Standardising Terminology, Classification, Diagnostic Recognition and Reporting of Patient-ventilator Interactions: Protocol for an International Delphi Consensus Study (INTERACT)

Introduction:

Optimal patient-ventilator interaction (PVI) is a state of harmony in which ventilatory support is aligned with the patient's neural respiratory drive and breathing pattern, thereby achieving the core goals of mechanical ventilation, including comfort, safety, and timely liberation from the ventilator. Disruption of this harmony results in abnormal PVI patterns, which have been associated with increased work of breathing (WOB), a higher risk of ventilator-induced lung injury (VILI), greater requirement for sedation and neuromuscular blockade, and prolonged duration of mechanical ventilation and intensive care unit (ICU) stay.

Achieving and maintaining optimal PVI requires clinicians to be familiar with its terminology, understand the underlying physiological mechanisms, and recognise, interpret, prevent, and manage abnormal interactions at the bedside. However, PVI terminology remains highly inconsistent. 'Early triggering', 'reverse triggering', and 'neural or ventilator-induced entrainment', are often used interchangeably, while 'double triggering' and 'double cycling' are inconsistently defined and applied. Descriptors, including 'starvation', 'flow mismatch', and 'under-assistance' are variably applied, and terms including 'missed efforts', 'ineffective triggering', 'ineffective efforts', and 'failed trigger' are frequently used to describe the same abnormality. In parallel, broader labels such as dyssynchrony, asynchrony, and PVI patterns coexist without a standardised conceptual hierarchy. This inconsistency reflects a fundamental lack of consensus on how PVI phenomena should be defined and classified.

This lack of standardisation creates confusion and hinders accurate recognition, education, and communication, limits comparability in reporting across studies, and impedes the development of automated detection systems. Although recent initiatives have moved toward greater standardisation, adoption remains incomplete and legacy terminology continues to be widely used.

To address these challenges, we plan an international Delphi consensus, The INTERACT (International Consensus on Patient-Ventilator Interaction Terminology, Classification, Diagnostic Recognition and Reporting) study, to develop a standardised framework for PVI. Specifically, we aim to harmonise terminology, establish a clear and clinically meaningful classification of PVI patterns, define the key diagnostic features and minimal monitoring requirements for their recognition, and propose standardised reporting recommendations for both research and clinical practice. By building expert consensus across disciplines and regions, this initiative seeks to create a shared and practical foundation to improve communication, education, research comparability, and ultimately patient care.

Methods:

Design:

The INTERACT study will be conducted as an international, modified Delphi process using multiple rounds of anonymous surveys, with iterative feedback to refine and consolidate expert opinion. Delphi rounds will be administered via the Calibrum platform (www.calibrum.com), with individualised secure links provided to each panellist. Two Delphi methodologists (PK and PN) will oversee the entire Delphi process, and all statements will be reviewed for appropriateness by two experienced Delphi-investigators. The study will adhere to the ACCORD (ACcurate COnsensus Reporting Document) reporting guideline to ensure transparency and methodological rigour.

Ethics, Consent, and Confidentiality:

The study will be conducted in accordance with the Declaration of Helsinki. Formal ethics approval has been granted by the institutional review board, University Clinical Centre of the Republic of Srpska (IRB approval number: 01-19-206-2/26). Electronic informed consent will be obtained from all participants. Participation will be voluntary, and panellists may withdraw at any time, and their data will be excluded from the analysis. Panellists will remain anonymous to each other, while the Delphi methodologists will retain identifying information solely for administrative purposes.

Steering Committee:

The steering committee consists of 12 multidisciplinary members, with a balanced representation including adult ICU physicians, anaesthesiologists, respiratory monitoring experts, mechanical ventilation educators, and Delphi consensus methodology investigators.

Steering committee members (co-authors) were selected with the following criteria: at least 10 years of clinical experience in invasive mechanical ventilation, respiratory physiology monitoring, and PVI; or peer-reviewed publications related to mechanical ventilation, respiratory monitoring, or PVI; or recognised educational, leadership, or guideline-related role in mechanical ventilation, respiratory physiology, respiratory physiology monitoring, or PVI; or experience in conducting Delphi studies and consensus statement.

Steering committee composition followed purposive sampling and a diversity criteria: balanced representation through inclusion of experts from different world regions and continents; and professional discipline via representation from intensive care medicine, anaesthesiology, respiratory monitoring, mechanical ventilation education and Delphi consensus; and gender and institutional balance, with efforts made to avoid over-representation from a single sex, institution, or country.

The steering committee is responsible for defining study scope, developing and refining the study domains, overseeing literature search, evidence synthesis, identification of knowledge gaps and statement drafting, reviewing feedback reports and supervising protocol fidelity, revising survey statements between rounds on the basis of predefined revision rules, panel-level response distributions and panellist feedback, and leading manuscript preparation and dissemination.

Delphi Methodologists:

Two members of the steering committee will serve as Delphi methodologists. They will provide dedicated methodological oversight throughout the consensus process. Designated Delphi methodologists will be responsible for survey architecture and round design, item revision rules and audit trails. Moreover, the Delphi process will include predefined strategies to minimise cognitive biases among both the steering committee and expert panellists. These strategies and their implementation will be transparently reported in accordance with established reporting standards.

Conflict of Interest:

All participants will be required to disclose conflicts of interest in accordance with the recommendations of the ICMJE (International Committee of Medical Journal Editors). Disclosure will include any financial, professional, or advisory relationships that may be relevant to the scope of the study, including (but not limited to) relationships with ventilator or monitoring technology industries, ties related to Electrical Activity of the diaphragm (EAdi), Electrical Impedance Tomography (EIT), or waveform analysis technologies, and paid consultancy roles, speaker's honoraria, or advisory roles in direct or indirect relation to the study scope.

Panel Selection:

A total of 40-50 international panellists will be invited. Where feasible, balanced representation will include ICU physicians and anaesthesiologists, respiratory therapists, ICU nurses with advanced ventilation expertise, and researchers or engineers specialising in waveform analysis or respiratory monitoring. Panellists will be recruited through a purposive sampling from key publications and established international networks. Invitations will include a study protocol, and an expected time commitment. If an invited panellist declines participation, a replacement will be selected from a predefined reserve list.

Panellist must meet at least one of the following: at least 10 years of experience managing invasively ventilated adult patients with demonstrated PVI expertise (education, quality improvement, or leadership roles); authorship of at least 2 peer-reviewed publications related to PVI, asynchrony, waveform analysis, or respiratory monitoring; or participation in at least one regional or international clinical guidelines or task forces related to mechanical ventilation or PVI.

Panel selection will follow best practices from established Delphi protocols, and ensuring diversity: no more than 70% of panellists from one sex; representation across diverse resource settings, using a pragmatic ICU capability-based definition rather than World Bank income categories alone; and no more than one panellist will be selected from the same centre or institution, to minimise institutional overrepresentation and promote broader international and multidisciplinary input.

Delphi Domains:

The steering committee defined a set of domains that reflect the core areas required to standardise PVI terminology, classification, diagnostic recognition, and reporting. These domains were derived from a focused review of the existing literature (based on PRISMA checklist), and recognised gaps in current practice.

Domain A, terminology and definitions, will address standardisation of terms and definitions for core PVI patterns, including triggering abnormalities, cycling abnormalities, effort-ventilation mismatch, and entrainment. This domain will also harmonise the existing terms used across relevant disciplines, including respiratory physiology, mechanical ventilation, engineering, and waveform analysis with the aim for identification of redundant, overlapping, or conflicting terms. Of note, examples are illustrative and non-exhaustive, and panellists will be encouraged to propose additional concepts and terminology throughout the Delphi process.

Domain B, classification framework, will focus on the principles for grouping PVI patterns (e.g., phase-based versus mechanism-based classification). This domain will explore whether PVI should be organised within hierarchical or non-hierarchical framework, define the criteria for classifying abnormal versus normal variability, consider inclusion or exclusion of effort-based versus ventilator-based components, and assess alignment with physiological mechanisms and clinical relevance.

Domain C, diagnostic features and waveform recognition, will address the key waveform signatures associated with each PVI pattern. It will also consider the role of ventilator scalars, loops, and advanced monitoring tools, such as oesophageal pressure and EAdi, in supporting diagnostic recognition. In addition, this domain will seek agreement on the minimum monitoring signals required to classify specific PVI patterns and on standard descriptors for waveform abnormalities.

Domain D, reporting standards, will focus on development of a minimum reporting dataset for documenting abnormal PVI. Rather than redefining PVI itself, this domain will standardise how identified abnormalities should be reported, including the type of the type of abnormality, patient phenotype, ventilator mode, monitoring signals used for detection, frequency of occurrence and when feasible severity.

Delphi Rounds and Procedures:

A formal literature review informs the statements for Round 1; the initial survey will be shared along with the supporting literature.

In Round 1, panellists will be invited to provide initial ratings of statements, free-text feedback on clarity of the statements, and proposal of missing terms or domains. Of note, the steering committee members will not vote in Delphi rounds.

In Round 2, feedback obtained during Round 1 will be utilised to revise the statements as necessary. Statement modifications will require congruent responses from at least 10% of the participating panellists for that round. Any recommendations regarding missing terms and domains will be addressed by the steering committee to determine their potential inclusion. Revised statements, including group-level response distributions and anonymised comments, will be shared in Round 2.

In subsequent rounds, further rounds will be conducted until consensus and stability are reached across statements. Items that remain ambiguous or unstable may be revised and re-tested according to predefined revision rules.

Question Formats:

The questions in the survey will be either 7-point Likert scale or multiple-choice items, with single or multiple options. The 7-point Likert scale consists of the following items: 1 signifies strong disagreement, 2 indicates disagreement, 3 reflects slight disagreement, 4 represents a neutral stance, 5 conveys slight agreement, 6 denotes agreement, and 7 denotes strong agreement. For the purpose of analysis, ratings of 1 to 3 will be counted as disagreement, while ratings of 5 to 7 indicate agreement.

Consensus and Stability Criteria:

Consensus will be defined as ≥ 75% agreement or disagreement on Likert-scale items, or ≥ 75% selection of a single option in multiple-choice items. Stability will be assessed from Round 2 onwards using Kruskal-Wallis tests for Likert-scale items and chi-square tests for multiple-choice items. Items demonstrating thematic ambiguity and congruent feedback from ≥ 10% of panellists will be revised and re-tested.

Closing of the Delphi Process:

The total study duration is anticipated to be approximately 12 months, encompassing panel recruitment, at least three Delphi rounds, and final ratification of the consensus and resulting practice statement. The Delphi process will continue until stability is reached across all statements. Final stable statements, including consensus percentages and practice recommendations, will be circulated to panellists for comment-only review without additional voting.

Attrition Management:

Attrition will be minimised through explicit clarification of roles and responsibilities in the invitation email, at least three e-mail reminders per Delphi round, and the provision of technical support throughout the process. An attrition rate exceeding 20% will trigger sensitivity analyses to assess the impact of non-response on results.

Data Management and Analysis:

All data will be stored on secure institutional or approved cloud-based systems with restricted access controls. Data will be retained for a minimum of 5 years. Quantitative analyses will include descriptive statistics, assessment of consensus attainment, and stability testing. Qualitative data will be analysed using thematic analysis with dual independent coders and adjudication by a third reviewer when necessary.

Bias mitigation:

Several measures have been incorporated to minimise the risk of bias across panel selection, statement development, the voting process, and dissemination.

To reduce selection bias and spectrum bias, panellists will be recruited through purposive sampling with explicit diversity criteria including geography, professional discipline, sex, and institutional affiliation. Panel composition includes ICU physicians, anaesthesiologists, respiratory therapists, and ICU nurses. No more than one panellist will be selected from a single institution, and no more than 70% from one sex.

To address conflict of interest, all participants, including steering committee members, are required to disclose financial, professional and advisory relationships in accordance with ICMJE recommendations.

To mitigate framing bias arising from steering committee's role in statement development, two designated Delphi methodologists independently review all statements for clarity and neutrality before each survey round is deployed. Pre-specified revision rules govern statement modification between rounds, ensuring that changes respond to panellist feedback rather than steering committee preference.

To prevent dominance and conformity, panellists remain fully anonymous to one another throughout the study. The 7-point Likert scale will be used to reduce response scale bias.

To manage attrition bias, response rates will be monitored across all rounds. To prevent dropout, a minimum of three reminders and dedicated technical support will be provided per round.

To prevent reporting bias, both consensus and non-consensus statement items will be reported in the final manuscript.

Authorship and Dissemination:

Participation as a Delphi panellist will not automatically confer authorship. Authorship will be determined in accordance with International Committee of Medical Journal Editors (ICMJE) criteria. Members of the steering committee who meet authorship criteria will be listed as co-authors of the final manuscript. Delphi panellists will be acknowledged collectively as collaborators unless they meet formal authorship requirements. This authorship policy will be clearly communicated in all invitation materials and study documentation. Findings will be disseminated through peer-reviewed publications, conference presentations, and an open-access toolkit, including the glossary and reporting checklist.

Discussion:

Current literature shows substantial variability in terminology, heterogeneous classification schemes, and inconsistent reporting methods across disciplines and institutions, which collectively hinder clinical interpretation, education, cross-study comparability, and the development of automated detection systems. The INTERACT study address a recognised gap in critical care by structuring a Delphi process around clearly defined domains to creates a systematic foundation for consensus building. The study domains reflect the four core objectives of the initiative: harmonising terminology, standardising classification, defining diagnostic features, essential monitoring requirements, and standardising reporting. The domains ensure that the panellists evaluate PVI from multiple perspectives: physiological mechanisms, waveform characteristics, clinical implications, and practical feasibility across diverse healthcare systems and ventilator platforms.

Importantly, the domains are intentionally broad and iterative. Round 1 will allow panellists to refine the conceptual structure, identify missing elements, and suggest alternative formulations. This flexibility supports a transparent, expert-driven process and ensures that the final framework is inclusive, internationally relevant, and aligned with both current evidence and real-world clinical experience. Ultimately, these domains are intended to support the development of clear, standardised, and clinically meaningful recommendations for PVI terminology, classification, diagnostic recognition, and reporting. These outputs are expected to improve consistency in bedside recognition, education, research reporting, inter-study comparability and the future development of automated detection systems.

The INTERACT study is designed to address this gap through three outputs: the INTERACT Glossary will establish the preferred terms; the INTERACT Taxonomy will provide a structured classification reflecting physiological mechanisms and clinical relevance; and the INTERACT Minimum Reporting Dataset will standardise how abnormal PVI is documented in both research and clinical practice. These outputs will collectively aim to create a shared language for PVI usable across diverse healthcare systems and professional disciplines.

Several methodological strengths support the credibility of the INTERACT study. The Delphi panel is designed to be broad, multidisciplinary and multinational, with representation from ICU physicians, anaesthesiologists, respiratory therapists, ICU nurses and researchers with relevant expertise. Purposive sampling, limits on institutional over-representation and explicit diversity criteria are intended to reduce selection and spectrum bias. Independent methodological oversight, anonymised voting, non-voting steering committee members, predefined revision rules and reporting of both consensus and non-consensus statements are incorporated to reduce framing, dominance, conformity and reporting bias. Open-text feedback in early rounds will allow panellists to identify missing domains, propose additional terminology and will improve the clarity of the statements before final consensus is assessed.

The main limitation is that the output will represent consensus-based recommendations rather than clinically validated standards. Therefore, the glossary, taxonomy and reporting dataset should be interpreted as a structured consensus framework requiring prospectives validation. The study is also limited to adult patients receiving invasive mechanical ventilation; consequently, its direct applicability to non-invasive ventilation, home ventilation and paediatric ventilation will be limited. In addition, although independent methodologists and predefined rules are used to mitigate framing bias, the steering committee still has central role in developing and refining survey statements. Finally, the study will not test whether implementation of the INTERACT outputs improve diagnostic reliability, clinical decision-making, educational outcomes or patient-centered outcomes; these questions will require subsequent validation and implementation studies.

Following completion of the Delphi process, the consensus outputs will be translated into practical tools for clinical, educational, and research use. These may include a glossary document, a taxonomy figure suitable for publication and educational, a reporting checklist derived from the minimum reporting dataset, and potentially an educational waveform atlas. The Delphi results may identify future research priorities, particularly for items that fail to reach consensus or require clearer operational definitions. Subsequent studies should evaluate the reliability, clinical usability, and educational value of the INTERACT glossary and taxonomy as well as implementation studies assessing whether standardising PVI terminology and classification improves recognition and communication among ICU teams.

In conclusion the INTERACT study is designed to address a well-recognised but unresolved gap in critical care: the absence of standardised, internationally agreed terminology, classification, diagnostic recognition, and reporting standards for PVI. Through a structured modified Delphi process, the study aims to generate the INTERACT Glossary, the INTERACT Taxonomy, and the INTERACT Minimum Reporting Dataset. These outputs are expected to improve consistency in bedside recognition, education, research reporting, inter-study comparability, and the future development of automated detection systems for abnormal PVI.