Beacon 3 is a system which provides advice on three ventilator settings, Inspiratory oxygen (FIO2), tidal volume (VT) or inspiratory pressure (Ps, Pc), and respiratory frequency (Rf). Beacon 3 adapts to the ventilation mode of the patient, providing advice from the acute phase where the patient may be highly sedated, through to the latter phases of weaning where the patient’s own spontaneous breathing is supported. The core of the Beacon 3 system is a set of physiological, mathematical models which predict what is likely to happen to the patient. Using these, Beacon 3 recommends a ventilator strategy which is most likely to be beneficial to the individual patient.
On recommending advice Beacon 3 displays the following advice window
This window illustrates the current ventilator settings and those advised by Beacon 3 for a patient in pressure controlled mo nde. The system recommends increasing FIO2 from 43% to 49%, reducing Pc from 28 cmH20 to 25 cmH20 and increasing Rf from 15 breaths/min to 18 breaths/min. One of the unique features of Beacon 3 is that this advice can be understood. This understanding comes from the following simulation screen which illustrates both ventilator strategies and their likely effect on the patient.
The left hand side of the screen illustrates ventilator settings, these being the current (FIO2 = 43%, Pc = 28 cmH20, Rf = 15 breaths/min), a central wheel which illustrates the advice provide by Beacon or simulation of the user ( FIO2 = 49%, Pc = 25 cmH20, Rf = 18 breaths/min) and the target settings (FIO2 = 49%, Pc = 16 cmH20, Rf = 27 breaths/min). The right hand side of the screen illustrates the physiological model simulated response to these strategies for this patient presented on the “preference zone” hexagon. Three shapes are drawn on this hexagon representing the current (blue), advice/simulated (white) and target (shaded) strategy. Each of these shapes represents the balance between the conflicting goals of ventilator strategy with these goals represented at each corner of the hexagon. Shapes located at the center rather than the outer rim of the hexagon represent benefit to the patient with this reflected in the color coding of the hexagon. The center is colored green gradually progressing to red at the outer rim. The three factors at the top represent over-ventilation, and those at the bottom under-ventilation. This provides the clinician with a single description of the likely effects of ventilator strategies.
For this patient the system suggests reducing pressure control from 28 to 25 cmH20, and increasing frequency from 15 to 18 breaths/min, plus a small increase in inspiratory oxygen. This advice is a clinical step towards the target settings. The potential consequence of adopting the advice can be seen by comparing the blue (current) and white (advised) shapes on the preference zone with the eventual target represented by the shaded area. A fall in respiratory pressure results in the white shape moving away from lung trauma. However, reducing pressure and hence tidal volume results in an increased risk of acidosis. The advice proposed by Beacon 3 is therefore one which will encourage permissive hypercapnia in this patient with acute lung damage.
The placement of the preference zone shapes on the 6 axes of the hexagon depends upon the values of physiological variables, for example placement on the acidosis axis depends upon the arterial pH, and placement on the lung trauma axis depends upon tidal volume, inspiratory plateau pressure and respiratory frequency. These values can be displayed by touching the screen, with the value of pH for the advice, i.e. pH= 7.368, shown above. These values come from simulations performed using the physiological models, with these models tuned to represent the individual patient, as will be presented later.
A further example is shown below for a patient in pressure support ventilation mode. The system recommends increasing FIO2 from 34% to 38%, and reducing Ps from 13 cmH20 to 10 cmH20. This patient is in the latter phase of ventilator management and therefore a zoom has been performed on the preference window and only the green area is shown.
As with the previous example, the potential consequence of adopting the advice can be seen by comparing the blue (current) and white (advised) shapes on the preference zone. The preference zone for patients in support modes of ventilation looks a little different from those in control modes. In support mode the risk of muscle atrophy, due to over-ventilation must be weighed against the risk of respiratory stress and fatigue on under ventilation. In this patient a reduction of Ps from 13 cmH20 to 10 cm H2O is likely to result in a fall in inspiratory pressure and an increase in respiratory frequency. Lower inspiratory pressure results in the white shape moving away from lung trauma. Increase in frequency both reduces the risk of muscle atrophy but at the same time increases the risk of respiratory stress, the resulting compromise being these pushing the patient towards weaning.
The core of the Beacon Caresystem is a number of linked mathematical models of physiology as illustrated in the figure below. Input to these models are measurements, including: end tidal O2 and CO2, SpO2, respiratory flows and pressures and user input values of arterial blood gas, along with the ventilator’s current settings.
Output from the system is twofold. Beacon provides advice on the optimal ventilator strategy, finding the strategy which is most likely to benefit the individual patient, as described previously. In addition, by tuning the mathematical models to measured data, Beacon “learns” about the patient, described the individual in terms of their: pulmonary gas exchange, lung mechanics, metabolism, circulation, the properties of the blood, and the patient’s respiratory drive. This learning allows for “intelligent monitoring” of the patient in two ways. It enables the system to “direct” the clinician to situations where further measurements, for example blood gas, are required to understand the patient. It also allows the system to alarm the clinician on detrimental physiological state in the patient. These functionalities – known as “directed measurement” and “intelligent alarms” respectively, are the basis for the “intelligent monitoring strategy of Beacon 3, described further here.
Physiological Model details
The following figure illustrates the physiological model components included in Beacon 3.
The model of Pulmonary Gas Exchange allows simulation of the effects of changes in ventilator settings on arterial blood gasses. By incorporating variables describing Metabolism (VO2, VCO2) this model can also describe the relationship between inspiratory and expiratory values of O2 and CO2. Full details of the model and its scientific background will later be described below.
The models of Acid-base chemistry of blood, interstitial fluid and tissues allow representation of how changes in arterial blood gases effect pH, bicarbonate and complete acid-base and oxygenation status. Full details of the model and its scientific background will later be described below.
The model of Acid-base chemistry of cerebral spinal fluid (CSF) allows calculation of pH in the CSF from the arterial acid-base status, an essential input to central chemoreceptor respiratory drive. Full details of the model and its scientific background will later be described below.
The model of Respiratory drive allows calculation of the expected alveolar ventilation resulting from chemoreceptor response to arterial blood and CSF acid-base and oxygenation status. Full details of the model and its scientific background will later be described below.
The model of Muscle function allows calculation of the alveolar ventilation resulting from the ability of the muscles to respond to the respiratory drive. Full details of the model and its scientific background will later be described below.
The model of Ventilation allows calculation of the alveolar ventilation resulting from ventilator settings, lung mechanics and serial dead space. Full details of the model and its scientific background will later be described below.
Beacon 3 provides two standard monitoring facilities as well as an intelligent monitoring strategy.
The above window illustrates the standard monitoring window included in Beacon 3. It can be customized by the user to include any of the following variables measured continuously by the Beacon Caresystem. These include pulse oximetry oxygen satuation (SpO2), inspiratory (FiO2, FiCO2) or expiratory (FeO2, FecO2) oxygen and carbon dioxide fractions, and calculated oxygen consumption (VO2) or carbon dioxide production (VCO2).
In addition Beacon 3 provides a historical monitoring view of the data as illustrated below
This view illustrates measured continuous variables, plus the measurement actions performed related to ventilator management. These include arterial blood gas sampling, measurement of pulmonary gas exchange or measurement of cardiac output. A full description of all data associated with the individual action can be accessed by clicking on the symbol marked on the action line.
Beacon 3 uses the current measurements and ventilator settings along with the physiological mathematical models to “learn” the patient’s physiological condition in terms of their: pulmonary gas exchange, lung mechanics, metabolism, circulation, the properties of the blood, and the patient’s respiratory drive.
This learning allows for “intelligent monitoring” of the patient in two ways. It enables the system to “direct” the clinician to situations where further measurements, for example blood gas, are required to understand the patient. It also allows the system to alarm the clinician on detrimental physiological state in the patient. These functionalities – known as “directed measurement” and “intelligent alarms” respectively, are the basis for the “intelligent monitoring strategy of Beacon 3
The figure below illustrates the advice provided by Beacon 3 for a patient on pressure support ventilation. As the core of the Beacon Caresystem is the physiological models, then it is possible to assess when the models do not simulate the patient correctly and hence the need for further measurements to understand the patient. The example of reducing opioid therapy is given below.
Reducing opioid therapy often causes an increase in the patient’s respiratory drive. An increase in respiratory drive will mean that the patient has a greater alveolar ventilation, and lower end tidal CO2 than predicted by the physiological models included in the Beacon Caresystem. This is recognized by the system and the user is directed to measure an arterial blood gas, with the following screen displayed.
The Beacon Caresystem uses the measured arterial acid-base status and alveolar ventilation to re-tune the model estimating the patient’s new respiratory drive. The advice provided by the Beacon Caresystem will then reflect this new respiratory drive, potentially reducing support to allow the patient greater control. The ability of the Beacon Caresystem to identify when the models are a poor description of patient state can direct the user both to measurement of blood gas and to measurement of pulmonary gas exchange.
Beacon 3 uses the current measurements and ventilator settings along with the physiological mathematical models to “learn” the patient’s physiological condition in terms of pulmonary gas exchange, lung mechanics, metabolism, circulation, the properties of the blood, and the patient’s respiratory drive. These values also allow the system to automatically notify the clinician when the patient’s state is changing. These notifications are Beacon 3’s “intelligent alarms”. They are intelligent as, unlike other alarms, they are not based upon single measurements such a SpO2, but rather upon deeper physiological changes in the patient’s state such as pulmonary function, metabolism, blood and respiratory drive. These are beneficial as alarms based on single measurement variables are a poor help in aiding the clinician in understanding the deeper cause of the problem. For example, a decrease in oxygenation of arterial blood can be due to numerous causes. The patient may have become active or developed a fever, in which case metabolic demand may have increased. The patient’s lung function may have deteriorated such that oxygen transports less freely from the lungs to blood, or it may just be the case that the clinician has inadvertently reduced inspired oxygen to a level where arterial values are compromised. Measurements and alarms indicating reduced arterial oxygenation are therefore useful but are not the complete picture. The ability of Beacon 3 to provide deeper “intelligent alarms” completes the monitoring picture and provides valuable support to the clinician.
Related to the implementation and evaluation of Beacon 3
Karbing DS, Allerød C, Thomsen LP, Espersen K, Thorgaard P, Andreassen S, Kjærgaard S, Rees SE. Retrospective evaluation of a decision support system for controlled mechanical ventilation. Med Biol Eng Comput. 2012 Jan;50(1):43-51.
Rees SE. The Intelligent Ventilator (INVENT) project: The role of mathematical models in translating physiological knowledge into clinical practice. Computer Methods and Programs in Biomedicine. 2011, Supplement, Vol 4. S1-S29.
Allerød C, Karbing DS, Thorgaard P, Andreassen S, Kjærgaard S, Rees SE. Variability of preference towards mechanical ventilator settings: a model based behavioral analysis. J Crit Care. 2011 Dec;26(6):637.e5-637.e12.
Karbing DS, Allerød C, Thorgaard P, CariusAM, FrilevL, Andreassen S, Kjærgaard S, Rees SE. Prospective evaluation of a decision support system for setting inspired oxygen in intensive care patients. Journal of Critical Care, 2010, 25(3):367-74.
C Allerød, S.E Rees, B.S Rasmussen, D.S Karbing, S Kjærgaard, P Thorgaard, S Andreassen. A Decision Support System for suggesting ventilator settings: Retrospective evaluation in cardiac surgery patients ventilated in the ICU. Computer Methods and Programs in Biomedicine, 2008, vol. 92, nr. 2, s. 205-212
S.E Rees, C Allerød, D Murley, Y Zhao, B.W Smith, S Kjærgaard, P Thorgaard, S Andreassen. Using physiological models and decision theory for selecting appropriate ventilator settings. Journal of Clinical Monitoring and Computing, 2006; Dec;20(6):421-429.
Related to the mathematical model of gas exchange
Karbing DS, Kjærgaard S, Andreassen S, Espersen K, Rees SE. Minimal model quantification of pulmonary gas exchanges in intensive care patients. Med Eng Phys. 2011 Mar;33(2):240-8.
S.E Rees, S Kjærgaard, S Andreassen, G. Hedenstierna. Reproduction of inert gas and oxygenation data – a comparison of the MIGET and a simple model of pulmonary gas exchange. Intensive Care Medicine 2010, 36:2117-2124.
D.S Karbing, S Kjaergaard, B.W Smith, K Espersen, C Allerod, S Andreassen, S.E Rees. Variation in the PaO2/FiO2 ratio with FiO2: Mathematical and experimental description, and clinical relevance. Critical Care. 2007 ; Vol. 11, No. 6.
B.S Rasmussen, H Laugesen, J Sollid, J Grønlund, S.E Rees, E Toft, J Gjedsted, C Dethlefsen, E Tønnesen. Oxygenation and release of inflammatory mediators after off-pump compared to after on-pump coronary artery bypass surgery. Acta Anaesthesiologica Scandinavica, 2007, 51(9):1202-10.
S.E Rees, S Kjærgaard, S Andreassen, G. Hedenstierna. Reproduction of MIGET retention and excretion data using a simple model of gas exchange in lung damage caused by oleic acid infusion. Journal of Applied Physiology, 2006 Sep;101(3):826-32.
B.S Rasmussen, J Sollid, S.E Rees, S Kjærgaard, D Murley, E Toft. Oxygenation within the first 120 h following coronary artery bypass grafting. Influence of systemic hypothermia (32 degrees C) or normothermia (36 degrees C) during the cardiopulmonary bypass: a randomized clinical trial. Acta Anaesthesiologica Scandinavica. 2006 Jan;50(1):64-71.
S Kjærgaard, S.E Rees, J Grønlund, E.M Malte, P Lambert, P Thorgaard, E Toft, S Andreassen. Hypoxaemia after cardiac surgery: Clinical application of a model of pulmonary gas exchange. European Journal of Anaesthesiology. 2004 Apr;21(4):296-301.
S Kjærgaard, S Rees, J Malczynski, J.A Nielsen, P Thorgaard, E Toft, S Andreassen. Non-invasive estimation of shunt and ventilation-perfusion mismatch. Intensive Care Medicine 2003 May;29(5):727-34.
S Kjærgaard, S.E Rees, J.A. Nielsen, M Freundlich, P Thorgaard and S Andreassen. Modelling of hypoxaemia after gynaecological laparotomy. Acta Anaesthesiologica Scandinavica 2001 Mar;45(3):349-356.
Thomsen LP, Karbing DS, Smith BW, Murley D, Weinreich UM, Kjærgaard S, Toft E, Thorgaard P, Andreassen S, Rees SE. Clinical refinement of the automatic lung parameter estimator (ALPE). J Clin Monit Comput. 2013, June 27(3), 341-50.
S.E Rees, S Kjærgaard, P Thorgaard, J Malczynski, E Toft, S Andreassen. The Automatic Lung Parameter Estimator (ALPE) system: Non-invasive estimation of pulmonary gas exchange parameters in 10-15 minutes. Journal of Clinical Monitoring and Computing, 2002, Vol 17, No.1, pp 43-52.
Related to the mathematical model of acid-base chemistry of blood, interstitial fluid and tissues
S.E Rees, E Klæstrup, J Handy, S Andreassen, S.R Kristensen. Mathematical modelling of the acid-base chemistry and oxygenation of blood – A mass balance, mass action approach including plasma and red blood cells. European Journal of Applied Physiology, 2010 Feb;108(3):483-94
S.E Rees, S Andreassen. Mathematical models of oxygen and carbon dioxide storage and transport: The acid-base chemistry of blood. Critical Reviews in Biomedical Engineering, 2005; 33(3):209-64.
S Andreassen, S E Rees. Mathematical models of oxygen and carbon dioxide storage and transport: Interstitial fluid and tissue stores and whole body transport. Critical Reviews in Biomedical Engineering, 2005; 33(3): 265-98.
Related to the mathematical model of respiratory drive
Larraza S, Dey N, Karbing DS, Jensen JB, Nygaard M, Winding R, Rees SE.A mathematical model approach quantifying patients’ response to changes in mechanical ventilation: Evaluation in pressure support. J Crit Care. 2015, in press
Larraza S, Dey N, Karbing DS, Jensen JB, Nygaard M, Winding R, Rees SE. A mathematical model approach quantifying patients’ response to changes in mechanical ventilation: evaluation in volume support. Med Eng Phys. 2015 Apr;37(4):341-9. doi: 10.1016/j.medengphy.2014.12.006. Epub 2015 Feb 14.
Larraza S, Dey N, Karbing DS, Nygaard M, Winding R, Rees SE. A mathematical model for simulating respiratory control during support ventilation modes. 19th World Congress of the International Federation of Automatic Control, IFA, 24-29 August 2014, Cape Town, South Africa. IFAC, 2014, s. 8433-8438
J. Duffin, “Role of acid-base balance in the chemoreflex control of breathing,” J Appl Physiol, vol. 99, pp. 2255-2265, 2005.
J. Duffin, “The role of the central chemoreceptors: A modeling perspective,” Respiratory Physiology and Neurobiology, vol. 173, pp. 230-243, 2010.
P. Ainslie and J. Duffin, “Integration of cerebrovascular CO2 reactivity and chemoreflex control of breathing: mechanisms of regulation, measurement an interpretation,” Am J Physiol Regul Integr Comp Physiol, vol. 296, pp. R1473-R1495, 2009.