Secondary Logo

Journal Logo

Basic Science Aspects

Evaluation of Heart Rate and Blood Pressure Variability as Indicators of Physiological Compensation to Hemorrhage Before Shock

Scully, Christopher G.*; Kramer, George C.; Strauss, David G.*

Author Information
doi: 10.1097/SHK.0000000000000340



The mammalian response to hemorrhage progresses through distinct phases characterized by transient hemodynamic compensation mechanisms that attempt to maintain cardiac output and blood pressure (1). Given continuous hemorrhage, the mechanisms will eventually be exhausted, leading to decompensated shock. Intersubject variability in the timing of the transition between phases given equivalent hemorrhage exists, making it difficult to know which compensating mechanisms are operating in an individual and how long those mechanisms may last (2). This leads to reliance on hypotension as a sign that interventions are required. However, waiting to perform interventions until significant hypotension is observed may result in a reduction of available interventions, the use of additional limited resources, and potentially missing the window when interventions may be most effective (3). New biomarkers are needed that can be continuously monitored to better describe an individual’s current response and predict future symptoms before significant changes in vital signs are observed.

Features from waveform analysis of the photoplethysmographic signal (4–6), the Shock Index (heart rate [HR]/systolic blood pressure [SBP]) (7), and noninvasive oxygenation measures (8–10) have been proposed as biomarkers for monitoring blood loss. Time-series analysis of physiological signals can be used to assess signatures generated by physiological processes. In this way, changes in physiological processes may be observed before a significant shift in vital signs has occurred. As a patient progresses through the stages of hemorrhage, different compensation mechanisms operate that may change the signal variability at different timescales. Time-frequency analysis can be used to monitor variability related to physiological processes operating at different timescales and quantify these changes across time. Such analysis could be clinically useful because it can be performed on routinely collected patient monitoring data. For example, withdrawal of high-frequency (0.15 – 0.4 Hz) HR variability has been associated with parasympathetic withdrawal, and increasing low-frequency (0.04 – 0.15 Hz) variability has been associated with sympathetic activity (11). This has led to HR and SBP variability measures being investigated as markers of the autonomic response to hemorrhage transitioning from parasympathetic to sympathetic activity and eventual sympathetic withdrawal before decompensation (12–14).

In the current study, we investigate the use of an index generated from time-frequency analysis representing changes in HR and SBP variability during compensation and decompensation phases of hemorrhagic shock in conscious sheep. We use this index to address the time-varying changes in HR and SBP and investigate if shifts in variability are specific to response phases of hemorrhage.


Experimental protocol

The experimental protocol was approved by the Institutional Animal Care and Use Committee of the University of Texas Medical Branch. Data were acquired from experiments performed for a previous study designed as a comparative analysis of closed-loop fluid resuscitation systems (15). All data analyzed in the current study were acquired before the closed-loop system was turned on.

Adult Merino cross range ewes, approximately 4 years old (N = 7; mean [SD], 37 [2.3] kg), were obtained from a USDA-licensed vendor (USDA license 74-B-0555; Talley Farms, Uvalde, Tex). Animals were acclimated at least 14 days before use in the vivarium and to confirm their health status. Animals were housed in small groups of two to four animals in the vivarium under 12-h light/dark cycles at a temperature of 23°C, where they had access to food and water ad libitum. Animals were surgically prepared at least 1 week before the experiment. Water and food were removed from the animals the night before surgery (∼18 h). Immediately before surgery, all animals received preventive intramuscular (i.m.) analgesia with long-acting (72 h) Buprenorphine i.m. (0.7 – 1 mg per sheep weighing ∼35 – 50 kg) and thereafter per need. Animals underwent anesthesia induced with 10 mg/kg ketamine i.m. followed by mask inhalation of halothane at 1% to 2%. Placement of an endotracheal tube allowed for maintenance of halothane anesthesia and controlled ventilation (Datex-Ohmeda) with FiO2 = 0.5. Initial tidal volume was set to 15 mL/kg and respiratory rate to 12 breaths/min. The FiO2 was adjusted to maintain SpO2 above 96%; tidal volume and respiratory rate were adjusted to maintain end-tidal CO2 between 35 and 40 mmHg. Animals received 1 L of intravenous (i.v.) Ringer’s lactate solution per hour during the surgical procedure. If hypotension occurred, the level of anesthesia was adjusted as necessary and additional Ringer’s lactate solution was infused. Femoral arterial, pulmonary arterial, and venous catheters were placed for hemorrhaging and pressure monitoring. A transit time Doppler flow probe (20 – 24 mm A-probe; Transonic Systems, Ithaca, NY) was placed on the common pulmonary artery for monitoring of continuous cardiac output. Catheter and probe connections were secured to wool on the back of the animal. A splenectomy was performed on each animal to eliminate splenic contraction and autotransfusion during hemorrhage (16). After instrumentation, animals were housed and monitored 5 to 7 days to allow for acclimation and recovery, during which they had access to food and water ad libitum.

Animals underwent one, two, or three hemorrhage experiments, each separated by at least 7 days, resulting in a total of 14 experiments. Both water and food were removed the evening before a study (∼18 h). Animals received preventative analgesics (Buprenorphine i.m.) before the experiments. Animals were moved to the monitoring laboratory at least 2 h before the start of the hemorrhage and positioned in a cage that they could look over but not turn around in because of placement of a bridal. The monitoring laboratory was dim, and investigators maintained a calm environment at all times.

Animals were conscious throughout the hemorrhage trial. Physiological measurements were recorded continuously for a baseline period followed by the initiation of a 25-mL/kgBody Weight hemorrhage occurring for 15 min. Monitoring continued after completion of the hemorrhage for an additional period of up to 15 min. After 30 min from the start of hemorrhage, animals were administered i.v. Ringer’s lactate solution therapy to restore blood pressure to a near-normal target using closed-loop fluid therapy as reported in Ying et al. (15). Data from the fluid therapy period was not used for the analysis in the present study. No signs of considerable stress on the animals were observed during the experiments. At the point of greatest blood loss, sheep were lethargic and would most often be lying down. After completion of all hemorrhage experiments for an animal, animals were euthanized with ketamine i.v. (25 mg/kg) followed by saturated KCl (1 mL/kg).

Continuous recordings of femoral arterial pressure, three-lead electrocardiogram (ECG), cardiac output (Flowmeter-Model 206; Transonic Systems, Inc., Ithaca, NY), pulmonary arterial pressure, central venous pressure, and a breathing trace using a resistive band placed around the thorax were made at 1,000 Hz (PowerLab; AD Instruments, Castle Hill, Australia). The ECG recordings from the three experiments either were not available (one experiment) or contained large segments with significant noise or a flat-line signal that did not allow identification of R-peak locations (two experiments). High-quality blood pressure recordings were available in all experiments. For this reason, HR computed from the arterial blood pressure is reported instead of that derived from the ECG. The pulmonary artery flow probe used for cardiac output monitoring was unavailable in two experiments, and respiration belt measurements were unavailable in three experiments.

Data analysis

We performed retrospective data analysis with MATLAB R2013A (The Mathworks, Inc., Natick, Mass). The ECG beat locations were identified using a previously reported nonlinear QRS detector (17). Arterial blood pressure pulse onsets were identified using a custom pulse detector, and onset times were used to determine the pulse-pulse interval and SBP sequences.

We used the continuous wavelet transform to determine HR and SBP variability from the respective pulse-pulse interval or beat-to-beat SBP sequence. Each sequence was first evenly sampled to 20 Hz using linear interpolation. The continuous wavelet transform was applied using a Morlet mother wavelet to produce a time-frequency representation as described in Appendix 1 (18, 19). The continuous wavelet transform outputs a complex coefficient at each time and frequency location. We investigated the wavelet power by squaring the absolute value of each complex coefficient.

We focused on frequencies from 0.02 to 1 Hz that include the traditional HR variability low (0.04 – 0.15 Hz)- and high (0.15 – 0.4 Hz)-frequency regions. For this study, a high-frequency band from 0.15 to 1 Hz was used because the breathing rate was observed to vary within this range during the study. A low-frequency band was set from 0.06 to 0.15 Hz, and a very low frequency band was assessed ranging from 0.02 to 0.06 Hz. Wavelet power was integrated across each frequency band at each time instant. We applied a 60-s moving average filter to generate a smoothed time-series of the continuous wavelet power in each of the three frequency bands (SBPHigh, SBPLow, SBPVeryLow). The SBPLow was divided by the sum of SBPHigh and SBPVeryLow, equation 2, to generate a single index (SBP-LF) representing the change in the balance of spectral power during the hemorrhage.

The HR-LF was generated using the same process from the wavelet transform of the pulse-pulse interval sequence. This index was selected to accentuate the reduced high-frequency variability and rise in low-frequency variability expected during hemorrhage.

A threshold was derived for each HR-LF and SBP-LF as the maximum value across all hemorrhage trials during the baseline period to identify if the indices increased during hemorrhage. The HR-LF and SBP-LF onset times were defined as the first time the index rose above this threshold. A single threshold determined from the baseline of all animals was selected as a conservative approach with no false positives during the baseline period. A second threshold was set for HR-LF and SBP-LF indices as the minimum across all baselines to recognize a second transition of signal variability. The HR-LF and SBP-LF fall times were defined as the time the index first crossed the second (minimum) threshold after the onset time. Onset and fall times are reported relative to the time when SBP first drops 20 mmHg and the time of the peak HR.


Heart rate rose during hemorrhage from 96.3 (22.2) beats/min to a peak of 176.0 (25.4) beats/min (mean [SD] across all trials) (Fig. 1A). The time the peak HR occurred from the start of hemorrhage varied among hemorrhage trials from a minimum of 5.3 min to a maximum of 22.1 min (11.7 [1.6] min). The rate of hemorrhage was approximately linear during the 15-min withdrawal period, so that the amount of blood lost at the time of the peak HR ranged from 8.9 to 25 mL/kgBody Weight (18.2 [1.5] mL/kgBody Weight). Timing of the peak HR between experiments represents variations in the timing of the compensatory mechanisms and specifically variations in how long the HR is able to rise as a compensatory response; this leads to differences in the timing of the drop in SBP (Fig. 1B). Aligning vital signs to the time of the peak HR for each trial demonstrates a more consistent response that can be separated into what we consider HR compensation and decompensation phases (Fig. 1, C and D).

Fig. 1:
Individual HR and SBP responses to hemorrhage. HR (A) and SBP (B) measurements during each hemorrhage from the baseline period until 5 min after completion of the hemorrhage. The vertical black line at 5 min indicates the start of hemorrhage, and the vertical black line at 20 min indicates the end of the hemorrhage. HR (C) and SBP (D) after aligning measurements to the peak HR (time of 0 min). The first rectangle highlights the peak compensation period, and the second rectangle highlights the initial decompensation period.

An example of the continuous R-R interval and SBP sequences and their corresponding time-frequency representations for a single experiment are shown in Figure 2. Light grey to white color indicates limited variability in the signal at that frequency (y axis) and time (x axis) location, whereas dark color indicates substantial variability. A consistent band of spectral power exists during baseline in the high-frequency region between 0.15 and 0.4 Hz, highlighted by the label HF in Figure 2C, in the R-R interval sequence. This corresponds to the breathing rate and represents the respiratory influence on the HR. During hemorrhage, the spectral power in this band diminishes as time approaches the peak HR, indicating a reduced respiratory influence on HR control. At approximately the 10-min mark (5 min into the hemorrhage), low-frequency power increases in both the R-R interval and SBP time-frequency representations, highlighted by labels LF, that continues until approximately the time of the peak HR. After the peak HR, a dark band is visualized in the very low frequency range, representing the slow oscillation that can be observed in both the R-R interval and SBP signals in Figure 2, A and B, respectively, after the peak HR.

Fig. 2:
R-R interval (A) and SBP (B) sequences during hemorrhage for one animal. The vertical black line indicates the start of hemorrhage, and the vertical gray line indicates the time of the peak HR. Time-frequency representations generated by the continuous wavelet transform of the R-R interval (C) and the SBP (D) sequences during hemorrhage, demonstrating the spectral power at each time and frequency location. The three dashed horizontal lines identify the boundaries of the high-frequency (0.15 – 1.0 Hz), low-frequency (0.06 – 0.15 Hz), and very low frequency (0.02 – 0.06 Hz) regions. The labels HF, LF, and VLF refer to high frequency, low frequency, and very low frequency, respectively, and are guides demonstrating when each region has elevated power.

The HR-LF and SBP-LF indices capture the shifting of spectral powers during hemorrhage (Fig. 3). Both are elevated while the HR is increasing. The indices begin to decrease shortly before the peak HR occurs and around the time that SBP and cardiac output begin to substantially decrease. Five of the seven sheep show similar responses to the example in Figure 3 (see Figure, Supplemental Digital Content 1, at, which plots HR-LF and SBP-LF responses for all hemorrhage trials). These shifts in spectral balances, represented by HR-LF and SBP-LF, were repeatable for sheep that underwent multiple hemorrhages. Two sheep (see parts D and F in Figure, Supplemental Digital Content 1, underwent multiple hemorrhage trials, with none presenting elevated HR-LF or SBP-LF responses during the hemorrhage, as shown for a single hemorrhage trial in Figure 4.

Fig. 3:
HR-LF (A) and SBP-LF (B) indices for the hemorrhage trial presented in Figure 2. C, HR (solid dark line), SBP (solid light line), and cardiac output (dashed line) during the trial. The vertical black line at 5 min indicates the start of hemorrhage, and the second vertical black line at 20 min indicates the end of hemorrhage. The first * indicates when HR-LF and SBP-LF cross the threshold, representing an elevated index, and the second * indicates when HR-LF and SBP-LF drop below the baseline minimum threshold.
Fig. 4:
HR-LF (A) and SBP-LF (B) indices for a hemorrhage trial that does not demonstrate an elevated response. C, HR (solid dark line), SBP (solid light line), and cardiac output (dashed line) during the trial. The vertical black line at 5 min indicates the start of hemorrhage, and the second vertical black line at 20 min indicates the end of hemorrhage.

The HR-LF and SBP-LF indices increased above the fixed threshold before a 20-mmHg drop in SBP for the five sheep that showed an increase during hemorrhage (Table 1). Onset times for these trials occurred 1 to 572 s (mean [SD], 168 [208] s) before a 20-mmHg drop in SBP. This indicates an elevated response in the indices before a large drop in SBP, but also that there is significantly variability in the timing of any warning. The elevated responses were limited to the HR compensation phase. After the peak HR, SBP-LF decreased below the minimum threshold (Table 1).

Table 1:
HR-LF and SBP-LF onset and fall times before the drop in SBP of 20 mmHg and peak HR


In this study, we characterized HR and SBP variability responses induced by a 25-mL/kgBody Weight hemorrhage in conscious sheep. The timing of the peak HR and transition from compensated to decompensated shock varied significantly between animals. The HR-LF and SBP-LF showed similar transient responses to hemorrhage and identified different balances of spectral power during baseline and HR compensation and decompensation. Low-frequency HR and SBP variability were elevated during the compensation period before falling during decompensation and may be useful for monitoring the current state of the physiological response to hemorrhage and potentially indicating an impending shift to decompensated shock. The similar HR-LF and SBP-LF responses indicate that the dynamics present in each may be from a common physiological source. Decreasing high-frequency power is associated with vagal withdrawal and increasing low-frequency power is associated with increased sympathetic activity (11). The benefit of using a single index generated from time-frequency analysis is that it captures transient changes from either vagal withdrawal or increased sympathetic activity.

Low-frequency HR variability results are consistent with a withdrawal of vagal activity and altered sympatho-vagal balance during the compensatory phases of hemorrhage (1), as well as with HR variability studies in humans (14). Batchinsky et al. (12) showed withdrawal of high-frequency R-R interval power in anesthetized sheep during a 40-mL/kgBody Weight (1 mL/kg per min) hemorrhage followed by withdrawal of R-R interval power in the 0.04- to 0.15-Hz range, representative of sympathetic withdrawal, after completion of the hemorrhage. Their study was performed in anesthetized animals with a slower hemorrhage rate than ours, and the analysis occurred at fixed intervals of blood loss while we presented continuous analysis that more closely mimics a clinical scenario. These experimental differences could explain why they showed a less dramatic rise in HR along with increased sympathetic activity throughout the 40-mL/kgBody Weight hemorrhage while we found a peak in the HR and low-frequency power withdrawal at approximately 20 mL/kgBody Weight that was dependent on the animal. The faster hemorrhage rate used in this study may alter the response of the compensation mechanisms (20, 21).

The SBP low-frequency variability increased during the peak HR compensation period and then decreased during the initial decompensation period. Figure 2 highlights the abrupt shift in the dominant frequencies occurring at the time of the peak HR. These observations suggest that the low- and very low frequency regions are representative of different physiological responses to hemorrhage. Madwed and Cohen (22) recognized differences between the 0.05- and 0.1-Hz oscillations after hemorrhage in anesthetized dogs. They hypothesized that the shorter time delay of the cardiac sympathetic nerves allows HR oscillations at 0.1 Hz to occur at earlier times during hemorrhage than peripheral sympathetic nerves triggering 0.05-Hz oscillations in SBP (22). This shift in the frequency of signal variability from 0.1 to 0.05 Hz results in our SBP-LF index dropping below the minimum threshold after the peak HR, allowing us to monitor the shift to decompensation. The 0.1-Hz oscillations in SBP have been reported to be inappropriate surrogates of sympathetic activity (23, 24); however, the dramatic temporal changes they undergo during peak compensation and initial decompensation suggest that tracking such dynamics could have merit for characterizing the physiological response when combined with additional physiological information. Rickards et al. (25) reported that elevated low-frequency oscillations in arterial pressure were associated with higher tolerance to central hypovolemia for subjects undergoing lower body negative pressure. Our results in this limited number of conscious sheep do not indicate any difference in the timing of the blood pressure response between animals that had strong low-frequency oscillations or those that did not. The hemorrhage with the shortest time to the peak HR and drop in SBP occurred for an animal with elevated HR-LF and SBP-LF, and one of the longest times occurred for an animal that did not show elevated HR-LF and SBP-LF. This difference may be a result of the fast hemorrhage conscious sheep model we used compared with lower body negative pressure.

Two animals did not show an elevated low-frequency response to hemorrhage. Individual differences has been a limiting factor for the expansion of HR variability analysis for hemorrhage monitoring, as previous studies have also shown individual HR variability measures to be poor predictors of central hypovolemia (26). Differences in the variability shifts between subjects limit the utility of such indices as a primary monitoring aid. However, observing trends in the changes of variability measures across time and combining this data with other clinical information and vital signs may aid in describing a patient’s current response to hemorrhage. Our results suggest that there may be a limited time span when HR-LF and SBP-LF are elevated because of the time-varying physiological changes during hemorrhage. This could enable a preventive clinical response before a significant change in traditional vital signs occurs or allow warning that a noticeable change in the patient’s status is likely to occur. The transient nature of the dynamics indicates the need for early and continuous vital sign monitoring, otherwise, the time period when such indices may be most revealing may not be observed. The advantage of generating new indices from analysis of HR and SBP data is that they can be computed from routinely collected patient monitoring data. Although HR and SBP variability responses were similar, monitoring HR dynamics may be more clinically relevant because of the availability of field ECG monitors that can continuously acquire the RR interval. Systolic blood pressure variability monitoring as shown here requires a continuous arterial blood pressure waveform from an arterial line; however, continuous noninvasive blood pressure monitors are increasingly available that may have utility for this purpose. Because of the limitations in intersubject variability, these markers may be more useful as an adjunct to monitoring but not a tool to be relied on. Alternatively, they may represent one of many features that could be combined using computational machine learning approaches to better characterize the overall response.


Our data were acquired from a previous study assessing resuscitation protocols and, therefore, has limitations associated with it. Our intention in the current study was to investigate changes in variability between the compensation and decompensation phases. A primary limitation of this study is that it was performed in heavily instrumented animals that may not represent a clinical scenario. There are a number of other confounding factors that could affect the variability measures that we could not assess with the current data such as the depth and recoverability of shock, the presence of painful stimuli, the effect of ongoing fluid resuscitation, and the rate of hemorrhage. Artifacts in HR and SBP signals also represent a challenge to implementing these techniques (27). One challenge of developing and validating new methods to monitor hemorrhage and predict decompensation is the limited models available to examine the effect of varying the injury. An advantage of our current study is that it is performed in a conscious large animal that may more closely resemble trauma scenarios than anesthetized animals. Future studies using conscious animal models may provide the opportunity to study the influences of injury variability on new biomarkers of hemorrhage to better understand their limitations in a way that cannot be done with clinical studies that rely on traditional blood donation protocols or lower body negative pressure.


Transient periods of physiologic compensation mechanisms were observed by HR and SBP variability markers during hemorrhage in conscious sheep. By aligning data to the peak HR, we assessed variability measures as indicators of the transient response to hemorrhage. This approach allows the study of measures that may be induced by specific compensation mechanisms that are known to have intersubject variability in their timing. By focusing the analysis on key vital sign events, we can identify biomarkers that represent a more personalized response to hemorrhage.


1. Schadt JC, Ludbrook J: Hemodynamic and neurohumoral responses to acute hypovolemia in conscious mammals. Am J Physiol 260 (2 Pt 2): H305–H318, 1991.
2. Schafer K, Van Sickle C, Hinojosa-Laborde C, Convertino VA: Physiologic mechanisms underlying the failure of the “shock index” as a tool for accurate assessment of patient status during progressive simulated hemorrhage. J Trauma Acute Care Surg 75 (2 Suppl 2): S197–S202, 2013.
3. Puyana JC, Pinsky M, Leuenberger U, Berkow J: Model-based prediction of autoregulatory exhaustion in response to lower-body negative pressure–induced shock. J Trauma Acute Care Surg 75 (2 Suppl 2): S190–S196, 2013.
4. McGrath SP, Ryan KL, Wendelken SM, Rickards CA, Convertino VA: Pulse oximeter plethysmographic waveform changes in awake, spontaneously breathing, hypovolemic volunteers. Anesth Analg 112 (2): 368–374, 2011.
5. Rickards CA, Vyas N, Ryan KL, Ward KR, Andre D, Hurst GM, Barrera CR, Convertino VA: Are you bleeding? Validation of a machine-learning algorithm for determination of blood volume status: application to remote triage. J Appl Physiol (1985) 116 (5): 486–494, 2014.
6. Scully CG, Selvaraj N, Romberg FW, Wardhan R, Ryan J, Florian JP, Silverman DG, Shelley KH, Chon KH: Using time-frequency analysis of the photoplethysmographic waveform to detect the withdrawal of 900 mL of blood. Anesth Analg 115 (1): 74–81, 2012.
7. Vandromme MJ, Griffin RL, Kerby JD, McGwin G Jr, Rue LW 3rd, Weinberg JA: Identifying risk for massive transfusion in the relatively normotensive patient: utility of the prehospital shock index. J Trauma 70 (2): 384–388; discussion 388–90, 2011.
8. Soller BR, Yang Y, Soyemi OO, Ryan KL, Rickards CA, Walz JM, Heard SO, Convertino VA: Noninvasively determined muscle oxygen saturation is an early indicator of central hypovolemia in humans. J Appl Physiol 104 (2): 475–481, 2008.
9. Soller BR, Ryan KL, Rickards CA, Cooke WH, Yang Y, Soyemi OO, Crookes BA, Heard SO, Convertino VA: Oxygen saturation determined from deep muscle, not thenar tissue, is an early indicator of central hypovolemia in humans. Crit Care Med 36 (1): 176–182, 2008.
10. Shackelford SA, Colton K, Stansbury LG, Galvagno SM Jr, Anazodo AN, DuBose JJ, Hess JR, Mackenzie CF: Early identification of uncontrolled hemorrhage after trauma: Current status and future direction. J Trauma Acute Care Surg 77 (3 Suppl 2): S222–S227, 2014.
11. Palazzolo JA, Estafanous FG, Murray PA: Entropy measures of heart rate variation in conscious dogs. Am J Physiol Heart Circ Physiol 274 (4): H1099–H1105, 1998.
12. Batchinsky AI, Cooke WH, Kuusela TA, Jordan BS, Wang JJ, Cancio LC: Sympathetic nerve activity and heart rate variability during severe hemorrhagic shock in sheep. Auton Neurosci 136 (1–2): 43–51, 2007.
13. Ji SY, Belle A, Ward KR, Ryan KL, Rickards CA, Convertino VA, Najarian K: Heart rate variability analysis during central hypovolemia using wavelet transformation. J Clin Monit Comput 27 (3): 289–302, 2013.
14. Zollei E, Paprika D, Makra P, Gingl Z, Vezendi K, Rudas L: Human autonomic responses to blood donation. Auton Neurosci 110 (2): 114–120, 2004.
15. Ying H, Bonnerup CA, Kirschner RA, Deyo DJ, Michell MW, Kramer GC: Closed-loop fuzzy control of resuscitation of hemorrhagic shock in sheep. Engineering in Medicine and Biology, 2002. Proceedings of the Second Joint 24th Annual Conference and the Annual Fall Meeting of the Biomedical Engineering Society EMBS/BMES Conference 2: 1575–1576, 2002.
16. Rafie AD, Rath PA, Michell MW, Kirschner RA, Deyo DJ, Prough DS, Grady JJ, Kramer GC: Hypotensive resuscitation of multiple hemorrhages using crystalloid and colloids. Shock 22 (3): 262–269, 2004.
17. Johannesen L, Galeotti L: Automatic ECG quality scoring methodology: mimicking human annotators. Physiol Meas 33 (9): 1479–1489, 2012.
18. Jimenez RF, Gunther B, Torres P: Time-frequency analysis of arterial pressure oscillations in anesthetized dogs: effects of standardized hemorrhages. Shock 15 (2): 143–150, 2001.
19. Torrence C, Compo GP: A practical guide to wavelet analysis. Bull Am Meteorol Soc 79 (1): 61–78, 1998.
20. Starc T, Stalcup SA: Time course of changes of plasma renin activity and catecholamines during hemorrhage in conscious sheep. Circ Shock 21 (2): 129–140, 1986.
21. Porter K, Ahlgren J, Stanley J, Hayward LF: Modulation of heart rate variability during severe hemorrhage at different rates in conscious rats. Auton Neurosci 150 (1): 53–61, 2009.
22. Madwed JB, Cohen RJ: Heart rate response to hemorrhage-induced 0.05-Hz oscillations in arterial pressure in conscious dogs. Am J Physiol 260 (4 Pt 2): H1248–H1253, 1991.
23. Taylor JA, Williams TD, Seals DR, Davy KP: Low-frequency arterial pressure fluctuations do not reflect sympathetic outflow: gender and age differences. Am J Physiol 274 (4 Pt 2): H1194–H1201, 1998.
24. Ryan KL, Rickards CA, Hinojosa-Laborde C, Cooke WH, Convertino VA: Arterial pressure oscillations are not associated with muscle sympathetic nerve activity in individuals exposed to central hypovolaemia. J Physiol 589 (Pt 21): 5311–5322, 2011.
25. Rickards CA, Ryan KL, Cooke WH, Convertino VA: Tolerance to central hypovolemia: the influence of oscillations in arterial pressure and cerebral blood velocity. J Appl Physiol 111 (4): 1048–1058, 2011.
26. Ryan KL, Rickards CA, Ludwig DA, Convertino VA: Tracking central hypovolemia with ecg in humans: cautions for the use of heart period variability in patient monitoring. Shock 33 (6): 583–589, 2010.
27. Sethuraman G, Ryan KL, Rickards CA, Convertino VA: Ectopy in trauma patients: cautions for use of heart period variability in medical monitoring. Aviat Space Environ Med 81 (2): 125–129, 2010.


The continuous wavelet transform provides a spectral representation of a signal (x) at each time (t) for a range of scales (s) that can be converted to frequency (f). It is defined as the following integral:

Ψ* represents the complex conjugate of the wavelet function that is shifted across time and translated across scales to determine the spectral representation at each time and scale location. We used the Morlet wavelet function defined as in (A2) with a center frequency ωo of 6.

The scale of the Morlet wavelet function can be used to determine the corresponding frequency as (19):

Wavelet power is then integrated across the scales corresponding to the appropriate frequency regions (i.e., low frequency from f1 of 0.06 to f2 of 0.15 Hz) at each time instant:


Hemorrhage; compensation; heart rate variability; blood pressure variability; HR — heart rate; SBP — systolic blood pressure

Supplemental Digital Content

© 2015 by the Shock Society