Increases in Heart Rate Variability Signal Improved Outcomes in Rapid Response Team Consultations: A Cohort Study

Background Reduced heart rate variability (HRV) indicates dominance of the sympathetic system and a state of “physiologic stress.” We postulated that, in patients with critical illness, increases in HRV might signal successful resuscitation and improved prognosis. Methods We carried out a prospective observational study of HRV on all patients referred to the rapid response team (RRT) and correlated with serial vital signs, lactate clearance, ICU admission, and mortality. Results Ninety-one patients were studied. Significantly higher HRV was observed in patients who achieved physiological stability and did not need ICU admission: ASDNN 19 versus 34.5, p=0.032; rMSSD 13.5 versus 25, p=0.046; mean VLF 9.4 versus 17, p=0.021; mean LF 5.8 versus 12.4, p=0.018; and mean HF 4.7 versus 10.5, p=0.017. ROC curves confirmed the change in very low frequencies at 2 hours as a strong predictor for ICU admission with an AUC of 0.772 (95% CI 0.633, 0.911, p=0.001) and a cutoff value of −0.65 associated with a sensitivity of 78.6% and a specificity of 61%. Conclusions Reduced HRV, specifically VLF, appears closely related to greater severity of critical illness, identifies unsuccessful resuscitation, and can be used to identify consultations that need early ICU admission.


Background
Extreme alterations in heart rate, blood pressure, and consciousness are parameters recognized as markers of "severe illness" and are used to mobilize rapid response teams or ICU consultations. RRT evaluation generally involves bedside assessment and resuscitation with eventual ICU admission if the patient does not stabilize. Prognostication in these patients is problematic. Early warning scores and, more recently, the quick Sequential Organ Failure Assessment (qSOFA) scores have been validated as strong predictors of outcome in septic patients and may be used to decide on early ICU transfer [1,2]. Lactic acidosis develops with hypoperfusion (shock). e aim of resuscitation is to normalize perfusion, which can be measured by clearance of lactate [3]. However in cases of mitochondrial disorders (Type B lactic acidosis) or with failure of usual clearance routes (renal or hepatic failure), physicians cannot rely on lactic acid clearance as a goal of resuscitation.
Fluctuations of the R-R interval between consecutive heartbeats as well as the oscillations between consecutive, instantaneous heart rates are conventionally known as heart rate variability (HRV) and are accepted as an indicator of the dynamic equilibrium between sympathetic and parasympathetic divisions of the autonomic nervous system [4]. Time domain parameters measure HRV over a given period.
ese are calculated based on the time interval between successive normal sinus heart beats (NN interval which is expressed in milliseconds). e variability in these NN intervals can be expressed by several di erent parameters. SDNN refers to the standard deviation of the NN interval. SDANN is obtained by averaging NN intervals for each 5-minute segment and calculating its standard deviation. SDNN index or ASDNN is the average of the SDNN of each 5-minute segment over 24 hours. rMSSD (root mean square of successive di erences) is calculated by squaring the di erence in milliseconds between successive NN intervals, averaging it, and then taking its square root. pNN50 is the percentage of successive NN intervals that di er by more than 50 milliseconds (ms).
Heart rate variability follows cyclical patterns. Di erent physiological parameters can cause these cyclical changes albeit at di erent cycle lengths or frequencies. Oscillations in heart rate due to respiration, for example, occur in a rhythmic fashion over a cycle with frequency between 0.15 and 0.40 Hz (high frequency (HF)). Contribution of di erent factors a ecting the heart rate variability can be calculated by analyzing the heart rate variability at very low frequency (VLF, 0.0033-0.04 Hz), low frequency (LF, 0.04-0.15 Hz), and high frequency (HF, 0.15-0.4 Hz).
Heart rate variability (HRV) has been described to indicate a balance between the sympathetic and parasympathetic nervous systems with reduced HRV, indicating dominance of the sympathetic system and a state of "physiologic stress." In patients after myocardial infarction, reduced HRV is predictive of cardiac mortality [5] and sudden cardiac death [6]. Similarly, reduced HRV in septic patients presenting to the emergency department has been linked to higher mortality and greater likelihood of progression to shock [7][8][9] and, in ICU patients, with higher organ failure scores [10]. In patients surviving cardiac arrest, reduced HRV appears to be predictive of early mortality [11]. Heart rate variability is evaluated as both time domain and frequency domain measures [4,12]. Time domain parameters estimate HRV over a 24-hour period. Frequency domain parameters can be measured hourly and as a mean over the monitoring period. Previous investigators have found that HRV measurements in both domains correlate with poor outcomes [7][8][9]11].
Patients seen in RRT consultation need early markers that would indicate either stabilization or deterioration requiring ICU admission. HRV data may be useful in such a situation as it can re ect improvement or worsening of illness, but there is no literature addressing this particular population. We postulated that an increase in HRV with resuscitation may signal stabilization and may serve as a guide of clinical recovery along with the more conventional lactate and hemodynamic variables such as blood pressure and heart rate. e objectives of our study were to study HRV patterns in patients referred for critical illness and to determine if trends in HRV variables could identify patients not responding to resuscitation and therefore requiring ICU admission.

Methods
is was a prospective, observational study of consecutive rapid response team (RRT) consultations carried out from June 2015 to May 2016. Patients were evaluated and resuscitated by the RRT/ICU consultation teams as per usual routine. Adult patients without atrial or ventricular arrhythmias or previous pacemaker or internal cardiac debrillator insertion were included in the study. Written consent was obtained for participation and monitoring. In addition to ongoing resuscitation, continuous EKG was recorded by a Holter monitor attached for 24 hours. Holter recordings were analyzed by using MARS Holter monitoring system (GE Healthcare) and proprietary software. All studies were manually scanned to ensure sinus rhythm, and that all abnormal beats were placed in appropriate bins. Recordings with atrial brillation were excluded from the analysis. Heart rate variability analysis was according to guidelines established by the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology [4]. Heart rate variability was measured as time domains measured over 24 hours (SDNN, ASDNN, rMSSD, pNN50%, SDANN, and mean NN) and frequency domains measured hourly (very low frequency (VLF), low frequency (LF), high frequency (HF), and low/high ratio) as well as a mean value taken over the 24-hour monitoring period. Changes from baseline were calculated. Frequency domains were reported as power in ms 2 . As per institutionapproved criteria, patient improvement and stabilization was de ned by normalization or ≥10% reduction in serum lactate levels, ≥15% reduction in heart rate, or increase in systolic blood pressure over the rst few hours of resuscitation or weaning o vasopressors and clinical judgment. ese patients would not routinely be admitted to the ICU. Patients not improving were admitted to the ICU as decided by the treating RRT/ICU consulting physician. Holter monitoring was continued in all patients for the prespeci ed 24-hour period. Patients could be treated at the bedside by the RRT using institution-approved therapeutic interventions that include dopamine up to a dose of 5 mcg/kg/minute administered through a peripheral line, any crystalloid uids or 5% albumin boluses, noninvasive ventilation by face mask, antibiotics, furosemide, endotracheal intubation, and emergency medications from the crash cart (epinephrine, atropine, naloxone, fentanyl, and 20% glucose). All patients were followed for 72 hours after enrollment with serial measurements of physiological data, biochemical data, and for outcomes (ICU admission and mortality) until day 28 from enrollment. HRV variables were collected both before and after ICU admission. Our study was designed to assess if measurement of HRV and change from baseline could assist clinicians in prognostication. However, as it was not known how soon during resuscitation a change may be seen, we measured hourly HRV for a full 24-hour period regardless of whether the patient was or was not admitted to the ICU.
Heart rate variability amongst the entire patient cohort was SDANN/ms 56 (37.7-91.  Table 2).  HRV was signi cantly higher in patients who did not need ICU admission.
ese patients also showed signi cantly greater hourly improvements in VLF during the resuscitative period. A signi cant divergence was identi ed in the VLF as early as 2 hours into resuscitation (Table 3 and Figure 1).
Patients who survived the ICU admission had signicantly lower APACHE II and SOFA scores with signi cantly greater improvements in serum lactate. HRV was signicantly higher (mean VLF 5.5 (4.2, 11.7) versus 11.9 (8. Patients surviving to 28 days had signi cantly higher HRV; time domains (ASDNN 14 versus 24, p � 0.012; rMSSD 13 versus 19, p � 0.037; mean NN 581 versus 685, p � 0.004) and frequency domains (mean VLF 5.7 versus 11.7, p � 0.005; mean LF 4.6 versus 7.5, p � 0.038) had signi cantly lower SOFA scores, heart rates, serum lactate levels, and greater lactate clearances at 12 and 24 hours. No signi cant di erences were found in lactate clearance at 4 and 6 hours and in baseline mean arterial pressure between survivors and nonsurvivors ( Figure 3).

Discussion
In this study, we demonstrate that heart rate variability has clinical utility in the assessment and resuscitation of critically ill patients. Patients in whom the critical illness stabilized tended to have a higher heart rate variability and showed greater hour-by-hour increases, compared to those who failed to improve and had to be admitted to the ICU. Survival di erences were also predictable by mean and hourly heart rate variability. ough our study is amongst the rst to look at HRV trajectories in RRT consultations, low HRV has been previously described as a marker of greater illness and worse outcomes. In 1994, Tsuji et al. [13] reported that, of the 736 original subjects in the Framingham Heart Study, analysis of the rst 2 hours of ambulatory ECG revealed a signi cant association between all-cause mortality and the very low frequency (p < 0.0001), low frequency (p < 0.0001), high frequency (p � 0.0014), total power (p < 0.0001), and the SDNN (p � 0.0019). is was followed by reports of associations between low HRV and sudden cardiac death [14], stroke outcomes [15,16], prognosis in heart failure [17] and risk of cardiac arrest [18].
Amongst the ICU population, Schmidt et al. [19], in an observational study of 90 patients with 24-hour ECG monitoring, described signi cantly reduced HRV in patients with multiorgan failure and lnVLF as an independent predictor of 28-day mortality (AUC 0.68, 95% CI 0.55, 0.8).
Papaioannou et al. [20] measured HRV as variance (exponent alpha2) and approximate entropy (ApEn) by analyzing daily heart rates recorded from bedside monitors. ey described lower ApEn in nonsurvivors compared to survivors (0.53 ± 0.25 versus 0.62 ± 0.23, p � 0.04) and higher variance and ApEn in patients with low SOFA scores (0.47 ±0.51 versus 0.10 ± 0.65, p < 0.001; 0.67 ± 0.28 versus 0.49 ±0.24, p < 0.001). In 2016, Bishop and coworkers [21] compared HRV with APACHE II scoring in 55 ICU patients. ey described a robust independent predictive ability for 30-day mortality with OR 0.6 and 95% CI 0.396, 0.911. Similar to these reports, we have also demonstrated that reduced HRV is associated with lower ICU and 28-day survival.
e VLF band falls between 0.0033 and 0.04 Hertz in the HRV spectrum. Kember et al. [22,23] demonstrated that the VLF band is generated by the heart's intrinsic nervous system and is modulated by e erent sympathetic activity. It is postulated that the activity of the autonomic nervous system, especially regulation of the reninangiotensin system and thermoregulation, may contribute to this HRV band [24]. e VLF band has been described to re ect a high in ammatory state [25,26] and has speci cally been described to have the highest association with adverse outcomes [13,17,19,21]. erefore, the VLF band should be considered an intrinsic rhythm necessary to health and well-being. In our study, we also found the VLF band to be strongly associated with both unsuccessful resuscitation and increased mortalities. Additionally, we found the VLF to have predictive ability above that of lactate clearance. e strengths of our study are that our study population was a diverse group similar to most populations that conventional RRT consultation would comprise. We were also able to demonstrate that vasopressor use had no e ect on HRV. Limitations include the inability to generalize our results to patients with pacemakers or atrial arrhythmias.
We set out to demonstrate was that, in patients who met criteria for RRT consultation (and of whom 85% were deemed sick enough to require ICU admission), HRV variables showed signi cant initial di erences and diversion between patients who stabilized with minimal resuscitation by the RRT team and did not require ICU admission compared to those who were admitted to the ICU. Since most patients were admitted within a mean period of 3.6 ± 2.3 hours (range 1-12), we also assessed HRV variables as predictors of 28-day survival, that is, whether the trajectories of HRV variables were an indicator of response. We con rmed this by observing signi cantly di erent trajectories of change in HRV variables between survivors and those who died. We were also able to identify a cuto value of the VLF variable that separated these two groups.
rough our observation of the di erences in HRV variables amongst RRT consults requiring ICU admission and in 28day survival of the entire study cohort, we surmise that, in  the future, these variables and the cuto values maybe helpful to clinicians to predict outcomes. We have shown that "real-world" heart rate variability monitoring is a practical tool that can be used to assess the adequacy of resuscitation and improvement in short-term hourly intervals and allows for rapid assessment in RRT/ICU consultations.
e growing availability of smart phone applications that measure HRV may allow RRTphysicians to perform bedside HRV monitoring. Certainly, validation of these applications with comparison to EKG recordings appears to be the next step.

Conclusions
Reduced HRV, speci cally VLF, appears closely related to greater severity of critical illness, identi es unsuccessful resuscitation, and can be used to identify consultations that need early ICU admission. Based on our results, prognostication using real-time HRV assessment at the bedside is a promising next step.

HRV:
Heart rate variability RRT: Rapid response team SOFA score: Sequential Organ Failure Assessment score qSOFA score: Quick Sequential Organ Failure Assessment score SDNN: Standard deviation of NN intervals SDANN: Standard deviation of the average NN intervals rMSSD: Root mean square of successive di erences pNN50%: Proportion of NN50 divided by total number of NNs VLF: Very low frequency LF: Low frequency HF: High frequency L/H: Low/high ratio AUC: Area under the curve ROC: Receiver-operating characteristic OR: Odds ratio CI: Con dence intervals.
Ethical Approval e study protocol was approved by the King Faisal Specialist Hospital and Research Centre, O ce of Research A airs, Research Advisory Committee (RAC Proposal no. 2151069). e study was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.