Article Text
PDF
Abstract
Objective To study the feasibility of automated titration of oxygen therapy in the delivery room for preterm infants.
Design Prospective non-randomised study of oxygenation in sequential preterm cohorts in which FiO2 was adjusted manually or by an automated control algorithm during the first 10 min of life.
Setting Delivery rooms of a tertiary level hospital.
Participants Preterm infants <32 weeks gestation (n=20 per group).
Intervention Automated oxygen control using a purpose-built device, with SpO2 readings input to a proportional-integral-derivative algorithm, and FiO2 alterations actuated by a motorised blender. The algorithm was developed via in silico simulation using abstracted oxygenation data from the manual control group. For both groups, the SpO2 target was the 25th–75th centile of the Dawson nomogram.
Main outcome measures Proportion of time in the SpO2 target range (25th–75th centile, or above if in room air) and other SpO2 ranges; FiO2 adjustment frequency; oxygen exposure.
Results Time in the SpO2 target range was similar between groups (manual control: median 60% (IQR 48%–72%); automated control: 70 (60–84)%; p=0.31), whereas time with SpO2 >75th centile when receiving oxygen differed (manual: 17 (7.6–26)%; automated: 10 (4.4–13)%; p=0.048). Algorithm-directed FiO2 adjustments were frequent during automated control, but no manual adjustments were required in any infant once valid SpO2 values were available. Oxygen exposure was greater during automated control, but final FiO2 was equivalent.
Conclusion Automated oxygen titration using a purpose-built algorithm is feasible for delivery room management of preterm infants, and warrants further evaluation.
- resuscitation
- neonatology
- intensive care units
- neonatal
Data availability statement
Data are available on reasonable request.
Statistics from Altmetric.com
What is already known on this topic?
Observational studies have revealed the difficulties in maintaining SpO2 within the prescribed shifting oxygen saturation target range during the first 10 min of life.
Automated control of oxygen delivery is well established for preterm infants in the neonatal intensive care unit and has been proven to increase time spent in the target SpO2 range compared with manual oxygen titration by clinicians.
Automated oxygen titration soon after birth has been trialled in a preterm animal model but needs evaluation in preterm infants under standard clinical conditions.
What this study adds?
A purpose-built oxygen control algorithm (VDL-DR1.0) was found to be equally effective to routine manual control in targeting the prescribed escalating SpO2 range.
Automated oxygen titration was associated with fewer SpO2 values above the target range when receiving supplemental oxygen.
No manual override adjustments were required during automated oxygen control, but the overall oxygen exposure was higher than with manual control.
Introduction
Administration of oxygen is a cornerstone of intensive care at any age, including at the time of birth.1 For oxygen therapy in the delivery room, current guidelines recommend titration of FiO2 from a starting point of 0.21 (term) or 0.21–0.30 (preterm),1 2 guided by oxygen saturation (SpO2) and informed by nomograms for SpO2 trajectory over the first 10 min.3 4 These nomograms reflect the expected gradual oxygenation change over several minutes with accretion of lung gas volume from the de-aerated state at birth.5 Supra-physiological oxygenation in the first minutes appears to offer no immediate advantage, and iatrogenic hyperoxaemia is known to have lasting detrimental effects for both the term6 and preterm7 infant. Protracted hypoxaemia is also undesirable.8
The ideals for SpO2 targeting in the delivery room outlined in the International Liaison Committee on Resuscitation guidelines and other consensus statements are not matched by current reality. Several studies in preterm infants <32 weeks gestation have highlighted the difficulties in titrating oxygen therapy to prescribed SpO2 targets in the delivery room (DR), finding that as a group, preterm infants spend the majority of time outside the prescribed boundaries for SpO2 in the first 10 min of life.9–12 These findings are unsurprising given that clinicians are focused on multiple tasks at the time of preterm birth, including thermoregulation, airway stabilisation and application of respiratory support. They also emphasise the challenges of manual oxygen titration in this context, and mirror the findings of studies beyond the delivery room, in which SpO2 has been noted to be in target range only 30%–60% of the time.13 14
As an aid to SpO2 targeting, devices which automate the titration of oxygen therapy have now emerged, and oxygen control algorithms are now embedded in several commercially available neonatal ventilators.15–19 These algorithms are known to increase time in the SpO2 target range in preterm infants, and reduce the frequency of extreme and prolonged SpO2 deviations.20–22 To our knowledge, no human studies have reported the use of automated oxygen control starting from birth, where the challenges of SpO2 targeting are considerably magnified. The combination of rapidly changing pulmonary blood flow, high levels of intrapulmonary shunt, inconsistent respiratory activity and shifting SpO2 targets make for a highly dynamic situation in which frequent and emphatic FiO2 alterations are often required. Given their provenance, it seems unlikely that existing oxygen control algorithms can perform well in this setting.
We conducted a study aiming to develop and test an automated oxygen control device designed specifically for DR care of preterm infants. We hypothesised that control of oxygen therapy using a novel algorithm for automated oxygen titration may increase time within the SpO2 target range in the first 10 min of life.
Methods
Study setting
The study was conducted in the delivery suites and operating theatres of the Royal Hobart Hospital (RHH), the tertiary neonatal care facility for the state of Tasmania, Australia. The Neonatal and Paediatric Intensive Care Unit at the RHH provides care for 70–80 preterm infants <32 weeks gestation per year.
Study design
This was a three-stage non-randomised study (stage I: observational study of manual oxygen control in the DR; stage II: in silico simulation study; stage III: preliminary interventional study of automated oxygen control in the DR), with oxygenation outcomes being compared between infants in stages I and III (online supplemental file).
Supplemental material
Participants
For stages I and III, preterm infants 23 to 31+6 weeks gestation were eligible for inclusion if inborn and not known prenatally to have major congenital or chromosomal anomalies. Only one of multiples could be included due to the limitation of study equipment. Non-sequential convenience sampling was used for enrolment of participants in both stages.
Study procedures
Stage I (observational study of manual oxygen titration)
Physiological data were recorded during routine delivery room management of preterm infants carried out by clinicians not involved in the study. Oxygen therapy was initiated with an FiO2 of 30% and adjustments to FiO2 were made via an air-oxygen blender (Bird Ultrablender, Vyaire, Yorba Linda, USA), aiming to place SpO2 within the bounds of the 25th and 75th centiles of the Dawson nomogram for babies <32 weeks gestation.4 SpO2, heart rate, FiO2 and airway pressure were recorded to a study computer using connections previously described14 23 (see online supplemental text and figure S1 for further details).
Supplemental material
Oxygenation and oxygen therapy in the patient groups overall. Graph of pooled data showing the spread of SpO2 and FiO2 values in the two study groups over the first 10 min. Upper panels: SpO2, lower panels: FiO2. Panels A and C: manual control; panels B and D: automated control. Light grey shaded area: boundaries of the 10th to 90th centiles for the study group populations; dark grey shaded area: IQR. Solid and dashed black lines in upper panels: median and 25th/75th centiles of the Dawson nomograms, respectively.
Stage II (algorithm development and implementation)
Development of the VDL-DR1.0 algorithm, a proportional-integral-derivative (PID) oxygen control algorithm tuned specifically for the delivery room, followed the methodology of our previous work24 and involved (i) abstraction of 1 Hz SpO2 and FiO2 data recorded in stage I (manual control) by conversion to a sequence of paired values for ventilation-perfusion (V/Q) ratio and shunt (Qs/Qt), (ii) use of these V/Q and shunt datastreams from each infant in a counterfactual simulation in which the performance of a PID oxygen control algorithm was examined in silico with multiple different permutations of coefficients for the proportional, integral and derivative terms, (iii) identification of the set of PID algorithm coefficients that optimised SpO2 targeting, measured against prespecified criteria and (iv) implementation of this algorithm in a feedback loop within a device, with a digital SpO2 input and a digital FiO2 output actuated by a motorised blender (see online supplemental text and figures S1-S3 for further details).
Stage III (automated oxygen titration)
Methodology for the interventional study examining the function of the VDL-DR1.0 algorithm was as for stage I, with the following exceptions (online supplemental figure S1): (i) beyond the recording capability described above, the study computer housed the VDL-DR1.0 algorithm, which received SpO2 input and calculated a set FiO2 value each second; (ii) the air-oxygen blender was modified by addition of a custom-built geared servomotor as previously described,18 allowing calculated FiO2 alterations to be actuated automatically. This blender also permitted manual FiO2 adjustments if required. Both simulation testing and previous clinical studies have demonstrated this device to be reliable, with delivered FiO2 corresponding to set values18; (iii)while clinical staff were instructed that they should remain in control of FiO2, the VDL-DR1.0 algorithm was activated at birth, and aimed to target the 50th centile of the Dawson nomogram for babies <32 weeks gestation4 during the first 10 min. In the absence of an incoming SpO2 signal, no change to existing FiO2 was made other than those made by the clinicians. Manual FiO2 adjustment at any time would lead to a halt in algorithm-directed FiO2 alterations for 30 s, with automated control thereafter resuming at the latest clinician-set FiO2 value.
Data collection and analysis
Physiological recordings as well as demographic and clinical information were collected in both study groups. The following oxygenation outcomes were determined: time in target range (SpO2 within the 25th and 75th centile boundaries of the Dawson nomogram when receiving supplemental oxygen, and within this range or above when in room air); time with SpO2 within 25th and 75th centile boundaries; time below 25th centile and above 75th centile when receiving oxygen; time of first SpO2 signal (seconds after birth) and time with no SpO2 signal. The number of FiO2 adjustments (change in measured FiO2 by 1% or greater) during manual and automated control was determined, as was average oxygen exposure (mean FiO2) and the final FiO2 at 10 min. Durations of various modes of respiratory support (continuous positive airway pressure (CPAP), positive pressure ventilation, intubation) were ascertained from the recording of circuit pressure. Clinical outcomes were documented during the hospital stay in both groups.
Outcome measures
The primary outcome was time in SpO2 target range (TR), measured for each infant over the first 10 min. Secondary outcomes included time above and below TR, number of FiO2 adjustments, mean oxygen exposure and final FiO2.
Statistical analysis
Continuous data were expressed as median and IQR. Intergroup comparisons were made with χ2 test or Fisher’s exact test for categorical data, and Mann-Whitney U test for continuous outcomes. Statistical significance was assumed where p value was <0.05. A sample size of 20 in each group was chosen pragmatically in view of this being an initial study of delivery room oxygen control. The number of 20 per group allowed a 1 SD difference in TR time between groups to be detected with >80% power and alpha error 0.05.25
Results
Twenty infants were included in each of the manual oxygen titration (manual control) group (September 2015–August 2016) and the automated control group (June 2017–August 2018). One infant in the automated control group had missing oximetry data and hence was excluded from the analysis. Airway pressure recordings were missing in one infant in each group. Demographic and clinical characteristics were comparable between the groups (table 1). Most infants were born by caesarean section and fully exposed to antenatal steroids.
Demographic and clinical characteristics
The time to first receive SpO2 readings was around 90 s on average and was somewhat longer for the automated oxygen titration group. For one infant in the automated control group, there was a delay of 436 s in obtaining an SpO2 reading, during which time four manual FiO2 adjustments were made. For other infants in both groups, no FiO2 adjustments were deemed necessary by clinicians in advance of a valid SpO2 reading being available. The initial SpO2 value, total time with a valid SpO2 signal and time until heart rate exceeded 100 beats per minute were similar between groups (table 2). Parameters of respiratory support were similar other than for slightly higher applied positive end-expiratory pressure in the automated control group (table 2). One infant (in the automated control group) was intubated in the delivery room; all others were successfully transitioned to spontaneous breathing supported by CPAP.
Delivery room physiological indices and respiratory support
Representative delivery room recordings (online supplemental figures S4 and S5) demonstrate the variability of oxygenation and the challenges of SpO2 targeting in the newly born preterm infant, both during manual and automated oxygen control. In aggregated data, oxygenation loosely followed the trajectory of the Dawson nomograms (figure 1), with widely ranging SpO2 values in the first 5 min in both groups, narrowing to more precise SpO2 targeting beyond 400 s, particularly with automated control. Values for FiO2 in the aggregated data were higher in the automated control group through most of the 10 min recording, with a broad range of values particularly during the first 5 min (figure 1).
Oxygenation outcomes were equivalent or better with automated oxygen titration compared with the manual control group (table 3). Median time in the target range was approximately equivalent in the two groups, whereas time with SpO2 in the hyperoxaemic range (>75th centile when receiving oxygen) was less with automated control. In infants where a valid SpO2 signal was available by 5 min, the proportion with SpO2 ≥80% at this time was 15/20 (75%) and 14/18 (78%) in the manual and automated control groups, respectively. As expected, the FiO2 adjustments were made considerably more frequently by the oxygen control algorithm than by clinicians during manual control (table 3). No manual adjustments to FiO2 were made in any infant in the automated control group once a valid SpO2 signal was obtained. Overall oxygen exposure trended higher with automated control, but the final FiO2 at 10 min was similar (table 3). Clinical outcomes did not differ between groups (online supplemental table S1).
Delivery room outcome data
Discussion
Achieving satisfactory oxygenation is a key component of delivery room care for the preterm infant, but the current approach of manual FiO2 adjustment lacks finesse and precision. In a prospective sequential cohort study, we found that by automating oxygen titration with a PID control algorithm tuned for this setting, SpO2 values were in the desired target range in the first 10 min with equal frequency to manual oxygen control, with fewer SpO2 readings >75th centile when receiving oxygen. The delivered FiO2 was higher overall in the automated control group, but final FiO2 values were similar between groups.
Previous studies of oxygenation in the delivery room have highlighted the difficulty of SpO2 targeting in the first period after birth of a preterm infant.9–12 These studies each had different SpO2 targets, with either a stipulated range of acceptable SpO2 values,9 10 12 or a single value,11 in most cases shifting upwards during the first 10 min of life. Average proportion of time within a prescribed SpO2 target range was 11%–21% in the three groups in the trial of Rabi et al,9 37% and 52% in two groups studied by Gandhi et al 10 and 35% in a single group study.12 In comparison to these previous studies, infants in our manual oxygen titration group spent a relatively high proportion of time (median 60%) in the SpO2 target range. The relative maturity of these infants and the low number requiring intubation in the delivery room may be contributing factors to the higher proportion of target range time, as may be the choice of a starting FiO2 of 0.30 rather than 0.21 or 1.0.
Acknowledging the relative success of SpO2 targeting in our manual control comparator group, the application of automated oxygen control in a similar group of infants using the VDL-DR1.0 algorithm was associated with an equivalent proportion of time in the SpO2 target range, and fewer SpO2 readings in the hyperoxaemic range. This was achieved without a single episode of manual oxygen titration in any infant once valid SpO2 readings were obtained, suggesting that clinicians were comfortable with the algorithm-driven FiO2 adjustments that were unfolding during the course of the delivery room care.
Exposure to oxygen was higher with automated control in our study, although the final FiO2 was similar. This can be explained in part by the frugality of clinicians in relation to oxygen therapy in the manual control group. The design of the algorithm also permitted FiO2 to rise to a peak of 1.0 over a relatively short period (~60 s) if oxygenation remained poor. This approximated or exceeded the prescribed rate of rise in FiO2 in previous studies of manual titration of oxygen in the delivery room (1–4 min9 10 12), and likely contributed to the wider range of FiO2 values found with automated oxygen control compared with manual control in the present study. Importantly, however, the resultant SpO2 values in the two groups were similar in their distribution.
The process of tuning of an algorithm for delivery room oxygen titration followed that we have described previously for automation of oxygen control in the neonatal intensive care unit (NICU).24 Requisite steps included the gathering of oxygenation data under standard conditions of manual oxygen titration, and the manipulation of these data in an abstracted form allowing multiple permutations of coefficients for the PID algorithm to be tested in a counterfactual in silico simulation. This approach meant that the behaviour of the selected algorithm could be anticipated with some confidence, but made the assumption that the manual control data were broadly representative.
A single prior controlled trial involving preterm lambs examined the effectiveness of automated oxygen control in the delivery room, using a rule-based algorithm designed for oxygen titration during mechanical ventilation in the NICU.17 Time in the SpO2 target range (25th–75th centile of the Dawson nomogram4) was similar in the automated and manual control groups (median 44% and 41%, respectively), but there was a significant reduction in hyperoxaemia with automated oxygen titration.
Our study has the limitation of being preliminary in nature, involving a small number of preterm infants managed in a single centre. Only with experience of automated oxygen titration in a larger number of infants, with a larger subgroup requiring intubation, will the effects on delivery room oxygenation be gauged in full. Such investigations could be conducted as stand-alone clinical trials or embedded into delivery room oxygen therapy trials as substudies, and should be powered to examine effects on neonatal outcomes known to be influenced by oxygenation, including mortality, retinopathy of prematurity, bronchopulmonary dysplasia and necrotising enterocolitis.26
Conclusion
Automated oxygen titration using a purpose-built algorithm is feasible for delivery room management of preterm infants, and warrants further evaluation.
Data availability statement
Data are available on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
The study protocol was approved by the University of Tasmania Human Research Ethics Committee (approval number H15112) and written informed consent was obtained prenatally from parents of each infant.
Acknowledgments
The authors would like to thank the parents of participating infants and the Neonatal and Paediatric Intensive Care Unit staff of the Royal Hobart Hospital for their cooperation in conducting this study. Circuit pressure monitoring equipment was kindly loaned by Fisher & Paykel Healthcare Ltd, East Tamaki, New Zealand.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors SKMA: conceived the study and gained funding (with PAD and TJG), conducted the study, compiled and analysed the data, wrote the first draft of the manuscript and approved the final version. RJ: was involved in the tuning, simulation testing and implementation of the VDL-DR1.0 algorithm, reviewed and edited the manuscript and approved the final version. APM: was involved in the tuning, simulation testing and implementation of the VDL-DR1.0 algorithm, reviewed and edited the manuscript and approved the final version. TJG: conceived the study and gained funding (with SKMA and PAD), was involved in the tuning, simulation testing and implementation of the VDL-DR1.0 algorithm, reviewed and edited the manuscript and approved the final version. PAD: conceived the study and gained funding (with SKMA and TJG), was involved in the tuning, simulation testing and implementation of the VDL-DR1.0 algorithm, conducted the study, analysed the data, reviewed and edited the manuscript, approved the final version, and takes responsibility for the content as the study guarantor.
Funding This study was supported by a grant (15-101) from the Royal Hobart Hospital Research Foundation.
Competing interests The University of Tasmania holds a patent concerning a method, apparatus and system for automatically controlling inspired oxygen delivery. An oxygen control algorithm (VDL1.1) has been incorporated as the OxyGenie automated oxygen titration system in the SLE6000 infant ventilator (SLE, South Croydon, UK).
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.