This study was performed at the pediatric intensive care unit (PICU) of the Beatrix Children’s Hospital, University Medical Center Groningen between February and July 2015. The Institutional Review Board approved the study. Signed informed consent was obtained from both parents or legal caretakers. Mechanically ventilated children < 18 years of age were eligible for inclusion. Patients with congenital or acquired neuromuscular disorders, premature birth with gestational age corrected for post-conceptional age < 40 weeks, severe traumatic brain injury (i.e. Glasgow Coma Scale < 8), congenital or acquired damage to the phrenic nerve, congenital or acquired paralysis of the diaphragm, use of neuromuscular blockade, chronic lung disease (i.e. tracheostomy ventilation), severe pulmonary hypertension, contra-indication for placement of electrodes on the skin and patients unable to trigger the ventilator from any other cause were excluded.
During the study, all patients remained subjected to standard-of-care of the intensive care. Measurements took place within the 24 h prior to extubation. The attending physician defined the ventilator mode and settings in agreement with our local guideline. Expiratory tidal volume (VT) was targeted at 6–8 mL/kg actual bodyweight. The flow trigger was set at 1.0 L/min. A proximal flow-sensor was used in patients < 15 kg. In cases of decreased respiratory system compliance, permissive hypercapnia was applied (pH > 7.20). The level of pressure support ventilation (PSV) was routinely set as PSV = peak inspiratory pressure (PIP) minus positive end-expiratory pressure (PEEP). Ventilator settings were fixed during the measurement unless the clinical condition of a patient required an adjustment of the setting made by the attending physician. Patients were ventilated in a time-cycled, pressure-limited synchronized mode of ventilation with PSV, pressure controlled/synchronized intermittent mandatory ventilation (PC/SIMV + PSV), pressure-limited mode with preset tidal volume (VT), i.e. pressure regulated volume controlled with PSV (PRVC/SIMV + PSV) or pressure controlled assist control (PC/AC) using the AVEA ventilator (CareFusion, Yorba Linda, CA, USA). Continuous infusion of midazolam, oral lorazepam and morphine or fentanyl intravenously was given for analgesia–sedation. The COMFORT behavior scale was used to titrate the level of sedation [20, 21]. Ten minutes prior to the recordings, patients were suctioned and the circuit was cleared from any water. Patients were in a 30 degrees anti-Trendelenburg supine position.
Ventilator pressure waveforms and dEMG acquisition
Patients underwent a 5-min continuous recording of ventilator pressure waveforms and dEMG. Ventilator pressure tracings were acquired through the ventilator’s RS232 interface (Ventilator Open XML Protocol, VOXP) at a sampling frequency of 100 Hz. The dEMG was derived from one pair of single Ag/AGCl electrodes (EasyTrode TM Pre gelled Electrodes, Multi Bio Sensors Inc, El Paso, USA) bilaterally placed at the costo-abdominal margin in the nipple line. A common electrode was placed at the sternal level . The dEMG was recorded at a sampling frequency of 500 Hz using the Dipha (Inbiolab, Groningen, The Netherlands). Polybench software (Applied Biosignals GmbH, Weener, Germany) was used to record the pre-processed data from the ventilator and the EMG recording device. The ventilator pressure waveforms and electrical activity of the diaphragm were analyzed offline.
The recorded dEMG needed to be processed for reliable assessment of the respiratory neural drive. The electrical activity of the heart and other peak artifacts were isolated from the raw dEMG data by means of an extended version of the gating technique . The gates were filled with the running average of the processed dEMG signal. A 50 Hz notch filter was used to minimize electrical interference from electronic devices on the intensive care. After filtering and gating, the running root mean square (RMS) (time window T = 0.2 s) of the processed dEMG signal was calculated. The calculated dEMG was used for analysis.
Description of patient–ventilator interaction
To evaluate patient–ventilator interaction, the computed dEMG activity was both manually and automatically compared to the ventilator’s waveforms to calculate the dEMG-phase scale (dEMG-phase scaleMANU and dEMG-phase scaleAUTO, respectively), using the modified NeuroSync method previously described by Sinderby et al. . Two investigators (AK and RB) manually analyzed the ventilator pressure and dEMG tracings using a graphical interface designed in Polybench (Applied Biosignals GmbH, Weener, Germany). Each investigator individually placed markers in the interface at the onset of neural inspiration (NAON), at 1/3 decline in the dEMG from its peak, i.e. the termination of neural inspiration (NAOFF), at the beginning of ventilator pressurization (MVON) and at the end of ventilator pressurization (MVOFF). The obtained neural inspiration and ventilator pressurization timings: NAON, NAOFF, MVON and MVOFF were used to calculate trigger and cycle-off errors of each breath. The algorithm for automated analysis was designed according to the same rules as for manual analysis. Early trigger and cycle-off errors as well as late trigger and cycle-off errors could range between 0 and 100%. Limits whether a breath is synchronous, dyssynchronous were set, accordingly to Sinderby et al. , at ± 33% difference between NAON and MVON and NAOFF and MVOFF. Neural inspirations not related to ventilator pressurizations or vice versa were considered as asynchronous breaths and assigned 100%. Cases of asynchronous breaths included ventilatory pressurization without neural activity (MV without NA), neural activity without ventilatory pressurization (NA without MV), multiple ventilatory pressurizations with one neural activity (multiple MV with NA) and multiple neural activities within one ventilatory pressurization (multiple NA with MV). Obtained data are shown in a graphical representation of the dEMG-phase scale; the intra-breath patient–ventilator interaction diagram. The dEMG-phase scale was defined as the mean absolute error of all breaths. The dEMG-phase scaleMANU, which was obtained by both experts, was compared with the dEMG-phase scaleAUTO.
Patient baseline demographics included age, gender, weight, admission diagnosis, Pediatric Index of Mortality (PIM) II and 24-h Pediatric RISk of Mortality (PRISM) II score, time of recordings and admission diagnosis. Before initiation of the measurements, ventilator settings including mode, pressure above peep (PAP), PEEP, mean airway pressure (Pmean), PSV, expiratory tidal volume (VT), frequency of set breaths, fraction of inspired oxygen (FiO2) and inspiratory time were recorded. Clinical data included prior use of neuromuscular blockade, tube size, air leakage around the endotracheal tube (ETT), end tidal CO2 and received amount of analgesia-sedation in the last 4 h preceding the registration. The COMFORT score was evaluated during the recording [20, 21].
The Shapiro–Wilk test was used to test data for normal distribution. Descriptive data were expressed as median [first quartile; third quartile] or percentage (%) of total. The breath-by-breath inter-rater agreement, defined as the agreement between errors obtained by the two investigators, and inter-method agreement, defined as the agreement between errors obtained by automated analysis and the average errors obtained by the two investigators, were evaluated by means of the intra-class correlation coefficient (ICC). Reliability was considered to be acceptable if the ICC was greater than 0.75 and excellent if the ICC was greater than 0.90 [23, 24]. After confirmation of a good breath-by-breath inter-rater and inter-method reproducibility, the agreement between dEMG-phase scaleMANU and dEMG-phase scaleAUTO was evaluated. All statistical analyses were performed using SPSS version 24 (IBM, Armonk, USA).