Intensive Care Weaning (iCareWean) protocol on weaning from mechanical ventilation: a single-blinded multicentre randomised control trial comparing an open-loop decision support system and routine care, in the general intensive care unit

Introduction Automated systems for ventilator management to date have been either fully heuristic rule-based systems or based on a combination of simple physiological models and rules. These have been shown to reduce the duration of mechanical ventilation in simple to wean patients. At present, there are no published studies that evaluate the effect of systems that use detailed physiological descriptions of the individual patient. The BEACON Caresystem is a model-based decision support system that uses mathematical models of patients’ physiology in combination with models of clinical preferences to provide advice on appropriate ventilator settings. An individual physiological description may be particularly advantageous in selecting the appropriate therapy for a complex, heterogeneous, intensive care unit (ICU) patient population. Methods and analysis Intenive Care weaning (iCareWean) is a single-blinded, multicentre, prospective randomised control trial evaluating management of mechanical ventilation as directed by the BEACON Caresystem compared with that of current care, in the general intensive care setting. The trial will enrol 274 participants across multiple London National Health Service ICUs. The trial will use a primary outcome of duration of mechanical ventilation until successful extubation. Ethics and dissemination Safety oversight will be under the direction of an independent committee of the study sponsor. Study approval was obtained from the regional ethics committee of the Health Research Authority (HRA), (Research Ethic Committee (REC) reference: 17/LO/0887. Integrated Research Application System (IRAS) reference: 226610. Results will be disseminated through international critical care conference/symposium and publication in peer-reviewed journal. Trial registration number ClinicalTrials.gov under NCT03249623. This research is registered with the National Institute for Health Research under CPMS ID: 34831.

The purpose of this supplementary material is to provide the reader with an understanding of the structure and function of the Beacon Caresystem©, described here as the clinical decision support system (CDSS).
The material includes four sections describing: the mathematical models included in the system, including description of the models used to describe the effects of PEEP; the use of the system; a section illustrating the presentation of advice to the clinician; and a section illustrating screens relating to SBT and the extubation checklist. Much of the material presented here, including patient examples is taken from supplementary material from previous publications (1,2). Figure E1 illustrates the structure of the mathematical models with a full description of model formulation and evaluation published recently (1,2,3). The system includes mathematical models of pulmonary gas exchange; respiratory mechanics; acid-base chemistry of blood, interstitial fluid, tissues and cerebral spinal fluid; respiratory drive and ventilation. In addition, the models include the effects of PEEP on gas exchange, pulmonary mechanics and ventilation as illustrated in figure E2 and described below. All models are tuned to the individual patient's physiological status through measurement of respiratory gas flows and pressures; calorimetry and capnography measurements of respiratory gas fractions of O2 and CO2, and subsequent calculation of oxygen utilisation (VO2) and carbon dioxide production (VCO2); pulse oximetry measurement of arterial oxygen saturation; and arterial blood measurements of acid-base, oxygenation and haemoglobin fractions. The model of pulmonary gas exchange is tuned to the appropriate matching of ventilation and perfusion to account for O2 and CO2 differences between arterial and end tidal gas values.

Mathematical models included in the CDSS
To do so an arterial blood gas (ABG) is required on system start up. In some patients, the system also requires modification of FIO2 to two levels for 2-5 minutes at each level to tune the pulmonary gas exchange model to the patient, a procedure previously called the automatic lung parameter estimator (ALPE) technique (3,4). The respiratory mechanics model is tuned to dynamic compliance. The model of acid-base chemistry of the blood is tuned to measured values of arterial pH, PCO2, PO2, and SO2, and haemoglobin concentration, with the acid-base chemistry of the cerebrospinal fluid (CSF) regulated to arterial bicarbonate values to account for chemical changes in respiratory drive. The respiratory drive model is tuned to describe the relationship between alveolar ventilation (VA) and arterial acid-base and oxygenation status. The ventilation model is tuned to the measured serial dead space (Vds) to allow calculation of alveolar ventilation. A series of algorithms are present in the CDSS to re-tune the models as new measurements present, or if the patient condition changes. These models are used to simulate the effect of changes in ventilator settings, with the system generating advice based upon simulated values.   Figure E2 illustrates the mathematical models built to describe the effects of PEEP. These are formulated as empirical linear models, with the system only advising on PEEP changes of a maximum of 3 cmH2O in any step and the slope of these relationships modified automatically on measuring the response to PEEP. These models are integrated with those in figure E1 to enable simulation of the patient specific effects of PEEP and other ventilator settings simultaneously, enabling the system to advise on changing settings toward the best compromise for the individual patient given the model simulations.
For patients with acute respiratory failure, PEEP is usually set to improve gas exchange and lung mechanics, and the sub-figures A, B and C of figure E2 illustrate the baseline expected response to changes in PEEP of pulmonary shunt, respiratory system compliance and high ventilation/perfusion (V/Q). Changes in shunt, compliance and high V/Q automatically result in model-simulated changes in oxygenation, ventilation volumes and pressures, and carbon dioxide partial pressures through the models illustrated in figure E1. For shunt and compliance, the initial slopes of the models are dependent on the initial levels of shunt and compliance, with a greater improvement expected for a greater severity of respiratory abnormality. The slopes of these models are adapted automatically according to patient response to changes in PEEP, from measurements of oxygenation (SpO2), respiratory volumes and pressures, and end tidal CO2 values. Any increase in end tidal CO2 values that are simulated to result in severe acidosis result in the system requesting an arterial blood gas. The decision to increase PEEP is then based on the potential benefit of improved oxygenation and respiratory volumes and pressures, given the tuning of the models to the patient's state. For patients responding poorly to increases in PEEP, the absolute slope of shunt and compliance models would be substantially reduced following measurement of the response to PEEP changes. The decision to reduce PEEP at high SpO2 levels depends upon the FIO2 level. FIO2 will always be optimized first, such that the competing goals of sufficient oxygenation and oxygen toxicity are balanced.
Following this, the system will calculate the likely effects on oxygenation and respiratory volumes and pressures on reducing PEEP from the models illustrated on subfigures A, B and C of figure E2, combined with the models of E1. For patients recovering from respiratory injury, without substantial pulmonary shunt BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) BMJ Open doi: 10.1136/bmjopen-2020-042145 :e042145. 10 2020; BMJ Open , et al. Vizcaychipi MP or low respiratory system compliance, the system will calculate that the potential negative effects of reducing PEEP will be outweighed by the positive effects of reducing plateau pressure.
For patients on pressure support ventilation without substantial problems related to gas exchange or respiratory system mechanics, PEEP is often used to support the respiratory muscles, preventing abnormal respiratory muscle activity as indicated by abnormal breathing patterns. Figure E2, D and E represent the empirical linear models included in the system to account for changes in tidal volume (VT) due to PEEP. Figure E2D illustrates two potential abnormal situations; the first is a very high VT at a low PEEP value, and the second a very low VT at a high PEEP, possibly representing the situation where the diaphragm may be over distended due to high PEEP. In these situations, the initial slope of the PEEP versus VT relationship is negative. Figure E2E illustrates the situation of a low tidal volume at a low value of PEEP, possible due to under-support of the respiratory muscles. In this situation, the initial slope of the PEEP versus VT relationship is positive. As with other PEEP models, all slopes are adapted following measurement of the response to PEEP changes.  In addition to these examples, it is important to note that on occasion the system will present advice to switch ventilator mode. This occurs if the patient is in a controlled ventilator mode and the system identifies a total respiratory frequency or a total minute ventilation that is significantly greater than that set, where the system provides a message to the user suggesting changing to support mode. Similarly, if the patient is in support mode and the model simulated pH is very low then the system provides a message asking the user to consider either control mode or reduction of anesthesia. In modes containing mandatory breaths initiated by the ventilator or by patient effort, and where different settings are available for patient initiated and ventilator initiated breaths, the system provides advice on these depending upon the pattern of patient ventilation, and whether the patient is currently in a period with substantial spontaneous breathing activity. The system functions on a variety of manufactures ventilators and, as illustrated in these examples, provides advice for a variety of modes. Current exceptions to these include APRV, and modes with automatic control of ventilator settings such as ASV© or SmartCare©. Data is collected automatically by the Beacon Caresystem, and includes tidal volumes, both set if the mode dictates, and also measured.
The regulation of Vt and it relationship with mode can therefore be explored for both of the study arms.
The system records all ventilator settings via connection to the ventilator, all direct measurements of flow and gas concentrations, its own SpO2 measurement, and the timing and nature of all advice provided. The     Figure E4 illustrates the same screen as E3, but with the corners of the hexagon activated to show the current, simulated and advised values of variables simulated by the physiological models. In this patient, the current FIO2 setting of 41 %, results in a SaO2 value of 97.9 % and a current simulated SvO2 of 89.7 %.
The system illustrates that the current balance between over and under oxygenation may be inappropriate with the blue symbol on the hexagon pointing slightly toward oxygen toxicity. The system therefore suggests reducing FIO2. The system also illustrates an increased risk of respiratory muscle atrophy due to a low respiratory frequency, leading the system to suggest a reduction in PS from 14 to 12 cmH2O, simulating that this may result in increased Rf.    Figure E6 illustrates an advice including PEEP for a patient in pressure control ventilation. The vertical axis representing the balance between respiratory muscle atrophy and stress is disabled in controlled ventilation modes for cases where the patient has no spontaneous breathing activity. In this patient, the system indicated that the current balance between over and under ventilation might be inappropriate, with the blue symbol on the hexagon pointing toward lung trauma and with the patient being slightly alkalotic. The system suggested reducing PEEP and PC from 6 to 5 cmH2O and 12 to 11 cmH2O, respectively, and at the same time increasing Rf from 18 to 19 bpm.
The simulated effect of PEEP was a negligible increase in pulmonary shunt, with simulated patient response to the combined advice being to reduce inspiratory pressure, reduce pH and maintain appropriate oxygenation.   Figure E7 illustrates the result of following the advice illustrated in figure E8, and the subsequent advice provided by the system. The combined effect of PEEP and PC reduction and increase in Rf was as expected for both oxygenation and respiratory mechanics but less pronounced on pH. The advised level of PEEP was maintained in the next advice, but with an advice to decrease FIO2 expecting a resulting safe oxygenation. Simulations in the details boxes at the hexagon corners take into account differences between set and measured Rf, where measured Rf was 17 bpm despite set Rf of 19 bpm. Figure E9 -User interface of the system, with results according to the advice in Figure E6, and the next advice.  figure E10, the system provides a counter on the screen which counts down from 30 minutes, and as such indicates that the patient has been within ventilator settings consistent with an SBT and stable with regard other respiratory measurements, for this duration. Cut-off values for these other variables are illustrated in figure E10 Figure E10 -User interface illustrating variables used to time the SBT counter of the system. At any point when the counter is running, the user can view the screen illustrated in figure E11. This indicates which variables are within threshold, their values, and for how long this has been the case.