Acute respiratory failure is a common, life-threatening condition where the lungs cannot provide enough oxygen to the body. Many patients are treated with non-invasive respiratory support (NRS) such as high-flow nasal oxygen (HFNO), continuous positive airway pressure (CPAP), or bilevel positive airway pressure (BiPAP). However, up to half of patients receiving NRS still deteriorate and require intubation and invasive ventilation, which is linked to longer hospital stays, more complications, and slower recovery.

A major challenge in caring for these patients is that clinicians currently cannot directly see how well the breathing muscles (especially the diaphragm and parasternal intercostal muscles) and the lungs are working while the patient is using NRS. Existing bedside measures, such as respiratory rate or oxygen levels, only show part of the picture. They do not indicate how hard the patient is working to breathe or whether their respiratory muscles are becoming fatigued. This lack of information may delay important decisions about adjusting NRS settings or switching to other treatments.

This study aims to find out whether two advanced but non-invasive, radiation-free bedside monitoring tools can be used effectively in routine care:

1. Ultrasound, which can measure breathing muscle thickness, movement, and lung aeration
2. Electrical impedance tomography (EIT), which uses a soft belt of small electrodes around the chest to measure changes in air and blood flow within different regions of the lungs in real time

These tools have shown promise in earlier research, and interviews with patients and clinicians suggest they are comfortable, well-tolerated, and potentially useful. However, they have not yet been evaluated together in a real-world hospital environment where many acute respiratory failure patients are cared for outside the ICU.

What the study will involve:

Up to 100 adults with acute respiratory failure requiring any type of non invasive respiratory support will be recruited with the goal of obtaining complete data from at least 50 patients. Each participant will undergo ultrasound and EIT assessments up to seven times during the first 72 hours after starting NRS, plus an additional measurement if they improve enough to stop NRS or if they deteriorate and require intubation. These assessments take place at the bedside, require brief exposure of the upper chest, and last approximately 15-45 minutes. Routine clinical data-such as heart rate, oxygen levels, and breathing measures-will also be recorded.

In parallel, clinical staff caring for these patients will complete a short Healthcare System Usability Scale questionnaire to rate how useful, understandable, and practical they find the information generated by ultrasound and EIT. Some staff may also take part in optional interviews to explore usability in more depth.

What the study is trying to learn:

The primary aim is to determine the usability of these monitoring methods meaning understanding if they are practical, easy to use, and helpful for clinicians making decisions about NRS treatment.

Secondary aims include understanding:

• how the respiratory muscles and lungs change over time during NRS
• whether these changes are linked to treatment settings (e.g., flow rate, pressure support)
• whether certain patterns are associated with treatment success or failure (intubation or death)
• whether these tools could help identify patients at risk of deterioration earlier

Risks and benefits:

Both ultrasound and EIT are widely used, safe, and non-invasive. They involve no radiation, needles, or harmful exposure. Minor temporary discomfort from the gel or belt placement is possible. Participation will not change any clinical treatments. Although patients may not directly benefit, the study may help future patients by improving understanding of breathing muscle function and supporting more personalised respiratory care.

By contributing to this research, patients and clinicians will help determine whether advanced monitoring can be realistically implemented in busy hospital settings and whether it could lay the groundwork for future trials aimed at improving outcomes for people with acute respiratory failure.

Background and Rationale Acute respiratory failure (ARF) is a common and life-threatening syndrome characterised by inadequate gas exchange, resulting in hypoxaemia with or without hypercapnia, and frequently necessitating hospital admission and escalation of respiratory support. ARF is associated with substantial short-term mortality and long-term morbidity, including prolonged hospitalisation, reduced functional capacity, impaired quality of life, and increased healthcare utilisation. Despite advances in supportive respiratory therapies, outcomes remain poor for a significant proportion of patients, particularly when clinical deterioration is not recognised early.

Non-invasive respiratory support (NRS), including high-flow nasal oxygen (HFNO), continuous positive airway pressure (CPAP), and bilevel positive airway pressure (BiPAP), has become first-line therapy for many forms of ARF. These modalities aim to improve oxygenation, reduce work of breathing, and prevent the need for endotracheal intubation and invasive mechanical ventilation. Avoiding invasive ventilation is associated with reduced risk of ventilator-associated pneumonia, ventilator-induced lung injury, diaphragm disuse atrophy, delirium, and long-term neuromuscular weakness. Consequently, NRS is increasingly delivered not only in intensive care units (ICUs) but also in emergency departments, high-dependency units, and general wards.

However, despite widespread use, NRS failure rates remain substantial. A significant proportion of patients deteriorate and require delayed intubation, which is consistently associated with worse outcomes compared with early escalation. One of the major challenges in managing patients receiving NRS is the limited ability to directly assess respiratory muscle workload and lung mechanics at the bedside. As a result, clinicians often rely on indirect clinical markers that may lag behind physiological deterioration.

The primary pathophysiological determinant of ARF progression and NRS failure is the imbalance between ventilatory load and respiratory muscle capacity. Excessive inspiratory effort can lead to respiratory muscle fatigue, impaired ventilatory efficiency, and patient self-inflicted lung injury due to high transpulmonary pressures during spontaneous breathing. Importantly, these processes may occur even when conventional oxygenation metrics appear stable.

Traditional bedside metrics, such as respiratory rate, peripheral oxygen saturation, arterial blood gas measurements, and composite indices including the ROX index or HACOR score, provide indirect and incomplete insight into respiratory effort. While these measures are useful for population-level risk stratification, they cannot reliably quantify work of breathing or identify early respiratory muscle overload at the individual patient level. Furthermore, these indices are influenced by multiple confounders, including sedation, oxygen delivery settings, and clinician intervention.

Oesophageal manometry remains the reference standard for assessing inspiratory effort and work of breathing. However, its invasive nature, poor patient tolerance, technical complexity, and limited availability render it impractical for routine use in awake, non-intubated patients receiving NRS, particularly outside the ICU environment. Consequently, there is a critical unmet need for practical, non-invasive tools that provide real-time physiological insight into respiratory muscle function and lung mechanics during NRS.

Two non-invasive bedside technologies-ultrasound (US) and electrical impedance tomography (EIT)-offer complementary and physiologically meaningful assessments of respiratory mechanics and lung function. Respiratory muscle ultrasound enables direct visualisation and quantification of diaphragmatic and parasternal intercostal muscle structure and activity, providing surrogate markers of inspiratory effort, muscle recruitment, and mechanical efficiency. Lung ultrasound enables serial assessment of lung aeration and consolidation, capturing dynamic changes that may not be apparent on conventional imaging.

Electrical impedance tomography provides continuous, breath-by-breath assessment of regional lung ventilation and changes in end-expiratory lung volume, offering insight into ventilation distribution, lung homogeneity, and dynamic lung mechanics during spontaneous breathing supported by NRS. Together, US and EIT have the potential to bridge the gap between physiological understanding and bedside decision-making.

Although both modalities are increasingly used in research and selected clinical settings, neither has been systematically evaluated for usability, feasibility, and clinical applicability in patients receiving NRS across diverse hospital environments. In particular, it remains unclear how clinicians interpret, trust, and integrate this information into real-world decision-making processes. Understanding these aspects is essential before advanced monitoring can be embedded into routine care or tested in interventional trials.

Study Objectives Primary Objective The primary objective of this study is to evaluate the usability of respiratory muscle ultrasound and electrical impedance tomography as clinical decision-support tools for patients with acute respiratory failure receiving non-invasive respiratory support. Usability will be assessed using the Healthcare System Usability Scale (HSUS), focusing on effectiveness, efficiency, and clinician satisfaction when interpreting and applying physiological monitoring data in routine care.

Secondary Objectives

Secondary objectives are to:

• Assess the feasibility of performing repeated, protocolised ultrasound and EIT measurements across multiple time points during the early phase of NRS, including recruitment, retention, tolerability, data completeness, and technical reliability.
• Quantify temporal changes in respiratory muscle function, including diaphragmatic and parasternal intercostal muscle activity, and lung aeration and ventilation patterns over the first 72 hours of NRS.
• Examine the relationship between physiological measurements derived from US and EIT and NRS treatment settings, including flow rate, positive end-expiratory pressure (PEEP), and pressure support.
• Explore associations between respiratory muscle and lung physiological patterns and clinically relevant outcomes, including escalation to invasive ventilation and in-hospital mortality.
• Collect structured qualitative field notes describing workflow integration, interpretability, and real-world usability of advanced monitoring techniques from the perspective of the research team and clinical staff.

Study Design This is a prospective interventional study to be conducted across two hospital sites: the Royal London Hospital and the Newham University Hospital across Barts Health over 14 months. Data collection will be undertaken by the co-investigator, who is a member of the direct care team.

The study aims to obtain complete longitudinal physiological datasets from at least 50 adult patients. Up to 100 participants will be recruited to account for attrition due to early clinical deterioration, intolerance of monitoring, missing data, or withdrawal. In parallel, approximately 50 clinical staff members involved in the care of participating patients will complete usability assessments, and up to 20 may participate in optional semi-structured interviews.

For patients with acute respiratory failure requiring non-invasive respiratory supports serial measurements of respiratory muscle function will be taken across six time points within the first 72 hours (from day 1 to day 3) of commencing non-invasive respiratory support. Day 1 is defined as the first 24 hours from starting any non-invasive respiratory device.

The measurements taken from day 1 to day 3 are described below:

Ultrasound (US) data:

• Diaphragmatic excursion
• Parasternal intercostal muscle cross-sectional area and thickness at end inspiration and end expiration
• Diaphragmatic and parasternal thickening fraction
• Parasternal intercostal muscle strain from the US video
• Lung parenchyma aeration, consolidation and fluid burden following the recommended approach from current evidence of the Blue Protocol and the Lung Ultrasound score (as per literature).

Electrical Impedance Tomography (EIT) data:

The EIT lung imaging field will be divided into two regions of interest: from halfway down, the dependent dorsal lung region will be identified, and the other half represented the non-dependent ventral region. The following EIT parameters will be measured:

• Global and regional changes in end-expiratory lung impedance (corresponding to changes in end-expiratory lung volume) expressed in arbitrary units of impedance change from the baseline step (∆EELI, ∆EELInon-dep, and ∆EELIdep, respectively)
• Lung compliance and inhomogeneity These measurements will also be collected at a variable time point defined as when the patient is liberated from non-invasive respiratory or when is intubated.

For completeness, from day 1 to day 3 and at a variable time point, basic routinely measured data will also be collected such as respiratory rate, heart rate, peripheral oxygen saturation, partial arterial oxygen pressure, partial arterial carbon oxide pressure, fraction of inspired oxygen, ROX index (Respiratory rate Oxygenation) defined as the ratio of oxygen saturation (SpO2)/fraction of inspired oxygen (FiO2) over respiratory rate (RR), pain score (numerical scale), conscious level. Breathlessness score (using the Borg scale) also be collected from day 1 to day 3 and at a variable time point if the patient is not intubated.

Data about the in NRS treatment settings (i.e. flow, PEEP and pressure support) will be collected; as well as outcome data regarding treatment failure such as intubation rate and death.

The initial assessment will take place at the earliest possible point in their admission (e.g., once the patient has been deemed eligible and consent has been received). Evaluation of respiratory muscle function (ultrasound and EIT), will be completed across six timepoints from day 1 to day 3. Please see Table 1 below.

Usability will be evaluated across two times points at day 1 and at a variable time point either at day 2 or day 3 as clinical workload allows. To evaluate the usability of data acquired with US and EIT (in monitoring respiratory muscle function) to guide clinical decision making, the co-investigator (BF) will undertake the following steps:

1. Present the data acquired with US and EIT, alongside basic routinely measured data and information about the NRS settings to clinical staff

2. Administer the Healthcare System Usability Score (HSUS) questionnaire will be administered to two clinical staff (i.e. a senior doctor in training or consultant and a nurse or allied health care practitioner) involved in making decisions about patients treatments. This will allow to evaluate the usability of the data in supporting clinical decision making.

3. In addition, we will collect field notes defined as written records of observations, experiences, and insights while conducting this research to evaluate usability in depth.

   Only if additional manpower resources allow, semi-structured interview will be undertaken with up to 20 multidisciplinary clinical staff.

   All data will be managed using secure and anonymised databases. Data will be reported using descriptive and inferential statistics.

   The study is purely observational. The research team does not provide treatment recommendations or mandate changes to clinical management. Clinicians may view monitoring data as part of routine care but retain full autonomy over treatment decisions.

   Eligibility criteria:

   Inclusion criteria

   • Adult (≥18 years old)
   • with acute respiratory failure with hypoxia (i.e. arterial oxygen tension (PaO2) of <8.0 kPa), and/or with or without hypercapnia (i.e. arterial carbon dioxide tension (PaCO2) of >6.0 kPa) from any underlying disease or cause
   • requiring any non-invasive respiratory support (i.e. HFNO, CPAP, BiPAP)
   • Multidisciplinary critical care staff involved in the management of those recruited patients with acute respiratory failure requiring non-invasive respiratory supports. Staff will possibly have an interview and are also required to complete a questionnaire.

   Exclusion criteria

   • Patients in respiratory arrest defined as the total cessation of airflow and breathing effort and absent ventilation
   • Patients requiring immediate intubation
   • Patients with Glasgow Coma Scale (GCS) < 8
   • Patients with severe facial trauma or burns
   • Patients with fixed upper airway obstruction or inability to protect the airway
   • Patients with severe agitation and/or confusion that prevent use of the device mask
   • Patients with severe vomiting
   • Pregnancy
   • Patients with pacemakers and other electronic devices in the thorax
   • Patients on end-of-life care or palliative care (defined as expected to die and/or not receiving active treatment)
   • Contra-indication to EIT or ultrasound monitoring (e.g. burns, severe obesity, thoracic wounds limiting instrument placement, and thoracic drain)

   Study Procedures Ultrasound Assessments Respiratory muscle and lung ultrasound assessments are performed at the bedside using portable GE Venue Go ultrasound systems equipped with linear and phased array probes. All measurements follow standardised acquisition protocols to minimise operator variability.

   Measurements include:

   • Parasternal intercostal muscle cross-sectional area, thickness, and thickening fraction Parasternal intercostal muscle assessments include measurement of muscle thickness, cross-sectional area, and thickening fraction at end-expiration and end-inspiration. Video loops are acquired to enable offline strain analysis using speckle-tracking techniques, providing additional insight into muscle contractile behaviour.
   • Diaphragmatic thickness, thickening fraction, and excursion using B-mode and M-mode imaging Diaphragmatic ultrasound includes assessment of thickness, thickening fraction, and excursion using B-mode and M-mode imaging. Probe position and measurement timing are standardised, and multiple measurements are averaged to improve reliability.
   • Lung parenchyma aeration following the Lung Ultrasound Score (six-zone method) Lung ultrasound is performed using a six-zone scanning protocol to quantify lung aeration and consolidation using validated scoring methods. Static images and cine loops are archived for offline review and quality assurance.

   Videos will be stored for later strain analysis of parasternal muscle contraction.

   Electrical Impedance Tomography Electrical impedance tomography is performed using the INFIVISION ET1000 system. A 16-electrode belt is positioned circumferentially around the thorax at the 5th-6th intercostal space. After signal stabilisation, continuous impedance data are acquired.

   EIT-derived parameters include global and regional changes in end-expiratory lung impedance, indices of ventilation distribution and homogeneity, and estimates of lung compliance. Lung regions are segmented into dependent and non-dependent zones to assess gravitational effects on ventilation during NRS.

   Routine Clinical Data At each monitoring time point, routinely collected physiological and clinical data are recorded, including respiratory rate, heart rate, oxygen saturation, inspired oxygen fraction, arterial blood gas values when available, and NRS device settings. Conscious level and pain scores are documented. Subjective dyspnoea is assessed using the Borg scale when patients have capacity and are able to participate.

   Usability Assessments Usability is assessed using the Healthcare System Usability Scale (HSUS), a validated instrument aligned with international usability standards. The HSUS evaluates clinicians' perceptions of the usefulness, interpretability, and workflow integration of US and EIT data.

   Clinical staff complete the HSUS at two time points: early during NRS and at a later variable time point. In addition, the co-investigator records structured field notes during data acquisition to capture contextual factors, workflow challenges, and informal clinician feedback. Optional semi-structured interviews further explore clinician experiences, cognitive load, and decision-making processes.

   Outcome Measures

   • The primary outcome is the HSUS score reflecting usability of advanced respiratory monitoring data.
   • Secondary outcomes include feasibility metrics, temporal changes in physiological parameters, associations with NRS settings, clinical outcomes such as intubation and mortality, and qualitative usability insights derived from field notes and interviews.

   Assessment and management of risk

   All the data collected, and the monitoring instruments used as intervention are non-invasive and radiation free causing no complications or side effects for either participants or investigators. However, in patients who are confused or lack capacity asking them to score dyspnoea providing a subjective measure (i.e. Borg scale) comes with risks such as unreliable self-reported score or inability to provide the score due to limited comprehension or understanding. This can threaten the validity of the score and lead to misclassification of dyspnoea severity. Therefore, to mitigate this risk and avoid inappropriate treatment decisions, subjective scoring like the Borg scale will not be collected for confused/delirious patients. Additionally, performing additional procedures like US and EIT for patients who are confused or lack capacity may cause additional distress and agitation leading to artifacts and unusable data. To manage these risks, we will undertake the following steps:

   • Explain the procedures simply, even if comprehension is limited, and try to reassure patients as much as possible to minimize distress
   • Optimise the environment reducing noise and involve family member if possible and if this can offer further reassurance to the patient
   • Optimise the timing to collect the data, meaning performing US and EIT measurements when the patient is relatively calm and after basic needs (i.e. analgesia, repositioning) are addressed
   • Ensure the US probe and EIT belt are well tolerated using adequate gel and quick short sessions to reduce patient distress
   • Ensure a second clinician (i.e. nurse, doctor or physiotherapist) is also present by the bedside during the procedure to offer additional reassurance to the patient while the researcher is performing the measurement with US and EIT If despite taking all the steps above, patients with or without capacity are in any visible distress (i.e. verbally refuse to continue to participate in the procedure) then these procedures with US and EIT will be stopped to avoid causing further ditress to patients. This deviation from the protocol will be adequately documented in the patient's notes. Our patient representatives have advised on this process and they agree that this is a reasonable approach to ensure no further distress is caused to any patient.

   In terms of data handling and reporting, we will record and report when subjective scoring and data measurement could not be obtained and the reasons why. For transparency we will also report the proportion of missing and incomplete data.

   Statistical considerations Our primary aim is to evaluate the usability of the measurement data acquired with ultrasound and electrical impedance tomography. To assess usability we will use a simplified version of the Healthcare System Usability Score (HSUS). According to the International Organization for Standardization (ISO), usability is an outcome of use which can be defined as "the extent to which a system, product or service can be used by specified users to achieve specified goals with effectiveness, efficiency and satisfaction in a specified context of use". Therefore, to appropriately evaluate usability the sample size will also have to take into account the ability to evaluate effectiveness of the instrument in detecting changes in respiratory muscle function.

   For formative usability testing, a sample size of at least n= 30 is generally recommended for quantitative analysis or summative evaluations.

   To evaluate effectiveness, we aim to detect a change in muscle parasternal muscle cross-sectional area, diaphragmatic excursion and lung aeration of 20% within participants who respond and not respond to treatments. We have reviewed previous observational study on diseased participants, and calculated that we would need at least n=30 participants. This is based on data from paired t-tests for the parasternal intercostal muscle, diaphragmatic muscle and lung parenchyma to be evaluated. However, there is no data on electrical impedance tomography assessing the change in lung volume between responders and non-responders. We plan to assess the change in muscle thickness for 2 different muscles (i.e. parasternal intercostal and diaphragm), diaphragmatic excursion and lung parenchyma and volume using two instruments (i.e. ultrasound and electrical impedance tomography). Adjusting the type-1 error rate to alpha=0.015 for multiple testing of the 2 muscles and multiple lung conditions increases the required sample sizes to n= 50. A sample size of n=50 acute respiratory failure patients would therefore be well powered to detect these differences. However, we are unsure about the patients drop out rate, incomplete data and missing data as there is no data available about this. Therefore, allowing for an in hospital mortality of 20%, and a further 20-30% refusal rate/inability to tolerate US and/or EIT, missing and incomplete data we may aim to recruit up to 100 patients.

   For the semi-structured interview, a maximum variation sample size of up to 20 participants (nurses, doctors and AHPs across Barts Health NHS Trust) is the recommended sample size to reach saturation and diversity in qualitative interviews.

   Finally, to assess feasibility (which is a secondary aim) we will evaluate the following to detect events that could compromise the quality or flow of the study such as logistical problems that may disrupt study workflows and technical failures with data collection procedures:

   • Recruitment: Can eligible patients be identified and recruited? How long does it take to enrol the desired number of participants? Are recruitment rates sufficient to meet study targets? If unable to recruit understanding potential reasons and why participants may not wish to take part in the study.

   • Retention: Can we keep participants enrolled in the study throughout its duration?

   • Intervention delivery: can the intervention be delivered as designed and intended? Do participants adhere to the intervention as intended?

   • Data collection procedures: Can data be collected effectively and efficiently? Is data capture complete and reliable for these measures? What percentage of participants complete all the assessments methods?

   • General and safety consideration: How the setting (participant hospital location) impact the feasibility of the intervention? Does the intervention place a significant burden on participants or clinicians? Are adverse events and risks monitored and manageable within the study context? To evaluate feasibility we will use the traffic light system screen (red, amber, green) to quickly assess and communicate the progress, issues, or overall feasibility status of the trial or research study.

     • Green indicates that study in terms of feasibility is proceeding well without major issues, the criteria for success are being met, and the research methods and processes are viable.
     • Amber suggests caution, meaning there are some challenges or uncertainties in the research process that may require adjustments or further investigation but are not yet critical.
     • Red signals significant problems or barriers that may threaten the feasibility of the study, such as recruitment difficulties, methodological flaws, or resource issues that need urgent attention or may lead to stopping the research.

   This visual approach helps to quickly grasp the trial status and make decisions about continuing, modifying, or stopping the research based on early indicators.

   Evaluating these aspects will allow us to examine if the intervention can be realistically implemented, if patients and clinicians will engage with it, and if the necessary data can be gathered effectively in the clinical context.

   Sample size Based on the above considerations, we aim for at least 50 patients and 50 clinical staff members to be retained with full complete data measurements and up to 100 participants may be recruited to allow for incomplete or missing data.

   For the semi-structured interview, if time and resources allow, we will aim to recruit a maximum variation sample size of up to 20 participants is the recommended sample size to reach saturation and diversity in qualitative interviews.

   Method of analysis Characteristics of the study population will be described using descriptive statistics, as appropriate for parametric and non-parametric data. Multiple linear regression will evaluate the associations between outcome variables and the primary and secondary outcomes.

   To assess usability the Healthcare System Usability Scale (HSUS) will be used. The score is converted into percentage in a system ranging from 0 to 100 for rating of usability to allow interpretation.

   Interpretation follows the Acceptability scales range: "Not Acceptable"< 50, "Marginally acceptable" 50-70, "Acceptable"> 70. A usability score between 20% and 50% indicates a critical need to address the system's usability issues; between 50% and 70% indicates a need to address the system's usability concerns, some of which may be major; between 70% and 90% indicates a good usability score with the potential to improve; and between 90% and 100% indicates an excellent and easy to use system.

   For the field notes and the semi-structured interview, we will collect descriptive data about the participants and focus on common challenges, methods used and their issues. Interviews will be transcribed and analysed concurrently with data collection. Data from the initial interviews will be analysed inductively based on the constant comparative method, and informed by any sensitising concepts identified from the Healthcare System Usability framework. A set of initial codes and themes will be generated and used as a framework for further, more deductive, coding whilst remaining open to the possibility of new themes emerging. Finally, these sub-themes will be grouped into high-level themes for each study objective.

   To report the extent and rate of change in respiratory muscle thickness, excursion and lung aeration and volume in acute respiratory failure adults using bedside ultrasound and electrical impedance tomography across six time points, descriptive statistics (e.g., mean, SD) will be used. Repeated measures ANOVA, and independent samples t-test or Mann-Whitney U tests will be used as appropriate to evaluate changes in parasternal intercostal muscle, diaphragm and lung aeration over time between participants. Multiple linear regression analysis will be used to assess the relationship between changes in parasternal intercostal muscle, diaphragm and lung aeration and changes NRS settings and patient outcome (i.e. intubation, death). Correlations will be described using Pearson coefficients or Spearman rho for non-normally distributed or categorical data. Graphical representations will be used to visualise data trends. Statistical analysis will be performed using STAT or SPSS or R software, depending on the complexity of the analysis. Further exploratory statistical analyses may be performed depending on the results of the above analyses.

   Data management Data will be transcribed onto the electronic CRF (eCRF) on the secure data entry web portal. Submitted data will be stored securely against unauthorised manipulation and accidental loss. Only authorised users at Barts Health NHS Trust will have access. Desktop security is maintained through usernames and passwords. Data back-up procedures are in place and a full audit trail will be kept. Storage and handling of confidential trial data and documents will be in accordance with the Data Protection Act 2018 (UK). Access to the final data will be granted only to authorised representatives from the Sponsor, host institution and the regulatory authorities to permit study-related monitoring, audits and inspections to ensure compliance with regulations. We will not transfer clinical data outside of Barts Health NHS Trust.

   Consent Process Patients with acute respiratory failure frequently experience transient or fluctuating impairment in decision-making capacity as a result of hypoxaemia, hypercapnia, delirium, fatigue, or the effects of acute illness and respiratory support. The consent process for this study is therefore designed to be flexible, proportionate, and compliant with the UK Mental Capacity Act (2005), ensuring that participant autonomy and welfare are prioritised while allowing timely enrolment in a time-sensitive clinical context. All patients are formally assessed for capacity by appropriately trained members of the clinical or research team prior to enrolment. Where a patient is deemed to have capacity, written informed consent is obtained before any study-specific procedures are undertaken, following provision of a detailed participant information sheet and an opportunity to ask questions.

   For patients who lack capacity at the time of potential enrolment, a structured delayed consent approach is implemented. In such cases, advice regarding the patient's presumed wishes and best interests is sought from a personal consultee, typically a relative or close friend, where available. If a personal consultee cannot be identified within a clinically appropriate timeframe, agreement is sought from a nominated professional consultee who is independent of the research team and familiar with the patient's clinical care. This process allows inclusion of patients who would otherwise be systematically excluded from research due to acute incapacity, while ensuring that enrolment decisions are ethically justified and appropriately documented.

   Patients enrolled under consultee agreement are re-approached at the earliest appropriate opportunity should they regain capacity, at which point written informed consent is sought for continued participation and for the use of data already collected. Participants are informed that their involvement in the study is entirely voluntary and that they may withdraw at any time without providing a reason and without any impact on their clinical care. If a participant chooses to withdraw, no further data are collected, and data obtained prior to withdrawal are retained for analysis.

   Storage and archiving We will collect personal information (such as name, NHS number and contact details) only where necessary for consent, follow-up and study administration. These identifiable details will be stored securely on NHS systems at Barts Health NHS Trust and kept separate from research data. Research data (including clinical information, ultrasound images and videos, electrical impedance tomography data, questionnaires and interview transcripts) will be pseudonymised using a unique study code. In line with research regulations and Queen Mary University of London policy, essential study data will be stored securely for 25 years after the end of the study. Identifiable information will be securely destroyed once it is no longer required for study administration and follow-up.

Inclusion Criteria:

• Adult (≥18 years old)
• with acute respiratory failure with hypoxia (i.e. arterial oxygen tension (PaO2) of <8.0 kPa), and/or with or without hypercapnia (i.e. arterial carbon dioxide tension (PaCO2) of >6.0 kPa) from any underlying disease or cause
• requiring any non-invasive respiratory support (i.e. HFNO, CPAP, BiPAP)
• Multidisciplinary critical care staff involved in the management of those recruited patients with acute respiratory failure requiring non-invasive respiratory supports. Staff will possibly have an interview and are also required to complete a questionnaire.

Exclusion Criteria:

• Patients in respiratory arrest defined as the total cessation of airflow and breathing effort and absent ventilation (24,25)
• Patients requiring immediate intubation
• Patients with Glasgow Coma Scale (GCS) < 8
• Patients with severe facial trauma or burns
• Patients with fixed upper airway obstruction or inability to protect the airway
• Patients with severe agitation and/or confusion that prevent use of the device mask
• Patients with severe vomiting
• Pregnancy
• Patients with pacemakers and other electronic devices in the thorax
• Patients on end-of-life care or palliative care (defined as expected to die and/or not receiving active treatment)
• Contra-indication to EIT or ultrasound monitoring (e.g. burns, severe obesity, thoracic wounds limiting instrument placement, and thoracic drain)