
        • Trzebski A.
        Autonomic nerve and cardiovascular responses to changing blood oxygen and carbon dioxide levels in the rat.
        J Auton Nerv Syst. 1989; 28: 61-74https://doi.org/10.1016/0165-1838(89)90008-8
        • De Burgh D.M.
        • Angell-James J.E.
        • Elsner R.
        Role of carotid-body chemoreceptors and their reflex interactions in bradycardia and cardiac arrest.
        Lancet. 1979; 313: 764-767https://doi.org/10.1016/S0140-6736(79)91218-2
        • Yang T.
        • Senturia J.B.
        • Levy M.N.
        Antecedent sympathetic stimulation alters time course of chronotropic response to vagal stimulation in dogs.
        Am J Physiol. 1994; 266https://doi.org/10.1152/AJPHEART.1994.266.4.H1339
        • Glick G.
        • Wechsler A.S.
        • Epstein S.E.
        • Lewis R.M.
        • Mcgill R.D.
        Reflex Cardiovascular Depression Produced By Stimulation Of Pulmonary Stretch Receptors In The Dog Reflex Cardiovascular Depression Produced by Stimulation of Pulmonary Stretch Receptors in the Dog with the technical assistance of.
        J Clin Invest. 1969; 48https://doi.org/10.1172/JCI106004
        • Moore J.P.
        • Hainsworth R.
        • Drinkhill M.J.
        Reflexes from pulmonary arterial baroreceptors in dogs: interaction with carotid sinus baroreceptors.
        J Physiol. 2011; 589: 4041https://doi.org/10.1113/JPHYSIOL.2011.210732
        • Fleuren L.M.
        • Klausch T.L.T.
        • Zwager C.L.
        • et al.
        Machine learning for the prediction of sepsis: a systematic review and meta-analysis of diagnostic test accuracy.
        Intensive Care Med. 2020; 46: 383-400https://doi.org/10.1007/S00134-019-05872-Y
        • Kwon J.M.
        • Lee Y.
        • Lee Y.
        • Lee S.
        • Park J.
        An Algorithm Based on Deep Learning for Predicting In-Hospital Cardiac Arrest.
        J Am Hear Assoc Cardiovasc Cerebrovasc Dis. 2018; 7https://doi.org/10.1161/JAHA.118.008678
        • Johnson K.W.
        • Torres Soto J.
        • Glicksberg B.S.
        • et al.
        Artificial Intelligence in Cardiology.
        J Am Coll Cardiol. 2018; 71: 2668-2679https://doi.org/10.1016/J.JACC.2018.03.521
        • Yang J.J.
        • Hu X.
        • Boyle N.G.
        • Do D.H.
        Novel Approaches to Risk Stratification of In-Hospital Cardiac Arrest.
        Curr Cardiovasc Risk Rep. 2021; 15: 1-12https://doi.org/10.1007/S12170-021-00667-7/FIGURES/3