Turning clinical reasoning research toward intervention: Validating a process model using VR simulation data

Turning clinical reasoning research toward intervention: Validating a process model using VR simulation data | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Turning clinical reasoning research toward intervention: Validating a process model using VR simulation data Constanze Richters, Kapotaksha Das, Frank Fischer, Martin R. Fischer, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9497414/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract Clinical reasoning in acute care unfolds under time pressure as teams must continuously interpret evolving patient information while coordinating treatment. While research has predominantly focused on diagnostic reasoning, it remains insufficiently understood how care teams generate, coordinate, and enact interventions, and how these processes are organized. To address this gap, we introduce the Collaborative Diagnostic–Intervention Reasoning (CDI-R) model, a theory-driven process model that conceptualizes clinical reasoning as the interplay of diagnostic activities (DAs), intervention activities (IAs), and collaborative activities (CAs). We provide an initial empirical examination of the model using a virtual reality cardiac arrest simulation with 29 teams ( N = 116 participants). Team interactions were coded at the utterance level. Lag Sequential Analysis (LSA) was used to examine transitions between DAs and IAs, and Epistemic Network Analysis (ENA) was used to examine the structural organization of CAs. Findings reveal three key patterns. First, DAs and IAs constitute structurally distinct reasoning modes, characterized by different configurations of CAs. Second, reasoning unfolds in non-linear and recurrent sequences, with sustained engagement within and transitions between reasoning modes. Third, expertise shapes the organization of reasoning: expert-led teams engaged in a higher proportion of IA-oriented activity, transitioned more frequently from diagnosis to intervention, and showed more differentiated collaborative structures, whereas trainee-led teams exhibited more loops within reasoning modes. By explicitly integrating intervention reasoning into models of clinical reasoning, this study advances a process-oriented account of team-based reasoning. The CDI-R model provides a framework for examining how teams coordinate diagnosis and intervention in dynamic clinical settings. Clinical reasoning Intervention reasoning Team-based reasoning Acute care Clinical expertise Simulation-based training Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 1 Introduction Medical routine cases, such as a finger laceration or an epileptic seizure in a patient with known and pre-diagnosed epilepsy, can be managed by a single clinician. By contrast, high-risk situations such as cardiac arrest or undifferentiated symptom(s) (e.g., headache, abdominal pain, or undifferentiated fever) often require team-based collaboration of multiple clinicians. Clinical reasoning in such teams can be conceptualized as involving two interdependent processes: diagnostic reasoning (Fischer et al., 2014 ), which aims at reducing uncertainty regarding the etiology of symptoms by gathering and integrating patient information, and intervention reasoning (Richters et al., 2024 ), which focuses on generating different options, selecting and executing treatments to address the diagnosed problem. These reasoning processes may unfold concurrently, for instance, at the patient’s bedside, or sequentially, as in referrals between clinicians. In acute care, such as cardiac arrest management, diagnostic and intervention reasoning unfold concurrently under time pressure, as clinicians work together at the patient’s bedside. These situations are highly dynamic as conditions may change quickly over short periods of time and require constant diagnostic and intervention adjustments (see Stadler et al., 2019 ). Care team leaders and members navigate these time-sensitive activities together through collaborative activities, such as joint information processing, coordination, and monitoring, as well as structured communication techniques, such as closed-loop communication (Liu et al., 2016 ; Meier et al., 2007 ; Salas et al., 2008 ; Witti et al., 2023 ). Recent research has begun to examine how collaborative processes support diagnostic reasoning (Brandl et al., 2024 ; Radkowitsch et al., 2022 ; Richters et al., 2023 ). However, intervention reasoning has received much less attention (Cook et al., 2019 ; Monajemi & Rikers, 2011 ; Richters et al., 2024 ). A recent scoping review further shows that intervention reasoning research remains poorly integrated, with studies rarely building on each other and lacking a coherent, connected body of literature (Duong et al., 2023 ). It remains unclear how teams coordinate knowledge and actions to reach and implement decisions on interventions in dynamic clinical settings - and where and how they struggle. This is surprising given the central role of intervention reasoning for patient outcomes (Abdoler et al., 2023 ; Cook et al., 2019 ; Kumar & Basheer, 2021 ; Monajemi & Rikers, 2011 ; Penner et al., 2025 ). To address the conceptual and empirical gap in research on intervention reasoning in clinical reasoning research, this study proposes and validates the CDI-R process model that integrates diagnostic reasoning, intervention reasoning, and collaboration. The model aims to capture how these reasoning processes unfold in real-time team interactions. We validate CDI-R using VR-based cardiac arrest simulations, comparing diagnostic, intervention, and collaborative activity patterns across levels of clinical expertise. 1.1 Collaborative diagnostic reasoning and clinical expertise Diagnostic activities (DAs) involve the goal-directed collection, evaluation, and integration of patient-specific information to reduce uncertainty (Fischer et al., 2014 ; Heitzmann et al., 2019 ). DAs have been examined with respect to hypothesis generation, evidence evaluation, and pattern recognition (e.g., Fink et al., 2021 ; Tay et al., 2016 ; Gruppen, 2017 ). From a cognitive perspective, diagnostic reasoning relies on both backward and forward reasoning (Boshuizen & Schmidt, 1992 ), with expert clinicians often using forward reasoning and rapid pattern recognition (System 1), supported by reflective, analytical processes (System 2) when confronted with uncertainty (Kahneman, 2003 ; Norman et al., 2024 ). Diagnostic expertise is reflected through structured mental representations (see Simon & Newell, 1971 ), such as illness scripts, which enable fluid diagnostic inferences and efficient cue interpretation (Charlin et al., 2007 ). Expertise theory holds that experts solve problems with fewer steps, relying on structured knowledge and pattern recognition (e.g., Boshuizen & Schmidt, 1992 ; Ericsson et al., 1991). In collaborative diagnostic reasoning, collaborative activities (CAs) play a critical role. In acute care settings, such CAs typically include sharing information between and eliciting information from team members, coordination, mutual monitoring, and structured communication techniques such as closed-loop communication (e.g., Salas et al., 2005 ; Meier et al., 2007 ; Liu et al., 2016 ; Witti et al., 2023 ). The relevance of structured communication for safety-critical coordination is emphasized in frameworks like TeamSTEPPS (Department of Defense & Agency for Healthcare Research and Quality, 2010). Expertise in these CAs can be described in terms of internalized collaboration scripts (Fischer et al., 2013 ) and transactional routines that guide teams through complex, time-pressured decisions (Janssen & Kirschner, 2020 ). The collaborative diagnostic reasoning (CDR) model (Radkowitsch et al., 2022 ) systematically integrates CAs with DAs, demonstrating how shared understanding and coordinated reasoning contribute to diagnostic accuracy in teams. The CDR model has recently been validated in a clinical context involving diagnostic collaboration between an internist and a radiologist, including the request and interpretation of radiological imaging (Brandl et al., 2024 ). While this model is valuable for understanding how diagnosticians share and coordinate evidence and hypotheses, it remains diagnosis-centered. Structurally, it maps reasoning across distributed knowledge spaces and emphasizes social and cognitive skills in building shared diagnostic representations. However, it does not distinguish intervention reasoning as a separate or equally complex process, and the dynamic interplay between diagnostic reasoning and intervention reasoning in teams remains unclear. 1.2 Collaborative intervention reasoning: Conceptualization and theoretical integration Although intervention reasoning is essential, it has rarely been explicitly modeled. Primarily it is described as part of broader clinical reasoning skills (e.g., Liaw et al., 2010 ; Monajemi et al., 2012 ) or in the context of a concrete therapeutic case (e.g., Wilcox & Heudes, 2017 ). Richters et al. ( 2024 ) recently defined intervention activities (IAs) as cognitively and interactionally distinct but functionally interdependent subcomponents, namely intervention generation, selection, implementation (i.e., planning and executing), and evaluation, that often build on each other and vary in visibility depending on the clinical context. For example, in cardiac arrest, where high acuity and time pressure demand urgent intervention, implementation and immediate coordination are often crucial and explicit, whereas in chronic care, planning and adaptation may be more prominent but implicit (see Stadler et al., 2019 ). This variation between acute and chronic care contexts highlights the importance of not only what IAs are performed, but also how they are reasoned through and coordinated. The role of CAs in orchestrating intervention sequences, particularly in groups and teams, remains understudied. Moreover, IAs place distinct demands on CAs, requiring immediate, targeted communication such as task distribution or confirmation loops. 1.3 CDI-R: A process model for team-based diagnostic and intervention reasoning To address the lack of systematic research on intervention reasoning as a distinct cognitive process in collaborative clinical settings, we propose CDI-R as a theory-driven process model for collaborative diagnostic and intervention reasoning in dynamic clinical settings (see Fig. 1 ). Building on the CDR model (Radkowitsch et al., 2022 ), and drawing on research from cognitive science, clinical reasoning research, and computer-supported collaborative learning (CSCL), the CDI-R model introduces IAs as distinct from yet interdependent with DAs, both dynamically regulated by CAs. The model visualizes these components through concentric circles. This cyclical structure reflects the recursive, time-sensitive nature of collaborative reasoning in complex environments, such as cardiac arrest, where diagnosis and intervention evolve through continuous team collaboration. The outer layer of the model represents the domain-general task space (see cross-domain task space suggested by Hetmanek et al., 2018 ), encompassing primary DAs and IAs. The relations between these activities are non-linear because diagnostic and intervention reasoning often occur in parallel, interrupt each other, or recur iteratively in response to changing clinical conditions and team dynamics, capturing the iterative, non-sequential nature of real-time reasoning. Importantly, CDI-R conceptualizes this reasoning process as embedded within broader input variables and outcomes (see Rosen et al., 2018 ). Individual characteristics (e.g., prior knowledge, cognitive skills, emotions, expertise; e.g., Duffy et al., 2020; Pekrun, 2006), team characteristics (e.g., composition, familiarity; Fiore, 2011), task characteristics (e.g., complexity, interdependence; e.g., Cohen, 1987), and environmental features (e.g., authenticity, fidelity) shape how collaborative diagnostic and intervention reasoning unfolds. Likewise, the process gives rise to outcomes at the task level (e.g., diagnostic and intervention-related outcomes), the individual level (e.g., learning gains, motivation), and the team level (e.g., quality of collaboration, emergent states). While these boundary conditions and consequences are integral to a full CDI-R framework, the present model description centers on the dynamic process component interlinking DAs, IAs, and CAs. Underpinning the CDI-R model are several core propositions. First, DAs and IAs constitute conceptually and functionally distinct reasoning processes (Proposition A). Second, these processes may repeat, interrupt each other, co-occur, or shift rapidly in response to unfolding conditions and task demands (Proposition B). This includes both sustained engagement within a single reasoning mode (e.g., repeated DA→DA or IA→IA sequences) and transitions between modes (e.g., DA→IA or IA→DA), reflecting different temporal organizations of collaborative reasoning. This structure illustrates that diagnostic and intervention reasoning are not staged or unidirectional but dynamically interact under time pressure. At the core of the model lies the rotating system of CAs, divided into four interrelated categories: joint information processing, structured communication, coordination, and monitoring. While communication underpins all CAs, structured communication with the example of closed-loop communication is modeled separately due to its verification-based format and distinct team function (Department of Defense & Agency for Healthcare Research and Quality, 2010). These CAs, in their dynamic organization, shape and coordinate ongoing DAs and IAs, while also being shaped by them. For example, closed-loop communication helps verify that instructions such as administering a specified medication at the correct dosage are both received and executed correctly, reducing the risk of errors. Similarly, structured information sharing allows team members to integrate rapidly changing, frequently recorded clinical cues such as ECG information or vital signs, enabling timely and coordinated intervention. The rotational design reflects that different types of CAs become more or less prominent depending on the phase of the task and team needs. For example, coordination intensifies during intervention implementation, such as in cardiac arrest resuscitation when the team leader assigns roles (e.g., “You do chest compressions, I’ll summarize the case”) or delegates tasks (e.g., “John, administer 1 mg epinephrine now”), while information sharing dominates during hypothesis generation, such as when team members verbalize findings (e.g., “No pulse, rhythm shows ventricular fibrillation”) or elicit input (e.g., “Do we have a shockable rhythm?”; “What am I missing, team?”). Accordingly, the model proposes that CAs vary in their relevance and function depending on the evolving demands of the task and reasoning phase (Proposition C). Role of expertise The CDI-R model also addresses the role of clinical expertise in shaping reasoning dynamics. Rather than assuming general differences in team performance, we propose that clinical expertise shapes how DAs and IAs are coordinated within teams. Clinical expertise is expected to shape how teams coordinate and transition between DAs and IAs under time pressure (Proposition D1), as more experienced clinicians are better able to align DAs and IAs in a timely and goal-directed manner. This assumption is grounded in the idea that more advanced knowledge structures enable clinicians to anticipate developments, prioritize actions, and coordinate activities more effectively within dynamic team settings. We propose that expert-led teams engage in timely, targeted IA-oriented reasoning supported by collaborative processes aligned with evolving task demands. This includes structured coordination and verification processes (e.g., role allocation or check-backs; Liu et al., 2016 ; Salas et al., 2008 ), which enable smooth and safe implementation of interventions. In contrast, trainee-led teams are expected to show greater persistence within reasoning modes and a reduced likelihood of transitioning from DAs to IAs, reflecting less differentiated coordination under time pressure and reduced fluidity in transitions between reasoning modes. Thus, expertise is reflected not simply in better outcomes, but in how teams organize and transition between DAs and IAs in real time. Building on this, the model further proposes that clinical expertise is associated with differences in the extent and organization of explicitly enacted IAs (Proposition D2). Importantly, prior research on expertise in diagnostic reasoning suggests that experts often require fewer observable steps due to pattern recognition and more efficient processing (Boshuizen & Schmidt, 1992 ; Norman et al., 2024 ). From this perspective, one might expect experts to engage in fewer observable reasoning activities overall. However, we argue that this assumption does not directly transfer to IAs. In contrast to DAs, IAs often require the explicit enactment of multiple steps (e.g., generation, selection, implementation, and evaluation; Richters et al., 2024 ), particularly under time pressure and uncertainty, where actions must be coordinated, communicated, and verified within the team. As a result, even when experts recognize situations rapidly, they may still need to explicitly enact and coordinate intervention steps, making IAs less compressible than DAs. From this perspective, greater clinical expertise may be reflected in a more extensive and systematically organized engagement in IAs, rather than in a reduction of observable activity. This does not imply inefficiency, but rather reflects analytic engagement with the situation, such as anticipating complications, running parallel strategies, or deliberately verifying outcomes. Expertise may thus manifest not in doing fewer IAs, but in organizing and sequencing them more effectively under dynamic conditions. Prior empirical findings support this perspective: intervention-focused reasoning tends to increase with clinical expertise (Monajemi et al., 2012 ), and even experienced clinicians invest considerable cognitive effort in dynamic care situations (Hartjes et al., 2024 ). Consistent with this, experts tend to anticipate critical developments earlier and organize team actions accordingly when recognizing the severity of a situation (Popov et al., 2025 ). 2 Empirical model validation Based on the CDI-R process model, we conducted an empirical, theory-guided study to examine its core propositions in a simulated acute care context. We analyzed data from a virtual reality (VR) simulation of cardiac arrest management (Kentros et al., 2025 ; Popov et al., 2023 ) involving expert- and trainee-led teams. The study investigates how DAs, IAs, and CAs unfold and interact over time, and how these patterns vary based on clinical expertise. Guided by the model, we explored three central aspects: First, we explored whether diagnostic and intervention reasoning can be empirically distinguished in both their temporal organization and communicative structure (Proposition A), and whether they exhibit non-linear, recurrent transition patterns over time (Proposition B). Second, we investigated whether CAs vary systematically depending on whether teams are engaged in diagnostic or intervention reasoning (Proposition C). Third, we examined how expertise shapes these dynamics (Propositions D1 and D2), focusing on differences in (a) the distribution of DAs and IAs, (b) the fluidity of transitions between them (i.e., how flexible teams switch between DAs and IAs), and (c) the structural organization of CAs within each reasoning mode. The study evaluates the extent to which the observed temporal and structural patterns in team communication correspond to the theoretical propositions of the CDI-R model. 3 Method 3.1 VR simulation and participants This study uses a multi-user “open-world” VR simulation (Popov et al., 2023 ) that enables a team of four clinicians to interact with each other and a virtual patient in a dynamic and non-deterministic in-hospital cardiac arrest resuscitation scenario. VR offers a realistic yet controlled setting for capturing the time-pressured interplay of teams’ DAs, IAs and CAs. To effectively complete the ~ 15-minute simulation, the team must designate individual roles and responsibilities, such as team lead, chest compressions, airway management, electrical shock management, and medication administration (Fig. 2 ). The six-stage clinical scenario begins with the care team receiving handoff from a bedside nurse non-player character and a virtual patient being in Ventricular Tachycardia with a weak pulse (Stage 1), requiring synchronized cardioversion and intubation. The patient rapidly deteriorates into pulselessness (Stages 2–3), Asystole (Stage 4), and Ventricular Fibrillation (Stage 5), demanding continuous CPR, unsynchronized defibrillation, and appropriate medication (epinephrine, antiarrhythmics). The simulation is designed to conclude upon Return of Spontaneous Circulation (ROSC; Stage 6), when appropriate interventions are performed, prompting a shift to post-resuscitation care. Participants must communicate dynamically, exchanging information under time pressure and dynamically reallocating tasks. Participants were placed in the same room, each wearing a VR headset (i.e., facial cues unavailable) and standing in a semicircle to facilitate verbal coordination. A convenience sample was recruited as part of a mandatory monthly code team training at a large U.S. academic medical center with extensive experience in cardiac arrest simulation. In total, we analyzed data from 29 simulated VR cardiac arrest sessions ( N = 29 teams, 116 participants in total). Each 4-person session was directed by a single, non-repeating participant serving as the designated team leader, yielding nine expert-led teams and twenty trainee-led teams. Expert status was defined as current Advanced Cardiovascular Life Support (ACLS) certification and at least 5 years of experience leading cardiac arrest teams. The nine expert team leaders consisted of a diverse cohort of experienced clinicians: nurses ( n = 4), paramedics ( n = 2), emergency medicine physicians ( n = 2), and a family medicine physician ( n = 1). In five of these sessions, the two non-leader roles (e.g., procedural tasks of chest compressions and/or airway management) were filled by research staff. The twenty trainee-led sessions (hereafter referred to as trainee-led) were directed by junior emergency medicine ( n = 13) or family medicine ( n = 7) residents. These trainee leaders had completed prior ACLS training but possessed little to no experience leading real cardiac arrest teams. In these simulations, the three non-leader roles were filled by fellow residents. 3.2 Data coding Team interactions were automatically recorded and transcribed. Transcripts were segmented into communicative events at the utterance level using ELAN (Wittenburg et al., 2006 ). Annotations were performed by six domain experts in emergency medical services using a theory-informed coding scheme derived from the CDI-R model (Table 1 ). Two dimensions were coded: CAs and the reasoning mode, i.e., whether it refers to diagnosing (DA) or intervening (IA). To establish coding reliability, Kappa scores were calculated from three recordings (each approximately twelve minutes long). The analysis demonstrated inter-annotator agreement: 0.73 for the main codes, 0.64 for the sub-codes (specific CAs), and 0.82 for the reasoning focus (DA vs. IA). Although some subjectivity remains due to the task’s inherent complexity, these expert-validated transcripts constituted a robust ground-truth dataset for our analyses. Table 1 Coding scheme Main Code Sub-code Definition Examples Joint Information Processing Sharing Information (CA) Providing factual information (new or synthesis of existing), including recapping “We’re at one minute, 30 seconds right now” “Latest vitals show…” Questioning / Eliciting Information (CA) Requesting information from others “What are the latest labs?” “Any new findings?” Evaluating Shared Information (CA) Discussing/assessing provided information, explaining rationale and justifications “Vitals are concerning”! “I disagree” Structured communication CLC: Checkback Statement (CA) Receiver repeats instructions/information for verification (intervention-related only) “Yes, I will shock at 200J” Coordination Assigning Roles (CA) Designating roles or verifying expertise “You do airway, I'll do meds” “You handle respiratory, he’ll do IV” Allocating Tasks (CA) Distributing tasks and implementing plans “Shock at 150J” Monitoring Expressing Uncertainty (CA) Voicing doubts or lack of confidence “Not sure what’s causing this” “His other EKG was looking like he might be having an MI?” Sharing Hypothesis (CA) Formulating working diagnosis “Patient may become septic” Diagnostic- and Intervention reasoning Diagnosis-related (DA) Collection and interpretation of case-specific information to reduce diagnostic uncertainty Assessment of symptoms, lab results interpretation Intervention-related (IA) Activities improving patient condition through generation, implementation, monitoring, and evaluation Treatment planning, medication administration Note. CA = Collaborative Activity, DA = Diagnostic Activity, IA = Intervention Activity. 3.3 Analyses The analytic approach was designed to examine both the temporal dynamics and the structural organization of team-based diagnostic and intervention reasoning, in line with the study’s propositions regarding the distinction between DAs and IAs (A), their non-linear and recurrent organization (B), the variation of CAs across reasoning modes (C), and expertise-related differences in these dynamics (D1 and D2). Analyses were conducted at the team level. To capture these complementary dimensions, we combined Lag Sequence Analysis (LSA; Bakeman, R., & Quera, V., 2011 ; Sackett, G. P., 1979 ) to model temporal progression and Epistemic Network Analysis (ENA; Shaffer et al., 2016 ) to examine patterns of co-occurrence and connections between CAs. These combined analytical approaches provide both a sequential and a relational perspective on team-based reasoning. Lag Sequence Analysis (LSA) We employed LSA to quantify how teams transitioned between DAs and IAs over time, directly addressing Propositions B (non-linearity and recurrence) and D1 (expertise-related transition dynamics). We constructed transition matrices to calculate the conditional probability P(Next|Current) of transitions from a current activity to a subsequent activity. For example, we estimated how likely it was that a DA was followed by another DA or by an IA in the next utterance. These probabilities capture how teams move between reasoning modes during interaction. Sustained team behaviors were indicated by elevated conditional probabilities for self-transitions (DA→DA or IA→IA), reflecting persistence and recurrence within interaction dynamics. In contrast, directed forward progression was characterized by higher probabilities for DA→IA transitions alongside lower rates of IA→DA transitions, indicating a shift from diagnostic to intervention-oriented reasoning. All transition probabilities were calculated at the team level and then aggregated for expert- and trainee-led teams, rather than reported for individual sequences. To determine statistical reliability, we calculated adjusted residuals (z-scores) for each transition, applying a threshold of |z| ≥ 1.96 (p 1.96) or inhibited (< -1.96) transitions. These z-scores indicate whether a given transition occurs more or less often than expected by chance. To assess the stability of reasoning loops (“stuckness”) versus linear progression, we extended the analysis beyond immediate sequential behaviors (Lag-1) to multiple steps (Lags 2–5). This allowed us to examine whether teams remained within the same reasoning mode across several consecutive utterances or transitioned more flexibly between DAs and IAs. Persistence was operationalized as sustained self-transition probabilities and statistically significant z-scores across increasing lags, whereas faster decay of z-scores indicated greater temporal fluidity and reduced persistence. Epistemic Network Analysis (ENA) While LSA models sequential transitions, ENA models the complex co-occurrence of the activities (i.e., codes) within a moving temporal window (set to six utterances) to create weighted network associations. We used ENA to investigate structural distinctiveness across reasoning modes (Proposition A), collaborative activity variation (Proposition C), and expertise-related differences (Propositions D1 and D2). To ensure consistent evaluation, we applied a standardized approach to network comparison and interpretation (Zörgő et al., 2024). We projected networks into a shared geometric space and generated subtraction networks to visualize distinct connection topologies, highlighting connections that are stronger in one context than the other, where edge weights and colors indicate context dominance. We quantitatively compared the positions of network centroids using two-sample t-tests (assuming unequal variance) along the primary dimension (X-axis). Significant centroid separation (p < 0.05) and Cohen's d effect sizes served as evidence of systematic structural differences in communicative organization. We aggregated and compared data in two distinct phases. First, to test the structural dissociation between reasoning modes, we aggregated data across all teams to contrast a pooled Intervention Network (CA co-occurrences within IA-coded segments) against a Diagnostic Network (DA-coded segments). Structural dissociation was considered supported if the network centroids were significantly separated (p < .05) with a meaningful effect size, and not supported if no significant separation or only negligible effects were observed. Second, to evaluate expertise differences while controlling for these inherent structural variations, we separated the dataset into distinct groups based on team expertise. By ensuring that expert and trainee data points were not mixed, we compared their networks independently within diagnostic and intervention contexts. Expertise differences were considered supported if expert and trainee centroids differed significantly within a given context and not supported if no significant differences between groups were observed. 4 Results 4.1 Distribution of Diagnostic and Intervention Activities The analysis of the total activity distribution reveals distinct differences in reasoning focus between teams (see Fig. 3 and Fig. 4 ). Expert-led teams dedicated a significantly larger proportion of their communication to IAs (66.2%) compared to DAs (33.8%). In contrast, Trainees exhibited a more balanced split but a comparatively higher reliance on diagnosis, with 54.2% of activities coded as IAs and 45.8% as DAs. This pattern is consistent with Proposition D2, indicating that expert-led teams engaged in a greater proportion of explicit intervention-oriented activity, whereas trainee-led teams showed a comparatively higher proportion of diagnosis-oriented communication. Table 2 Frequency and percentage of communicative activities (diagnostic and intervention) by scenario stage and team Simulation Stage Across both Types of Teams Expert-led Teams Trainee-led Teams Stage 1 20.93% 13.59% 24.04% Stage 2 15.61% 15.04% 15.84% Stage 3 21.36% 22.48% 20.88% Stage 4 9.67% 10.94% 9.14% Stage 5 20.24% 21.79% 19.58% Stage 6 12.19% 16.15% 10.51% Note. Stages: Stage 1 = ventricular tachycardia with pulse (initial presentation ), Stage 2–3 = deterioration to pulselessness, Stage 4 = asystole, Stage 5 = ventricular fibrillation, Stage 6 = return of spontaneous circulation (ROSC). Examining the volume of activity across the six stages of the simulation highlights differences in team momentum and focus (Table 2 , Fig. 3 ). Trainees spent the highest percentage of their communicative effort in Stage 1 (24.04%). Experts, conversely, peaked in Stage 3 (22.48%), aligning their maximum communicative output with the scenario's active intervention phase. A notable divergence occurred at the Return of Spontaneous Circulation (ROSC; Stage 6). Trainees showed a resurgence of DAs at this stage, while experts maintained a steady focus on IA. Across all stages, experts engaged in a higher percentage of IA-based communication than trainees, who maintained higher DA ratios throughout the scenario. 4.2 Lag Sequence Analysis (LSA) Figure 5 presents the transition dynamics for the entire dataset at Lag-1. The overall pattern reveals a significant tendency for teams to maintain their collective attention on a single reasoning process. The strongly positive z-scores for self-transition (DA→DA and IA→IA) indicate that teams are highly likely to maintain their current state rather than rapidly shifting between modes. This highlights the clustered nature of team interactions and the general stability of reasoning blocks during the simulation. . When stratified by experience level (Fig. 6 ), distinct transition patterns emerged. Experts were significantly more likely to transition from DAs to IAs ( P (DA→IA) = .37) compared to trainees ( P (DA→IA) = .28). This reflects a more direct progression from problem formulation to solution implementation in expert-led teams. Conversely, experts were less likely to switch back from an IA to a DA ( P (IA→DA) = .19) compared to trainees ( P (IA→DA = .23). Analyzing transitions across multiple lags (Lags 1–5) revealed distinct patterns of cognitive persistence versus fluidity (Fig. 7 a and 7 b). Trainees exhibited a higher probability of remaining in a Diagnostic Loop (DA→DA) across all observed lags (Fig. 7 b). For example, at Lag-1, the self-transition probability for trainees was 0.72 compared to 0.63 for experts. As the lag increased, trainees remained more likely to be “stuck” in specific reasoning loops (both DA and IA), indicating that they tend to spend more time iterating within a single reasoning category when communicating. In contrast, the lower self-transition probabilities for experts across Lags 1–5 suggest a more linear progression. Experts transitioned from diagnostic to intervention sequences more quickly, showed less fragmented, cyclical reasoning patterns observed in trainee-led teams, thereby supporting the model's prediction regarding expertise-driven transition fluidity (Proposition D1). Further insight into the persistence of prior states across lags is provided by the decay of adjusted residuals (z-scores). For expert-led teams, the z-scores for self-transitions decayed rapidly, dropping below the significance threshold (|z| < 1.96) starting at Lag-5. This suggests that an expert-led team's current reasoning is relatively independent of actions taken five or more steps earlier, enabling fluid adaptation to new stimuli. In contrast, trainee z-scores remained significantly above the threshold up to Lag-8, indicating a strong, prolonged influence from the past that constrained their ability to switch gears. 4.3 Epistemic Network Analysis (ENA) Beyond temporal sequences, the Epistemic Network Analysis (ENA) models reveal that DAs and IAs form fundamentally different structural networks of communication (Fig. 8 ). The projection of the network centroids confirmed a statistically significant separation between the two reasoning modes along the primary dimension (Mean IA = -0.14, SD IA = 0.12 vs. Mean DA = 0.14, SD DA = 0.16; t(52.47)= -7.13, p < 0.001, Cohen's d = 1.8). This strong effect size indicates that IA and DA represent distinct reasoning modes with systematically different communicative states (Proposition A). Visual inspection of the network nodes (centroids) reveals clear thematic clustering by reasoning goal. In the intervention cluster, the codes “Allocating Tasks” and “CLC: Checkback Statement” are heavily positioned toward the IA side, serving as the primary mechanisms for executing and verifying care. The subtraction network, highlighting differences in connection strength between IA and DA segments, shows strong, cohesive links between these two codes on the IA side, suggesting a tightly coordinated execution pattern. Conversely, the diagnostic cluster on the DA side shows a more distributed pattern of connections among “Questioning / Eliciting Information”, “Sharing Information”, “Sharing Hypothesis”, and “Evaluating Shared Information”, reflecting the integration of multiple information-related activities during diagnostic reasoning, which tend to co-occur in close temporal proximity. We subsequently examined how experts and trainee teams differ in their communicative structure, specifically during DAs (Fig. 9 ). Expert-led teams (Mean = -0.17, SD = 0.16) were statistically distinct from trainee-led teams (Mean = 0.08, SD = 0.14) along the primary dimension (t(13.83)= -4.04, p < 0.001, Cohen's d = 1.71). This large effect size confirms that even when expert and trainee teams are ostensibly “diagnosing”, they are engaging in fundamentally different communicative processes. The positioning of nodes in the projected space illustrates a divergence in functional differentiation. Trainee networks clustered closely with “Questioning/Eliciting Information”, “Expressing Uncertainty”, and “Assigning Roles”, characterizing diagnosis as a homogeneous pattern of information seeking and role clarification. Expert networks, however, clustered with “Sharing Hypothesis”, “Allocating Tasks”, and “CLC: Checkback Statement”. Notably, experts utilized these action-oriented activities even within diagnostic segments, possibly “pre-loading” the intervention phase. The difference network is in support of these strategies: trainees showed stronger unique connections linking “Questioning / Eliciting Information” to “Sharing Information”, whereas experts displayed stronger connections linking “Sharing Hypothesis” to “Allocating Tasks”. Similar structural differences emerged when teams engaged in IAs (Fig. 10 ). Expert-led teams (Mean = -0.13, SD = 0.15) were significantly distinct from trainee-led teams (Mean = 0.06, SD = 0.16; t(15.90)= -3.04, p = 0.01, Cohen's d = 1.21). While the effect size is slightly smaller than in the diagnostic context, it remains robust, indicating that “intervening” means something structurally different to experts than it does to trainees. Trainee networks demonstrated continued use of diagnostic-oriented activities in the intervention phase, maintaining strong associations with “Questioning/Eliciting Information” and “Sharing Information” rather than switching purely to execution. Expert networks were tightly clustered around “Allocating Tasks” and “CLC: Checkback Statement”, prioritizing tightly coordinated execution patterns. The expert difference network was dominated by a strong triangle of connections between “Allocating Tasks”, “CLC: Checkback Statement”, and “Sharing Information”, reflecting a confirm-assign-verify protocol. In contrast, the strongest unique connections for trainees linked “Questioning / Eliciting Information” to “Allocating Tasks” and “Sharing Information”, likely reflecting hesitation and interrupted instructions. Comparing the results across both contexts reveals a fundamental difference in how the teams led by experts and those led by trainees organize their reasoning. Trainees exhibited a relatively homogenous communicative structure across both contexts. Whether diagnosing or intervening, their networks were consistently pulled toward “Questioning/Eliciting Information” and “Sharing Information”. This suggests they tend to lack a distinct “execution mode” and instead rely on a continuous, undifferentiated process of information gathering to manage uncertainty. Experts, however, demonstrated high functional differentiation. In the DA context, their structure bridged hypothesis generation with preparatory assignments, acting as a planning mode. In the IA context, they shifted entirely to a tight loop of assignment and verification, acting primarily in an execution mode. 5 Discussion 5.1 Summary of findings This study provides an initial empirical examination of the CDI-R process model in a high-stakes acute care simulation. All observable team communication was coded as CAs and additionally tagged as DA- or IA-oriented, allowing us to analyze distributional, temporal, and structural characteristics of DAs and IAs and their variation by expertise. At the distributional level, expert-led teams devoted a larger proportion of their communication to IA-oriented segments, whereas trainee-led teams showed a comparatively higher proportion of DA-oriented communication. Experts completed the simulation in less time while showing a higher proportion of IA-oriented activity. Across stages, trainees showed greater DA resurgence in later segments, whereas experts maintained IA dominance. These findings align with Proposition D2 and indicate that explicit engagement in IAs increases with expertise. At the temporal level, LSA revealed significant maintenance probabilities for both DA→DA and IA→IA transitions, indicating recurrent reasoning phases and supporting Proposition B. Expertise shaped transition dynamics (Proposition D1). Expert-led teams showed higher DA→IA transition probabilities and lower IA→DA reversions at Lag-1. Across multiple lags, expert self-transition effects decayed more rapidly, whereas trainee-led teams showed sustained DA→DA and IA→IA persistence across lags, indicating less dynamic progression. At the structural level, pooled ENA models demonstrated clear dissociation between DA- and IA-oriented segments (Proposition A). DA-oriented segments showed stronger co-occurrences among Sharing Information, Questioning / Eliciting Information, Evaluating Shared Information, Sharing Hypothesis, and Expressing Uncertainty. IA-oriented segments were characterized by stronger connections between Allocating Tasks and CLC: Checkback Statement. This phase-dependent variation in CA configuration supports Proposition C. Stratified ENA further revealed expertise-related differences within each reasoning orientation (Proposition D1). In DA-oriented segments, expert-led teams showed stronger links between Sharing Hypothesis and Allocating Tasks, whereas trainee-led teams showed denser connections among Sharing Information, Questioning / Eliciting Information, and Expressing Uncertainty. In IA-oriented segments, expert-led teams demonstrated tightly connected Allocating Tasks and CLC: Checkback Statement configurations, while trainee-led teams maintained stronger co-occurrences among joint information processing activities. Across contexts, expert networks showed greater differentiation between DA- and IA-oriented configurations, whereas trainee networks exhibited more homogeneous structures. Taken together, the validation findings provide converging support for all core propositions of the CDI-R model, while also revealing how expertise modulates their interaction. 5.2 Implications for CDI-R Model The CDI-R model was introduced as a theory-driven process model for collaborative diagnostic and intervention reasoning. The present findings provide initial empirical support for its core assumptions and contribute to refining its theoretical propositions. Building on prior work on the CDR model (Radkowitsch et al., 2022 ; Brandl et al., 2024 ) and extending efforts to integrate contextual and interprofessional factors into healthcare problem-solving frameworks (Witti et al., 2023 ), CDI-R advances a process-oriented perspective on team-based clinical reasoning that integrates diagnostic, intervention, and collaborative dynamics. In the following, we discuss the implications of the findings for each proposition (A–D), while explicitly addressing how the propositions interrelate. Proposition A: Structural Distinction of DAs and IAs Proposition A states that DAs and IAs constitute conceptually and functionally distinct reasoning processes. The ENA findings provide structural support for this assumption. Across pooled analyses, DA-oriented segments were characterized by stronger connections among information sharing, questioning / eliciting information, evaluating shared information, sharing hypotheses, and expressing uncertainty. In contrast, IA-oriented segments showed structurally stronger connections involving allocating tasks and CLC (i.e., checkback statements). These results indicate that diagnostic and intervention reasoning are associated with distinct configurations of CAs. This supports the theoretical extension of the CDR model (Radkowitsch et al., 2022 ), which primarily focused on diagnostic knowledge coordination. By demonstrating that intervention-oriented segments are structurally dominated by coordination and verification processes, the findings support the view that intervention reasoning constitutes a distinct process rather than a direct extension of diagnostic reasoning. At the same time, the distinction observed under Proposition A is closely linked to Proposition C, which posits that CAs vary systematically depending on reasoning orientation. The distinction between DAs and IAs is reflected in systematic differences in the CAs associated with each reasoning mode. In this sense, the empirical support for Proposition A is closely linked to Proposition C, which specifies how these differences in collaborative activities emerge. Proposition B: Non-Linear and Recurrent Interaction of DAs and IAs Proposition B assumes that DAs and IAs unfold in non-linear, recurrent patterns rather than in a strictly staged sequence. The LSA findings confirm this dynamic organization. In line with our analytic definitions, significant maintenance probabilities (DA→DA; IA→IA) indicate sustained reasoning phases, while switch transitions (DA→IA; IA→DA) show that reasoning modes interact and alternate over time. The multi-lag analyses further clarified that recurrence does not imply rigidity. Although both expertise levels exhibited sustained reasoning loops (Proposition B), expert-led teams showed faster decay of maintenance effects across lags, indicating a faster decline in the influence of prior states. This suggests that while recurrent patterns characterize collaborative reasoning in general, their temporal dynamics vary as a function of expertise (Proposition D1). These findings align with theoretical perspectives emphasizing that reasoning in complex environments is recursive and adaptive rather than linear (Hetmanek et al., 2018 ; Radkowitsch et al., 2022 ). Within the CDI-R framework, non-linearity is not noise but a structural feature of collaborative reasoning under dynamic task conditions. Thus, non-linear recurrence is a general feature of collaborative reasoning (Proposition B), but its temporal dynamics vary systematically with expertise (Proposition D1). Proposition C: Phase-Dependent Modulation of Collaborative Activities Proposition C posits that CAs vary in relevance and configuration depending on whether teams are engaged in DAs or IAs. The ENA results provide clear empirical support for this. DA-oriented segments were structurally centered on joint information processing and monitoring activities, including sharing information, questioning / eliciting information, evaluating shared information, sharing hypothesis, and expressing uncertainty. IA-oriented segments were structurally centered on coordination and structured communication, particularly allocating tasks and CLC (i.e., checkback statement). These findings are consistent with theoretical work on team coordination and structured communication in high-risk domains (Salas et al., 2008 ; Liu et al., 2016 ), which highlights the role of verification and role clarity during action implementation. They thus support the extension of collaborative reasoning models toward coordinated action, as proposed in CDI-R. Importantly, expertise again modulated these phase-dependent configurations (Proposition D1). Expert-led teams showed more differentiated CA structures across DA- and IA-oriented segments, whereas trainee-led teams exhibited more homogeneous CA configurations across reasoning modes. This suggests that expertise shapes how clearly collaborative structures align with reasoning orientation. Thus, while CAs vary systematically with reasoning orientation (Proposition C), the degree of differentiation between these configurations depends on expertise (Proposition D1). Propositions D1 and D2: Expertise as Coordination and Explicit Intervention Engagement Proposition D1 states that expertise becomes visible in how teams coordinate and transition between DAs and IAs under time pressure. Proposition D2 states that experienced clinicians may engage in a higher proportion of explicitly enacted IAs. The present findings support both propositions while also refining them. First, expert-led teams engaged in a higher proportion of IA-oriented segments while completing the simulation in less time. This finding is theoretically significant. Classical expertise theory would predict fewer observable steps to reach the same solution due to pattern recognition and knowledge encapsulation (Boshuizen & Schmidt, 1992 ). In the context of intervention reasoning, however, this assumption appears insufficient. The present findings suggest that expertise may involve more, not fewer, explicitly enacted intervention actions when these are strategically timed and tightly coordinated. This aligns with evidence that intervention-focused reasoning rather increases than decreases with clinical experience (Monajemi et al., 2012 ) and that experienced clinicians invest substantial effort in dynamic interventions (Hartjes et al., 2024 ). These results therefore challenge the assumption, derived from classical expertise theory, that expert performance is characterized by fewer observable steps, and refine Proposition D2. Expertise in intervention-rich environments may manifest not as reduced observable activity, but as structured, anticipatory, and verification-driven coordination of action. This apparent deviation from minimalist expertise accounts invites further interpretation. One explanation for the increased IA frequency is that certain intervention steps serve deliberate safety and verification functions. Experts may enact additional IAs as structured double checks (e.g., confirming device functioning or task execution), reflecting monitoring and error-prevention strategies rather than inefficiency. Second, expertise shaped the temporal and structural organization of reasoning. Expert-led teams showed higher DA→IA transition probabilities, reduced persistence within single reasoning modes, and more differentiated CA configurations across reasoning orientations. These findings directly support Proposition D1 while also reinforcing Propositions B and C. A similar expertise-related pattern was visible in how teams responded to the onset of the ROSC phase of the simulation. Trainee-led teams often appeared to treat ROSC as signaling completion of the scenario and shifted toward diagnostic discussion and reflection on the patient’s condition. In contrast, expert-led teams tended to interpret ROSC as a trigger for the next intervention phase, immediately transitioning toward post-resuscitation care. This pattern further illustrates how expertise shapes the timing and coordination of intervention-oriented reasoning under dynamic task conditions. Beyond these differences in frequency (Proposition D2), the structural findings suggest a pattern of deeper integration. In expert-led teams, links between Sharing Hypothesis and Allocating Tasks in DA-oriented segments indicate that diagnostic formulation is already aligned with anticipated intervention. This suggests that intervention reasoning may function, in part, as a diagnostic probe, for example by systematically disconfirming alternative causal explanations (e.g., H’s and T’s reversible causes of cardiac arrest; Reddy & Hanmandlu, 2022 ), consistent with iterative models of generation, selection, and evaluation (Richters et al., 2024 ). Expertise may therefore involve a closer alignment between diagnostic formulation (DAs) and subsequent intervention-oriented actions (IAs). This pattern suggests that, with increasing expertise, the functional boundary between DAs and IAs becomes more permeable. This pattern is consistent with the CDI-R model, in which Propositions A–C describe structural properties of collaborative reasoning, whereas Proposition D specifies how these structures are more coherently integrated and adaptively coordinated with increasing expertise. Simulation Context and Model Generalizability Although the present empirical test was conducted in a VR-based cardiac arrest simulation, the CDI-R model itself is not restricted to this specific context. Rather, it conceptualizes how collaborative diagnostic and intervention reasoning unfold in dynamic, time-pressured clinical environments. The present study therefore provides an empirical test case within acute care simulation rather than delimiting the model to that context. Future research in live clinical settings and across different domains of care will be necessary to examine the generalizability of these dynamics. Building on prior applications of the CDR model in clinical contexts (Brandl et al., 2024 ) and efforts to incorporate contextual and interprofessional factors into healthcare problem-solving frameworks (Witti et al., 2023 ), CDI-R is positioned as a flexible process model adaptable across clinical environments. Interrelation of Propositions and Conceptual Refinement Although Propositions A–D were introduced as analytically distinct components of the model, their interrelations follow directly from the theoretical structure of CDI-R. Specifically, Propositions A–C describe structural and temporal properties of collaborative reasoning, whereas Proposition D specifies how these processes are shaped by expertise. The empirical support for the individual propositions therefore also supports the coherence of the overall model (see Table 3 ). Table 3 Empirical support for CDI-R propositions Proposition Theoretical claim Empirical support (this study) A DAs and IAs are structurally distinct reasoning processes ENA showed significant centroid separation and distinct CAs configurations B DAs and IAs unfold in non-linear, recurrent patterns LSA revealed significant self-transitions and switching patterns C CAs vary depending on reasoning mode ENA showed systematic differences in CAs configurations between DAs and IAs segments D1 Expertise shapes coordination and transition dynamics Experts showed higher DA→IA transitions, lower stuckness, more differentiated structures D2 Expertise linked to greater explicit IA frequency Experts showed higher proportion of IA-oriented activity Beyond supporting the individual propositions, the findings also point to a broader conceptual implication regarding the nature of collaborative reasoning. In team-based clinical settings, reasoning is not solely an individual cognitive process but is enacted through interaction. DAs and IAs involve both individual cognitive processing and collaboratively organized communicative activities. While the analytic separation of propositions supports theoretical clarity, the empirical patterns observed here underscore their tight coupling in practice. This should not be interpreted as a limitation of measurement but rather reflects the inherently interactive nature of collaborative reasoning. Within the CDI-R framework, collaboration is therefore not an external layer added to reasoning but constitutes the medium through which diagnostic and intervention processes are coordinated and realized. The present findings therefore support CDI-R as an integrated collaborative process model rather than as a set of isolated claims. They strengthen the theoretical necessity of intervention reasoning as a distinct construct within clinical reasoning research and contribute to refining how expert performance is conceptualized in dynamic, intervention-rich environments. Intervention reasoning emerges not as a residual category beyond diagnosis, but as a coordinated, action-oriented reasoning mode that dynamically interacts with diagnostic processes and is systematically shaped by expertise. Future research across diverse clinical contexts will be required to further test and elaborate this integrated framework. 5.4 Limitations and further research While the present findings provide initial empirical support for the CDI-R model, several limitations qualify their interpretation. First, the VR environment itself may introduce confounds related to navigating a novel interface. Prior research suggests that familiarity with VR reduces cognitive load during complex tasks (Lee et al., 2023 ; Bruyne et al., 2026 ). We did not assess participants’ previous VR experience. This could have influenced observed differences in reasoning patterns. Also, the VR headsets limited access to facial cues and reduced co-presence compared to bedside practice. While the simulation was designed to authentically represent acute care interactions, the technologically mediated context may remain a constraint on validity. Second, this initial validation of the CDI-R model focused on expert- versus trainee-led teams. However, we did not systematically vary team composition or role distribution. A more controlled design would need to systematically vary group configurations–e.g., mixed (dyadic, larger team) expertise levels, rotating leadership, prior team familiarity, interprofessional team composition–to isolate their independent and interactive effects on collaborative reasoning patterns. In addition, in five out of nine expert-led conditions, non-leader roles were filled by research staff knowledgeable with ACLS procedures and familiar with the VR environment and performing chest compressions and/or airway management, whereas trainee-led teams consisted exclusively of residents. This structural difference in team composition may have influenced collaborative dynamics independently of leadership expertise. To mitigate potential bias, research staff participating in supporting roles were instructed to perform their tasks in a standardized manner, avoiding behaviors that could unduly advantage or disadvantage any particular study condition. Although the leadership role was the primary experimental manipulation, future research should employ fully comparable team compositions to isolate the effects of expertise more precisely. Third, although the CDI-R model proposes general principles for collaborative diagnostic and intervention reasoning, we tested it within a single acute care context. Cardiac arrest has unique task characteristics, such as highly protocolized interventions (ACLS guidelines), compressed time frames, and relatively clear diagnostic decision points, that may not reflect the full range of acute care scenarios. Other contexts, such as multi-trauma or sepsis, involve greater diagnostic ambiguity, longer timeframes, and less standardized intervention pathways, which may substantially alter the relative distribution and sequencing of DAs and IAs. Beyond acute care, it would further be valuable to investigate whether similar collaborative reasoning dynamics emerge in other complex team-based decision-making domains, thereby testing the broader domain-generality of the CDI-R framework. Fourth, we analyzed teams’ DAs, IAs, and CAs patterns but did not directly link these patterns to patient outcomes or performance quality. The simulation concluded at ROSC for all teams, but we did not measure time-to-ROSC, adherence to ACLS protocols, or omission and commission errors. Future research should examine whether the patterns proposed in the CDI-R model predict measurable differences in care quality, patient safety, or treatment efficiency. Fifth, our analysis segments communication into discrete utterance-level codes, but clinical reasoning may involve cognitive processes that unfold between utterances or remain covert/internalized. Silent coordination, parallel processing by team members, or internal decision-making before verbalization are not captured. Experts may engage in more efficient covert reasoning that reduces their need for certain types of verbal collaboration. Multimodal methods such as eye-tracking, physiological measures, freeze probe recall, or post-simulation subjective rating approaches could provide insight into these latent cognitive processes (Salmon et al., 2009 ). Sixth, it remains unclear to what extent the patterns observed in VR-based cardiac arrest management transfer to real clinical settings, which involve higher stakes, greater stress, and additional environmental constraints. Simulation-induced inquiry may partly inflate the observable frequency of IAs while also enabling controlled investigation of intervention reasoning. Future research should disentangle simulation-specific exploration from clinically driven intervention behavior. 5.5 Practical Implications The present findings have implications for how intervention processes are conceptually framed within medical education. In many acute care curricula, intervention procedures, task allocation, and structured communication are already central training components (e.g., ACLS and crisis resource management). However, these elements are often approached from a procedural or protocol-driven perspective. The present results suggest that coordinated intervention under time pressure can also be understood as a structured form of collaborative reasoning that unfolds in dynamic interplay with diagnostic reasoning, rather than as a discrete downstream step(s). Our findings indicate that effective IA-oriented phases involve systematic task allocation, closed-loop communication, and deliberate monitoring, and that these processes are dynamically coupled with diagnostic formulation. Instructional approaches may therefore benefit from explicitly framing team-based intervention not only as execution of protocols but as a coordinated reasoning process that integrates hypothesis generation, action planning, and real-time evaluation. In simulation-based training, debriefings could emphasize more process analysis in addition to outcome evaluation and guideline adherence. By mapping the co-occurrence and sequencing transitions between DA- and IA-oriented phases and how CAs are configured within each, the ENA visualizations reveal precisely where and how reasoning transitions falter across expertise levels. These network structures can inform targeted feedback during post-simulation debriefs. For example, analyses may pinpoint when trainee-led teams remain anchored in diagnostic loops rather than transitioning to intervention in a timely manner. Such process-oriented analysis may also help identify silent failures , situations in which teams reach apparently successful outcomes (e.g., ROSC) while still demonstrating fragile reasoning patterns, such as missed intervention steps, errors of omission, or limited engagement in systematic cause evaluation (e.g., H’s and T’s). By revealing these patterns, ENA and LSA visualizations can support more targeted feedback during debriefings. However, realizing this potential requires adequate faculty development to equip instructors with the skills to interpret process-oriented analytics and translate them into effective feedback, particularly as prior work highlights the importance of aligning learning analytics tools with users’ needs and supporting stakeholders in understanding and using such data in practice (Alfredo et al., 2025 ). Using the CDI-R framework along with ENA visualizations as a reflective lens may help make these collaborative reasoning structures visible and discussable. Beyond individual debriefs, the CDI-R framework may enable systematic comparison of collaborative clinical reasoning patterns across teams and scenarios. If applied longitudinally, it could allow educators to trace how collaborative clinical reasoning develops over the course of training. Together, these implications do not argue for more intervention training per se, but for a more explicit conceptualization of intervention reasoning as a collaborative, action-oriented reasoning mode within existing team-based training programs. 6 Conclusion Clinical reasoning research has predominantly emphasized diagnostic reasoning, while intervention reasoning has received comparatively limited conceptual and empirical attention. Yet, in complex, high-risk team-based situations, such as cardiac arrest, reasoning cannot be reduced to uncertainty reduction alone. It unfolds as a dynamic interplay between DAs, IAs and CAs under time pressure. This paper responds to the conceptual and empirical gap in research on intervention reasoning by introducing CDI-R as a theory-driven process model that systematically integrates DAs, IAs, and CAs within a unified framework. Using VR-based acute care simulations as an initial empirical instantiation, we demonstrate that intervention reasoning constitutes a structurally distinct yet dynamically intertwined mode of collaborative reasoning. Moreover, expertise shapes not only how often interventions are enacted but also how diagnostic and intervention processes are temporally orchestrated and structurally integrated. By turning clinical reasoning research toward intervention, this paper expands the theoretical scope of collaborative reasoning models beyond diagnosis-centered accounts. CDI-R positions intervention reasoning as a coordinated, action-oriented component of team-based clinical reasoning that interacts dynamically with diagnostic processes. While further research is required to systematically test the model across diverse clinical reasoning contexts and link reasoning process patterns to performance outcomes, the present findings provide a conceptual and analytic foundation for studying how teams turn evolving situation assessments into coordinated action. Declarations Funding This work was funded by the U.S. National Science Foundation under Grant Nos. 2202451 and 2506865. In addition, the work of Constanze Richters was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under TRR 419/1 – 542251251 and FOR 2385; FI 792/11-2. Competing interests The authors have no competing interests to declare. Ethics approval The study protocol was approved by the Institutional Review Board (HUM00188482, HUM00193383). Consent to participate Informed consent was obtained from all individual participants included in the study. Data availability Due to the privacy of the video and sensitive nature of the simulation training, trainees were assured raw data would remain confidential and would not be shared. Acknowledgments We would like to thank Dr. Michael Cole and Dr. James Cooke for sharing their expertise on ACLS, VR, and clinical simulation. We also thank Benjamin Root, Cami Trendy, Umair Syed, Morgan Carpenter, and Nikolas Grotewold for their data annotation efforts. Author contributions: Constanze Richters: Conceptualization; Methodology; Visualization; Validation; Project administration; Writing – original draft; Writing – review & editing. Kapotaksha Das: Methodology; Software; Formal analysis; Visualization; Data curation; Writing – original draft; Writing – review & editing. Frank Fischer: Funding acquisition; Conceptualization; Supervision; Writing – review & editing. Martin R. Fischer: Writing – review & editing. Matthias Stadler: Funding acquisition; Writing – review & editing. Vitaliy Popov: Conceptualization; Methodology; Formal analysis; Investigation; Data curation; Resources; Supervision; Funding acquisition; Writing – original draft; Writing – review & editing. References Abdoler, E. A., Parsons, A. S., & Wijesekera, T. P. (2023). The future of teaching management reasoning: Important questions and potential solutions. Diagnosis , 10 (1), 19–23. https://doi.org/10.1515/dx-2022-0048 Alfredo, R., Milesi, M., Echeverria, V., Gašević, D., Buckingham Shum, S., Zhao, L., Yan, L., Jin, Y., Fan, J. X., Pammer-Schindler, V., Swiecki, Z., & Martinez-Maldonado, R. (2025). Co-designing AI-powered learning analytics: Bringing students and teachers together. International Journal of Educational Technology in Higher Education , 22 (1), 78. https://doi.org/10.1186/s41239-025-00572-8 Bakeman, R., & Quera, V. (2011). Sequential analysis and observational methods for the behavioral sciences . Cambridge University Press. Brandl, L., Stadler, M., Richters, C., Radkowitsch, A., Fischer, M. R., Schmidmaier, R., & Fischer, F. (2024). Collaborative Problem-Solving in Knowledge-Rich Domains: A Multi-Study Structural Equation Model. International Journal of Computer Supported Collaborative Learning 19, 341–368 (2024). https://doi.org/10.1007/s11412-024-09425-4 Boshuizen, H. P. A., & Schmidt, H. G. (1992). On the Role of Biomedical Knowledge in Clinical Reasoning by Experts, Intermediates and Novices. Cognitive Science , 16 (2), 153–184. https://doi.org/10.1207/s15516709cog1602_1 Charlin, B., Boshuizen, H. P. A., Custers, E. J., & Feltovich, P. J. (2007). Scripts and clinical reasoning: Clinical expertise. Medical Education , 41 (12), 1178–1184. https://doi.org/10.1111/j.1365-2923.2007.02924.x Cook, D. A., Durning, S. J., Sherbino, J., & Gruppen, L. D. (2019). Management Reasoning: Implications for Health Professions Educators and a Research Agenda. Academic Medicine , 94 (9), 1310–1316. https://doi.org/10.1097/ACM.0000000000002768 De Bruyne, J., Larmuseau, C., De Marez, L., Talsma, D., & Bombeke, K. (2026). Timing matters! Using delayed signaling to improve experiential learning in procedural VR training. Computers & Education , 105553. https://doi.org/10.1016/j.compedu.2025.105553 Department of Defense, & Agency for Healthcare Research and Quality. (2010). TeamSTEPPS®: Strategies and tools to enhance performance and patient safety . [Manual]. Department of Defense, TRICARE; Agency for Healthcare Research and Quality. Duong, Q. H., Pham, T. N., Reynolds, L., Yeap, Y., Walker, S., & Lyons, K. (2023). A scoping review of therapeutic reasoning process research. Advances in Health Sciences Education , 28 (4), 1289–1310. https://doi.org/10.1007/s10459-022-10187-7 Gruppen, L. (2017). Clinical Reasoning: De ning It, Teaching It, Assessing It, Studying It. Western Journal of Emergency Medicine , 18 (1), 4–7. https://doi.org/10.5811/westjem.2016.11.33191 Ericsson, K. A. (Ed.). (1991). Toward a general theory of expertise: Prospects and limits . Cambridge University Press. Fink, M. C., Reitmeier, V., Stadler, M., Siebeck, M., Fischer, F., & Fischer, M. R. (2021). Assessment of Diagnostic Competences With Standardized Patients Versus Virtual Patients: Experimental Study in the Context of History Taking. Journal of Medical Internet Research , 23 (3), e21196. https://doi.org/10.2196/21196 Fischer, F., Kollar, I., Stegmann, K., & Wecker, C. (2013). Toward a script theory of guidance in computer-supported collaborative learning. Educational Psychologist , 48 (1), 56–66. https://doi.org/10.1080/00461520.2012.748005 Fischer, F., Kollar, I., Ufer, S., Sodian, B., Hussmann, H., Pekrun, R., Neuhaus, B., Dorner, B., Pankofer, S., Fischer, M. R., Strijbos, J. W., Heene, M., & Eberle, J. (2014). Scientific Reasoning and Argumentation: Advancing an Interdisciplinary Research Agenda in Education. Frontline Learning Research , 2 (3), 28–45. https://doi.org/10.14786/flr.v2i2.96 Hartjes, M. G., Richir, M. C., Cazaubon, Y., Donker, E. M., Van Leeuwen, E., Likic, R., Pers, Y. M., Piët, J. D., De Ponti, F., Raasch, W., Van Rosse, F., Rychlícková, J., Sanz, E. J., Schwaninger, M., Wallerstedt, S. M., De Vries, T. P. G. M., Van Agtmael, M. A., Tichelaar, J., & On behalf of the EACPT Education Working group. (2024). Enhancing therapeutic reasoning: Key insights and recommendations for education in prescribing. BMC Medical Education , 24 (1), 1360. https://doi.org/10.1186/s12909-024-06310-4 Heitzmann, N., Seidel, T., Opitz, A., Hetmanek, A., Wecker, C., Fischer, M. R., Ufer, S., Schmidmaier, R., Neuhaus, B., Siebeck, M., Stürmer, K., Obersteiner, A., Reiss, K., Girwidz, R., & Fischer, F. (2019). Facilitating Diagnostic Competences in Simulations in Higher Education: A Framework and a Research Agenda. Frontline Learning Research , 7 (4), 1–24. https://doi.org/10.14786/flr.v7i4.384 Hetmanek, A., Engelmann, K., Opitz, A., & Fischer, F. (2018). Beyond Intelligence and Domain Knowledge. In F. Fischer, C. A. Chinn, K. Engelmann, & J. Osborne (Eds.), Scientific Reasoning and Argumentation (1st ed., pp. 203–226). Routledge. https://doi.org/10.4324/9780203731826-12 Janssen, J., & Kirschner, P. A. (2020). Applying collaborative cognitive load theory to computer-supported collaborative learning: Towards a research agenda. Educational Technology Research and Development , 68 (2), 783–805. https://doi.org/10.1007/s11423-019-09729-5 Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality. American Psychologist , 58 (9), 697–720. https://doi.org/10.1037/0003-066X.58.9.697 Kentros, M., Kamarianakis, M., Cole, M., Popov, V., Protopsaltis, A., & Papagiannakis, G. (2025). Non-linear, Team-based VR Training for Cardiac Arrest Care with enhanced CRM Toolkit. Eurographics 2025 - Short Papers . https://doi.org/10.2312/EGS.20251054 Kumar, D., & Basheer, A. (2021). Does management reasoning constitute the backbone of the clinical learning environment? Conceptual analysis of the existing notions. Journal of Advances in Medical Education & Professionalism , 9 (1), 54–58. https://doi.org/10.30476/jamp.2020.84431.1138 Lee, J. Y., de Jong, N., Donkers, J., Jarodzka, H., & van Merriënboer, J. J. (2023). Measuring cognitive load in virtual reality training via pupillometry. IEEE Transactions on Learning Technologies , 17 , 704–710. https://doi.org/10.1109/TLT.2023.3326473 Liaw, S. Y., Chen, F. G., Klainin, P., Brammer, J., O’Brien, A., & Samarasekera, D. D. (2010). Developing clinical competency in crisis event management: An integrated simulation problem-based learning activity. Advances in Health Sciences Education , 15 (3), 403–413. https://doi.org/10.1007/s10459-009-9208-9 Liu, L., Hao, J., von Davier, A. A., Kyllonen, P., & Zapata-Rivera, D. (2016). A tough nut to crack: Measuring collaborative problem solving. In Y. Rosen, S. Ferrara, & M. Mosharraf (Eds.), Handbook of Research on Technology Tools for Real-World Skill Development (pp. 344–359). IGI Global. https://doi.org/10.4018/978-1-4666-9441-5 Meier, A., Spada, H., & Rummel, N. (2007). A rating scheme for assessing the quality of computer-supported collaboration processes. International Journal of Computer-Supported Collaborative Learning , 2 (1), 63–86. https://doi.org/10.1007/s11412-006-9005-x Monajemi, A., & Rikers, R. M. J. P. (2011). The role of patient management in medical expertise development: Extending the contemporary theory. The International Journal of Person Centered Medicine , 1 (1), 161–166. https://doi.org/10.5750/ijpcm.v1i1.42 Monajemi, A., Rostami, E., Savaj, S., & Rikers, R. (2012). How does Patient Management Knowledge Integrate into an Illness Script? Education for Health , 25 (3), 153. https://doi.org/10.4103/1357-6283.109791 Norman, G., Pelaccia, T., Wyer, P., & Sherbino, J. (2024). Dual process models of clinical reasoning: The central role of knowledge in diagnostic expertise. Journal of Evaluation in Clinical Practice , 30 (5), 788–796. https://doi.org/10.1111/jep.13998 Penner, J. C., Minter, D. J., Abdoler, E. A., & Parsons, A. S. (2025). Reasoning on Rounds 2: A Framework for Teaching Management Reasoning in the Inpatient Setting. Journal of General Internal Medicine , 40 (6), 1424–1429. https://doi.org/10.1007/s11606-024-09039-4 Popov, V., Harmer, B., Raphael, S., Scott, I., Sample, A. P., Cooke, J. M., & Cole, M. (2025). Elucidating cognitive processes in cardiac arrest team leaders: A virtual reality-based cued-recall study of experts and novices. Annals of Medicine , 57 (1), 2470976. https://doi.org/10.1080/07853890.2025.2470976 Popov, V., Li, Y., Cooke, J., Sample, A., & Cole, M. (2023). Towards Supporting Technical and Non-Technical Skills Development by Using Multimodal Debriefing System After Multi-User VR-Based Simulation Training . 368–369. https://doi.org/10.22318/cscl2023.644173 Radkowitsch, A., Sailer, M., Fischer, M. R., Schmidmaier, R., & Fischer, F. (2022). Diagnosing Collaboratively: A Theoretical Model and a Simulation-Based Learning Environment. In F. Fischer & A. Opitz (Eds.), Learning to Diagnose with Simulations (pp. 123–141). Springer International Publishing. https://doi.org/10.1007/978-3-030-89147-3_10 Reddy, D. R. S., & Hanmandlu, A. (2022). The 4 H’s and T’s: How reliable is this mnemonic in classifying etiologies of in-hospital cardiac arrests? Resuscitation , 175 , 3–5. https://doi.org/10.1016/j.resuscitation.2022.03.035 Richters, C., Schmidmaier, R., Popov, V., Schredelseker, J., Fischer, F., & Fischer, M. R. (2024). Intervention skills – a neglected field of research in medical education and beyond. GMS Journal for Medical Education , 41 (4). https://doi.org/10.3205/ZMA001703 Richters, C., Stadler, M., Radkowitsch, A., Schmidmaier, R., Fischer, M. R., & Fischer, F. (2023). Who is on the right track? Behavior-based prediction of diagnostic success in a collaborative diagnostic reasoning simulation. Large-Scale Assessments in Education , 11 (1), 3. https://doi.org/10.1186/s40536-023-00151-1 Rosen, M. A., DiazGranados, D., Dietz, A. S., Benishek, L. E., Thompson, D., Pronovost, P. J., & Weaver, S. J. (2018). Teamwork in healthcare: Key discoveries enabling safer, high-quality care. American Psychologist , 73 (4), 433–450. https://doi.org/10.1037/amp0000298 Sackett, G. P. (1979). The lag sequential analysis of contingency and cyclicity in behavioral interaction research. Handbook of infant development , 1 , 623–649. Salas, E., Sims, D. E., & Burke, C. S. (2005). Is there a Big Five in Teamwork? Small Group Research , 36 (5), 555–599. https://doi.org/10.1177/1046496405277134 Salas, E., Wilson, K. A., Murphy, C. E., King, H., & Salisbury, M. (2008). Communicating, Coordinating, and Cooperating When Lives Depend on It: Tips for Teamwork. The Joint Commission Journal on Quality and Patient Safety , 34 (6), 333–341. https://doi.org/10.1016/S1553-7250(08)34042-2 Salmon, P. M., Stanton, N. A., Walker, G. H., Jenkins, D., Ladva, D., Rafferty, L., & Young, M. (2009). Measuring Situation Awareness in complex systems: Comparison of measures study. International journal of industrial ergonomics , 39 (3), 490–500. https://doi.org/10.1016/j.ergon.2008.10.010 Shaffer, D. W., Collier, W., & Ruis, A. R. (2016). A Tutorial on Epistemic Network Analysis: Analyzing the Structure of Connections in Cognitive, Social, and Interaction Data. Journal of Learning Analytics , 3 (3), 9–45. https://doi.org/10.18608/jla.2016.33.3 Simon, H. A., & Newell, A. (1971). Human problem solving: The state of the theory in 1970. American Psychologist , 26 (2), 145–159. https://doi.org/10.1037/h0030806 Stadler, M., Niepel, C., & Greiff, S. (2019). Differentiating between static and complex problems: A theoretical framework and its empirical validation. Intelligence , 72 , 1–12. https://doi.org/10.1016/j.intell.2018.11.003 Tay, S. W., Ryan, P. M., & Ryan, C. A. (2016). Systems 1 and 2 thinking processes and cognitive reflection testing in medical students. Canadian Medical Education Journal , 7 (2), e97–103. https://doi.org/10.36834/cmej.36777 Wilcox, G., & Heudes, A. (2017). Clinical Reasoning in the Assessment and Planning for Intervention for Oppositional Defiant Disorder. Canadian Journal of School Psychology , 32 (1), 46–58. https://doi.org/10.1177/0829573516658996 Wittenburg, P., Brugman, H., Russel, A., Klassmann, A., & Sloetjes, H. (2006). ELAN: A professional framework for multimodality research. In 5th international conference on language resources and evaluation (LREC 2006) (pp. 1556–1559). Witti, M. J., Zottmann, J. M., Wershofen, B., Thistlethwaite, J. E., Fischer, F., & Fischer, M. R. (2023). FINCA – a conceptual framework to improve interprofessional collaboration in health education and care. Frontiers in Medicine , 10 , 1213300. https://doi.org/10.3389/fmed.2023.1213300 Zörgő, S., Árva, D., & Eagan, B. (2024, November). Making sense of the model: Interpreting Epistemic networks and their projection space. In International Conference on Quantitative Ethnography (pp. 119–135). Cham: Springer Nature Switzerland. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 10 May, 2026 Reviewers agreed at journal 03 May, 2026 Reviewers invited by journal 01 May, 2026 Editor assigned by journal 24 Apr, 2026 Submission checks completed at journal 24 Apr, 2026 First submitted to journal 22 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9497414","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":633658263,"identity":"b2daceec-5042-492d-9c0f-786ff7658b2b","order_by":0,"name":"Constanze Richters","email":"data:image/png;base64,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","orcid":"","institution":"LMU Munich, LMU University Hospital","correspondingAuthor":true,"prefix":"","firstName":"Constanze","middleName":"","lastName":"Richters","suffix":""},{"id":633658264,"identity":"bddd00b1-54a4-4fe0-a8e1-ea6752b53bbf","order_by":1,"name":"Kapotaksha Das","email":"","orcid":"","institution":"University of Michigan Medical School","correspondingAuthor":false,"prefix":"","firstName":"Kapotaksha","middleName":"","lastName":"Das","suffix":""},{"id":633658265,"identity":"eebcabe4-4b15-4fe3-955d-7fd5fe7f0aa9","order_by":2,"name":"Frank Fischer","email":"","orcid":"","institution":"LMU Munich","correspondingAuthor":false,"prefix":"","firstName":"Frank","middleName":"","lastName":"Fischer","suffix":""},{"id":633658266,"identity":"fe64c97d-79fe-47a2-9857-e47c85c34490","order_by":3,"name":"Martin R. Fischer","email":"","orcid":"","institution":"LMU Munich, LMU University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Martin","middleName":"R.","lastName":"Fischer","suffix":""},{"id":633658267,"identity":"9c755289-3237-4741-b8f4-b9bd0b56c521","order_by":4,"name":"Matthias Stadler","email":"","orcid":"","institution":"LMU Munich, LMU University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Matthias","middleName":"","lastName":"Stadler","suffix":""},{"id":633658268,"identity":"943d8471-aecc-4e40-9946-76fe4704b98c","order_by":5,"name":"Vitaliy Popov","email":"","orcid":"","institution":"University of Michigan Medical School","correspondingAuthor":false,"prefix":"","firstName":"Vitaliy","middleName":"","lastName":"Popov","suffix":""}],"badges":[],"createdAt":"2026-04-22 14:08:32","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9497414/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9497414/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109075259,"identity":"0db00e7c-f1be-461d-baf2-c2c1bdb7ac89","added_by":"auto","created_at":"2026-05-12 10:54:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":98928,"visible":true,"origin":"","legend":"\u003cp\u003eCDI-R process model for collaborative diagnostic and intervention reasoning\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/628cacf6cee212eaca82b6f0.png"},{"id":109075298,"identity":"a8dd887c-8a03-4194-bdd4-2294a80fe117","added_by":"auto","created_at":"2026-05-12 10:55:06","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":36902,"visible":true,"origin":"","legend":"\u003cp\u003eMulti-user virtual reality (VR)-based cardiac arrest simulation\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/9f2f29649ba38719c236c834.jpg"},{"id":109075494,"identity":"ff3fe0b5-8ff0-4d26-865a-5aa73f72ccdd","added_by":"auto","created_at":"2026-05-12 10:56:34","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":34837,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of total communicative activities (diagnostic vs. intervention) by team type\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/cd7bce1132563a9fb034debc.png"},{"id":109075270,"identity":"a9bdfe41-d102-412e-b7dd-14b5d0c76a24","added_by":"auto","created_at":"2026-05-12 10:54:59","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":25271,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal distribution of diagnostic and intervention activities across simulation stages\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/69f4de83467192e7f33d80eb.png"},{"id":109203862,"identity":"ae4a437a-6ef3-465d-a9e8-b62574716fda","added_by":"auto","created_at":"2026-05-13 14:49:25","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":42444,"visible":true,"origin":"","legend":"\u003cp\u003eWhole dataset – transition frequencies, conditional probabilities, and z-scores at lag-1\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/e2510fd4d4ec4432574cd7a6.png"},{"id":109075498,"identity":"12ff770d-c141-4489-9231-bef8783fee82","added_by":"auto","created_at":"2026-05-12 10:56:35","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":93779,"visible":true,"origin":"","legend":"\u003cp\u003eGroup comparison – transition frequencies and conditional probabilities for expert- vs. trainee-led teams at lag-1\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/205179c2469f1a2e83144946.png"},{"id":109075509,"identity":"412d6fe1-8d31-4758-a33e-46bab3d4d646","added_by":"auto","created_at":"2026-05-12 10:56:41","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":249939,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e7a \u003c/strong\u003eExpert-led teams – characteristic transitions across multiple lags (lags 1–5)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;7b \u003c/strong\u003eTrainee-led teams – characteristic transitions across multiple lags (lags 1–5)\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/d7b64649b5914ee8ee22e4ae.png"},{"id":109076273,"identity":"08d0d6d4-3929-41cd-bf26-738ccf63fcf6","added_by":"auto","created_at":"2026-05-12 11:01:37","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":159132,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003ePooled ENA models comparing intervention (red) and diagnostic (blue) networks\u003c/em\u003e\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/a291e02b8925be598aa2ab30.png"},{"id":109075485,"identity":"0ed0be52-e72a-4448-a402-804ff326e7d6","added_by":"auto","created_at":"2026-05-12 10:56:32","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":160722,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of expert (red) and trainee (blue) networks during diagnostic activities\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/a210ac1b0e5830e82e8498ed.png"},{"id":109075486,"identity":"cd17a2c5-48c6-4c54-ab75-80d30187df45","added_by":"auto","created_at":"2026-05-12 10:56:32","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":126009,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of expert (red) and trainee (blue) networks during intervention activities\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/95321cce04f05363ce8e19c1.png"},{"id":109296503,"identity":"4edeb5d3-e6a4-4c15-8a50-9ee8418f46c8","added_by":"auto","created_at":"2026-05-15 08:47:36","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1277034,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497414/v1/c02127d7-bdef-454c-a238-37a5288d615e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Turning clinical reasoning research toward intervention: Validating a process model using VR simulation data","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eMedical routine cases, such as a finger laceration or an epileptic seizure in a patient with known and pre-diagnosed epilepsy, can be managed by a single clinician. By contrast, high-risk situations such as cardiac arrest or undifferentiated symptom(s) (e.g., headache, abdominal pain, or undifferentiated fever) often require team-based collaboration of multiple clinicians. Clinical reasoning in such teams can be conceptualized as involving two interdependent processes: \u003cem\u003ediagnostic reasoning\u003c/em\u003e (Fischer et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), which aims at reducing uncertainty regarding the etiology of symptoms by gathering and integrating patient information, and \u003cem\u003eintervention reasoning\u003c/em\u003e (Richters et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), which focuses on generating different options, selecting and executing treatments to address the diagnosed problem. These reasoning processes may unfold concurrently, for instance, at the patient\u0026rsquo;s bedside, or sequentially, as in referrals between clinicians.\u003c/p\u003e \u003cp\u003eIn acute care, such as cardiac arrest management, diagnostic and intervention reasoning unfold concurrently under time pressure, as clinicians work together at the patient\u0026rsquo;s bedside. These situations are highly dynamic as conditions may change quickly over short periods of time and require constant diagnostic and intervention adjustments (see Stadler et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Care team leaders and members navigate these time-sensitive activities together through collaborative activities, such as joint information processing, coordination, and monitoring, as well as structured communication techniques, such as closed-loop communication (Liu et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Meier et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Salas et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Witti et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRecent research has begun to examine how collaborative processes support diagnostic reasoning (Brandl et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Radkowitsch et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Richters et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, intervention reasoning has received much less attention (Cook et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Monajemi \u0026amp; Rikers, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Richters et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). A recent scoping review further shows that intervention reasoning research remains poorly integrated, with studies rarely building on each other and lacking a coherent, connected body of literature (Duong et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). It remains unclear how teams coordinate knowledge and actions to reach and implement decisions on interventions in dynamic clinical settings - and where and how they struggle. This is surprising given the central role of intervention reasoning for patient outcomes (Abdoler et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Cook et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Kumar \u0026amp; Basheer, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Monajemi \u0026amp; Rikers, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Penner et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo address the conceptual and empirical gap in research on intervention reasoning in clinical reasoning research, this study proposes and validates the CDI-R process model that integrates diagnostic reasoning, intervention reasoning, and collaboration. The model aims to capture how these reasoning processes unfold in real-time team interactions. We validate CDI-R using VR-based cardiac arrest simulations, comparing diagnostic, intervention, and collaborative activity patterns across levels of clinical expertise.\u003c/p\u003e \u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Collaborative diagnostic reasoning and clinical expertise\u003c/h2\u003e \u003cp\u003eDiagnostic activities (DAs) involve the goal-directed collection, evaluation, and integration of patient-specific information to reduce uncertainty (Fischer et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Heitzmann et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). DAs have been examined with respect to hypothesis generation, evidence evaluation, and pattern recognition (e.g., Fink et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Tay et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Gruppen, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). From a cognitive perspective, diagnostic reasoning relies on both backward and forward reasoning (Boshuizen \u0026amp; Schmidt, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1992\u003c/span\u003e), with expert clinicians often using forward reasoning and rapid pattern recognition (System 1), supported by reflective, analytical processes (System 2) when confronted with uncertainty (Kahneman, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Norman et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Diagnostic expertise is reflected through structured mental representations (see Simon \u0026amp; Newell, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e1971\u003c/span\u003e), such as illness scripts, which enable fluid diagnostic inferences and efficient cue interpretation (Charlin et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Expertise theory holds that experts solve problems with fewer steps, relying on structured knowledge and pattern recognition (e.g., Boshuizen \u0026amp; Schmidt, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1992\u003c/span\u003e; Ericsson et al., 1991).\u003c/p\u003e \u003cp\u003eIn collaborative diagnostic reasoning, collaborative activities (CAs) play a critical role. In acute care settings, such CAs typically include sharing information between and eliciting information from team members, coordination, mutual monitoring, and structured communication techniques such as closed-loop communication (e.g., Salas et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Meier et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Liu et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Witti et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The relevance of structured communication for safety-critical coordination is emphasized in frameworks like TeamSTEPPS (Department of Defense \u0026amp; Agency for Healthcare Research and Quality, 2010). Expertise in these CAs can be described in terms of internalized collaboration scripts (Fischer et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) and transactional routines that guide teams through complex, time-pressured decisions (Janssen \u0026amp; Kirschner, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The collaborative diagnostic reasoning (CDR) model (Radkowitsch et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) systematically integrates CAs with DAs, demonstrating how shared understanding and coordinated reasoning contribute to diagnostic accuracy in teams. The CDR model has recently been validated in a clinical context involving diagnostic collaboration between an internist and a radiologist, including the request and interpretation of radiological imaging (Brandl et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). While this model is valuable for understanding how diagnosticians share and coordinate evidence and hypotheses, it remains diagnosis-centered. Structurally, it maps reasoning across distributed knowledge spaces and emphasizes social and cognitive skills in building shared diagnostic representations. However, it does not distinguish intervention reasoning as a separate or equally complex process, and the dynamic interplay between diagnostic reasoning and intervention reasoning in teams remains unclear.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.2 Collaborative intervention reasoning: Conceptualization and theoretical integration\u003c/h2\u003e \u003cp\u003eAlthough intervention reasoning is essential, it has rarely been explicitly modeled. Primarily it is described as part of broader clinical reasoning skills (e.g., Liaw et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Monajemi et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) or in the context of a concrete therapeutic case (e.g., Wilcox \u0026amp; Heudes, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Richters et al. (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) recently defined intervention activities (IAs) as cognitively and interactionally distinct but functionally interdependent subcomponents, namely intervention generation, selection, implementation (i.e., planning and executing), and evaluation, that often build on each other and vary in visibility depending on the clinical context. For example, in cardiac arrest, where high acuity and time pressure demand urgent intervention, implementation and immediate coordination are often crucial and explicit, whereas in chronic care, planning and adaptation may be more prominent but implicit (see Stadler et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis variation between acute and chronic care contexts highlights the importance of not only what IAs are performed, but also how they are reasoned through and coordinated. The role of CAs in orchestrating intervention sequences, particularly in groups and teams, remains understudied. Moreover, IAs place distinct demands on CAs, requiring immediate, targeted communication such as task distribution or confirmation loops.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e1.3 CDI-R: A process model for team-based diagnostic and intervention reasoning\u003c/h2\u003e \u003cp\u003eTo address the lack of systematic research on intervention reasoning as a distinct cognitive process in collaborative clinical settings, we propose CDI-R as a theory-driven process model for collaborative diagnostic and intervention reasoning in dynamic clinical settings (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Building on the CDR model (Radkowitsch et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and drawing on research from cognitive science, clinical reasoning research, and computer-supported collaborative learning (CSCL), the CDI-R model introduces IAs as distinct from yet interdependent with DAs, both dynamically regulated by CAs. The model visualizes these components through concentric circles. This cyclical structure reflects the recursive, time-sensitive nature of collaborative reasoning in complex environments, such as cardiac arrest, where diagnosis and intervention evolve through continuous team collaboration. The outer layer of the model represents the domain-general task space (see cross-domain task space suggested by Hetmanek et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), encompassing primary DAs and IAs. The relations between these activities are non-linear because diagnostic and intervention reasoning often occur in parallel, interrupt each other, or recur iteratively in response to changing clinical conditions and team dynamics, capturing the iterative, non-sequential nature of real-time reasoning. Importantly, CDI-R conceptualizes this reasoning process as embedded within broader input variables and outcomes (see Rosen et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Individual characteristics (e.g., prior knowledge, cognitive skills, emotions, expertise; e.g., Duffy et al., 2020; Pekrun, 2006), team characteristics (e.g., composition, familiarity; Fiore, 2011), task characteristics (e.g., complexity, interdependence; e.g., Cohen, 1987), and environmental features (e.g., authenticity, fidelity) shape how collaborative diagnostic and intervention reasoning unfolds. Likewise, the process gives rise to outcomes at the task level (e.g., diagnostic and intervention-related outcomes), the individual level (e.g., learning gains, motivation), and the team level (e.g., quality of collaboration, emergent states). While these boundary conditions and consequences are integral to a full CDI-R framework, the present model description centers on the dynamic process component interlinking DAs, IAs, and CAs.\u003c/p\u003e \u003cp\u003eUnderpinning the CDI-R model are several core propositions. First, DAs and IAs constitute conceptually and functionally distinct reasoning processes (Proposition A). Second, these processes may repeat, interrupt each other, co-occur, or shift rapidly in response to unfolding conditions and task demands (Proposition B). This includes both sustained engagement within a single reasoning mode (e.g., repeated DA\u0026rarr;DA or IA\u0026rarr;IA sequences) and transitions between modes (e.g., DA\u0026rarr;IA or IA\u0026rarr;DA), reflecting different temporal organizations of collaborative reasoning. This structure illustrates that diagnostic and intervention reasoning are not staged or unidirectional but dynamically interact under time pressure.\u003c/p\u003e \u003cp\u003eAt the core of the model lies the rotating system of CAs, divided into four interrelated categories: joint information processing, structured communication, coordination, and monitoring. While communication underpins all CAs, structured communication with the example of closed-loop communication is modeled separately due to its verification-based format and distinct team function (Department of Defense \u0026amp; Agency for Healthcare Research and Quality, 2010).\u003c/p\u003e \u003cp\u003eThese CAs, in their dynamic organization, shape and coordinate ongoing DAs and IAs, while also being shaped by them. For example, closed-loop communication helps verify that instructions such as administering a specified medication at the correct dosage are both received and executed correctly, reducing the risk of errors. Similarly, structured information sharing allows team members to integrate rapidly changing, frequently recorded clinical cues such as ECG information or vital signs, enabling timely and coordinated intervention.\u003c/p\u003e \u003cp\u003eThe rotational design reflects that different types of CAs become more or less prominent depending on the phase of the task and team needs. For example, coordination intensifies during intervention implementation, such as in cardiac arrest resuscitation when the team leader assigns roles (e.g., \u0026ldquo;You do chest compressions, I\u0026rsquo;ll summarize the case\u0026rdquo;) or delegates tasks (e.g., \u0026ldquo;John, administer 1 mg epinephrine now\u0026rdquo;), while information sharing dominates during hypothesis generation, such as when team members verbalize findings (e.g., \u0026ldquo;No pulse, rhythm shows ventricular fibrillation\u0026rdquo;) or elicit input (e.g., \u0026ldquo;Do we have a shockable rhythm?\u0026rdquo;; \u0026ldquo;What am I missing, team?\u0026rdquo;). Accordingly, the model proposes that CAs vary in their relevance and function depending on the evolving demands of the task and reasoning phase (Proposition C).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eRole of expertise\u003c/em\u003e \u003c/p\u003e \u003cp\u003eThe CDI-R model also addresses the role of clinical expertise in shaping reasoning dynamics. Rather than assuming general differences in team performance, we propose that clinical expertise shapes how DAs and IAs are coordinated within teams. Clinical expertise is expected to shape how teams coordinate and transition between DAs and IAs under time pressure (Proposition D1), as more experienced clinicians are better able to align DAs and IAs in a timely and goal-directed manner. This assumption is grounded in the idea that more advanced knowledge structures enable clinicians to anticipate developments, prioritize actions, and coordinate activities more effectively within dynamic team settings. We propose that expert-led teams engage in timely, targeted IA-oriented reasoning supported by collaborative processes aligned with evolving task demands. This includes structured coordination and verification processes (e.g., role allocation or check-backs; Liu et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Salas et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), which enable smooth and safe implementation of interventions. In contrast, trainee-led teams are expected to show greater persistence within reasoning modes and a reduced likelihood of transitioning from DAs to IAs, reflecting less differentiated coordination under time pressure and reduced fluidity in transitions between reasoning modes. Thus, expertise is reflected not simply in better outcomes, but in how teams organize and transition between DAs and IAs in real time.\u003c/p\u003e \u003cp\u003eBuilding on this, the model further proposes that clinical expertise is associated with differences in the extent and organization of explicitly enacted IAs (Proposition D2). Importantly, prior research on expertise in diagnostic reasoning suggests that experts often require fewer observable steps due to pattern recognition and more efficient processing (Boshuizen \u0026amp; Schmidt, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1992\u003c/span\u003e; Norman et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). From this perspective, one might expect experts to engage in fewer observable reasoning activities overall. However, we argue that this assumption does not directly transfer to IAs. In contrast to DAs, IAs often require the explicit enactment of multiple steps (e.g., generation, selection, implementation, and evaluation; Richters et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), particularly under time pressure and uncertainty, where actions must be coordinated, communicated, and verified within the team. As a result, even when experts recognize situations rapidly, they may still need to explicitly enact and coordinate intervention steps, making IAs less compressible than DAs. From this perspective, greater clinical expertise may be reflected in a more extensive and systematically organized engagement in IAs, rather than in a reduction of observable activity. This does not imply inefficiency, but rather reflects analytic engagement with the situation, such as anticipating complications, running parallel strategies, or deliberately verifying outcomes. Expertise may thus manifest not in doing fewer IAs, but in organizing and sequencing them more effectively under dynamic conditions. Prior empirical findings support this perspective: intervention-focused reasoning tends to increase with clinical expertise (Monajemi et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), and even experienced clinicians invest considerable cognitive effort in dynamic care situations (Hartjes et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Consistent with this, experts tend to anticipate critical developments earlier and organize team actions accordingly when recognizing the severity of a situation (Popov et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"2 Empirical model validation","content":"\u003cp\u003eBased on the CDI-R process model, we conducted an empirical, theory-guided study to examine its core propositions in a simulated acute care context. We analyzed data from a virtual reality (VR) simulation of cardiac arrest management (Kentros et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Popov et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) involving expert- and trainee-led teams.\u003c/p\u003e \u003cp\u003eThe study investigates how DAs, IAs, and CAs unfold and interact over time, and how these patterns vary based on clinical expertise. Guided by the model, we explored three central aspects: First, we explored whether diagnostic and intervention reasoning can be empirically distinguished in both their temporal organization and communicative structure (Proposition A), and whether they exhibit non-linear, recurrent transition patterns over time (Proposition B). Second, we investigated whether CAs vary systematically depending on whether teams are engaged in diagnostic or intervention reasoning (Proposition C). Third, we examined how expertise shapes these dynamics (Propositions D1 and D2), focusing on differences in (a) the distribution of DAs and IAs, (b) the fluidity of transitions between them (i.e., how flexible teams switch between DAs and IAs), and (c) the structural organization of CAs within each reasoning mode.\u003c/p\u003e \u003cp\u003eThe study evaluates the extent to which the observed temporal and structural patterns in team communication correspond to the theoretical propositions of the CDI-R model.\u003c/p\u003e"},{"header":"3 Method","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.1 VR simulation and participants\u003c/h2\u003e \u003cp\u003eThis study uses a multi-user \u0026ldquo;open-world\u0026rdquo; VR simulation (Popov et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) that enables a team of four clinicians to interact with each other and a virtual patient in a dynamic and non-deterministic in-hospital cardiac arrest resuscitation scenario. VR offers a realistic yet controlled setting for capturing the time-pressured interplay of teams\u0026rsquo; DAs, IAs and CAs. To effectively complete the ~\u0026thinsp;15-minute simulation, the team must designate individual roles and responsibilities, such as team lead, chest compressions, airway management, electrical shock management, and medication administration (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The six-stage clinical scenario begins with the care team receiving handoff from a bedside nurse non-player character and a virtual patient being in Ventricular Tachycardia with a weak pulse (Stage 1), requiring synchronized cardioversion and intubation. The patient rapidly deteriorates into pulselessness (Stages 2\u0026ndash;3), Asystole (Stage 4), and Ventricular Fibrillation (Stage 5), demanding continuous CPR, unsynchronized defibrillation, and appropriate medication (epinephrine, antiarrhythmics). The simulation is designed to conclude upon Return of Spontaneous Circulation (ROSC; Stage 6), when appropriate interventions are performed, prompting a shift to post-resuscitation care.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e Participants must communicate dynamically, exchanging information under time pressure and dynamically reallocating tasks. Participants were placed in the same room, each wearing a VR headset (i.e., facial cues unavailable) and standing in a semicircle to facilitate verbal coordination. A convenience sample was recruited as part of a mandatory monthly code team training at a large U.S. academic medical center with extensive experience in cardiac arrest simulation. In total, we analyzed data from 29 simulated VR cardiac arrest sessions (\u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;29 teams, 116 participants in total). Each 4-person session was directed by a single, non-repeating participant serving as the designated team leader, yielding nine expert-led teams and twenty trainee-led teams.\u003c/p\u003e \u003cp\u003eExpert status was defined as current Advanced Cardiovascular Life Support (ACLS) certification and at least 5 years of experience leading cardiac arrest teams. The nine expert team leaders consisted of a diverse cohort of experienced clinicians: nurses (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;4), paramedics (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2), emergency medicine physicians (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2), and a family medicine physician (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1). In five of these sessions, the two non-leader roles (e.g., procedural tasks of chest compressions and/or airway management) were filled by research staff.\u003c/p\u003e \u003cp\u003eThe twenty trainee-led sessions (hereafter referred to as trainee-led) were directed by junior emergency medicine (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;13) or family medicine (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;7) residents. These trainee leaders had completed prior ACLS training but possessed little to no experience leading real cardiac arrest teams. In these simulations, the three non-leader roles were filled by fellow residents.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Data coding\u003c/h2\u003e \u003cp\u003eTeam interactions were automatically recorded and transcribed. Transcripts were segmented into communicative events at the utterance level using ELAN (Wittenburg et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). Annotations were performed by six domain experts in emergency medical services using a theory-informed coding scheme derived from the CDI-R model (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Two dimensions were coded: CAs and the reasoning mode, i.e., whether it refers to diagnosing (DA) or intervening (IA).\u003c/p\u003e \u003cp\u003eTo establish coding reliability, Kappa scores were calculated from three recordings (each approximately twelve minutes long). The analysis demonstrated inter-annotator agreement: 0.73 for the main codes, 0.64 for the sub-codes (specific CAs), and 0.82 for the reasoning focus (DA vs. IA). Although some subjectivity remains due to the task\u0026rsquo;s inherent complexity, these expert-validated transcripts constituted a robust ground-truth dataset for our analyses.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCoding scheme\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMain Code\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSub-code\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDefinition\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eExamples\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u003cb\u003eJoint Information Processing\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSharing Information (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eProviding factual information (new or synthesis of existing), including recapping\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;We\u0026rsquo;re at one minute, 30 seconds right now\u0026rdquo;\u003c/p\u003e \u003cp\u003e\u0026ldquo;Latest vitals show\u0026hellip;\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eQuestioning / Eliciting Information (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRequesting information from others\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;What are the latest labs?\u0026rdquo;\u003c/p\u003e \u003cp\u003e\u0026ldquo;Any new findings?\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEvaluating Shared Information (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDiscussing/assessing provided information, explaining rationale and justifications\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;Vitals are concerning\u0026rdquo;!\u003c/p\u003e \u003cp\u003e\u0026ldquo;I disagree\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eStructured communication\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCLC: Checkback Statement (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReceiver repeats instructions/information for verification (intervention-related only)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;Yes, I will shock at 200J\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eCoordination\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAssigning Roles (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDesignating roles or verifying expertise\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;You do airway, I'll do meds\u0026rdquo;\u003c/p\u003e \u003cp\u003e\u0026ldquo;You handle respiratory, he\u0026rsquo;ll do IV\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAllocating Tasks (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDistributing tasks and implementing plans\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;Shock at 150J\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eMonitoring\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExpressing Uncertainty (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVoicing doubts or lack of confidence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;Not sure what\u0026rsquo;s causing this\u0026rdquo;\u003c/p\u003e \u003cp\u003e\u0026ldquo;His other EKG was looking like he might be having an MI?\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSharing Hypothesis (CA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFormulating working diagnosis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026ldquo;Patient may become septic\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eDiagnostic- and Intervention reasoning\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDiagnosis-related (DA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCollection and interpretation of case-specific information to reduce diagnostic uncertainty\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAssessment of symptoms, lab results interpretation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntervention-related (IA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eActivities improving patient condition through generation, implementation, monitoring, and evaluation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTreatment planning, medication administration\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cem\u003eNote.\u003c/em\u003e CA\u0026thinsp;=\u0026thinsp;Collaborative Activity, DA\u0026thinsp;=\u0026thinsp;Diagnostic Activity, IA\u0026thinsp;=\u0026thinsp;Intervention Activity.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Analyses\u003c/h2\u003e \u003cp\u003eThe analytic approach was designed to examine both the temporal dynamics and the structural organization of team-based diagnostic and intervention reasoning, in line with the study\u0026rsquo;s propositions regarding the distinction between DAs and IAs (A), their non-linear and recurrent organization (B), the variation of CAs across reasoning modes (C), and expertise-related differences in these dynamics (D1 and D2). Analyses were conducted at the team level. To capture these complementary dimensions, we combined Lag Sequence Analysis (LSA; Bakeman, R., \u0026amp; Quera, V., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Sackett, G. P., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e1979\u003c/span\u003e) to model temporal progression and Epistemic Network Analysis (ENA; Shaffer et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) to examine patterns of co-occurrence and connections between CAs. These combined analytical approaches provide both a sequential and a relational perspective on team-based reasoning.\u003c/p\u003e \u003cp\u003e \u003cem\u003eLag Sequence Analysis (LSA)\u003c/em\u003e \u003c/p\u003e \u003cp\u003eWe employed LSA to quantify how teams transitioned between DAs and IAs over time, directly addressing Propositions B (non-linearity and recurrence) and D1 (expertise-related transition dynamics). We constructed transition matrices to calculate the conditional probability P(Next|Current) of transitions from a current activity to a subsequent activity. For example, we estimated how likely it was that a DA was followed by another DA or by an IA in the next utterance. These probabilities capture how teams move between reasoning modes during interaction.\u003c/p\u003e \u003cp\u003eSustained team behaviors were indicated by elevated conditional probabilities for self-transitions (DA\u0026rarr;DA or IA\u0026rarr;IA), reflecting persistence and recurrence within interaction dynamics. In contrast, directed forward progression was characterized by higher probabilities for DA\u0026rarr;IA transitions alongside lower rates of IA\u0026rarr;DA transitions, indicating a shift from diagnostic to intervention-oriented reasoning. All transition probabilities were calculated at the team level and then aggregated for expert- and trainee-led teams, rather than reported for individual sequences.\u003c/p\u003e \u003cp\u003eTo determine statistical reliability, we calculated adjusted residuals (z-scores) for each transition, applying a threshold of |z| \u0026ge; 1.96 (p \u0026lt; .05) to identify significantly characteristic (\u0026gt;\u0026thinsp;1.96) or inhibited (\u0026lt; -1.96) transitions. These z-scores indicate whether a given transition occurs more or less often than expected by chance.\u003c/p\u003e \u003cp\u003eTo assess the stability of reasoning loops (\u0026ldquo;stuckness\u0026rdquo;) versus linear progression, we extended the analysis beyond immediate sequential behaviors (Lag-1) to multiple steps (Lags 2\u0026ndash;5). This allowed us to examine whether teams remained within the same reasoning mode across several consecutive utterances or transitioned more flexibly between DAs and IAs. Persistence was operationalized as sustained self-transition probabilities and statistically significant z-scores across increasing lags, whereas faster decay of z-scores indicated greater temporal fluidity and reduced persistence.\u003c/p\u003e \u003cp\u003e \u003cem\u003eEpistemic Network Analysis (ENA)\u003c/em\u003e \u003c/p\u003e \u003cp\u003eWhile LSA models sequential transitions, ENA models the complex co-occurrence of the activities (i.e., codes) within a moving temporal window (set to six utterances) to create weighted network associations. We used ENA to investigate structural distinctiveness across reasoning modes (Proposition A), collaborative activity variation (Proposition C), and expertise-related differences (Propositions D1 and D2).\u003c/p\u003e \u003cp\u003eTo ensure consistent evaluation, we applied a standardized approach to network comparison and interpretation (Z\u0026ouml;rgő et al., 2024). We projected networks into a shared geometric space and generated subtraction networks to visualize distinct connection topologies, highlighting connections that are stronger in one context than the other, where edge weights and colors indicate context dominance. We quantitatively compared the positions of network centroids using two-sample t-tests (assuming unequal variance) along the primary dimension (X-axis). Significant centroid separation (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) and Cohen's d effect sizes served as evidence of systematic structural differences in communicative organization.\u003c/p\u003e \u003cp\u003eWe aggregated and compared data in two distinct phases. First, to test the structural dissociation between reasoning modes, we aggregated data across all teams to contrast a pooled Intervention Network (CA co-occurrences within IA-coded segments) against a Diagnostic Network (DA-coded segments). Structural dissociation was considered supported if the network centroids were significantly separated (p \u0026lt; .05) with a meaningful effect size, and not supported if no significant separation or only negligible effects were observed.\u003c/p\u003e \u003cp\u003eSecond, to evaluate expertise differences while controlling for these inherent structural variations, we separated the dataset into distinct groups based on team expertise. By ensuring that expert and trainee data points were not mixed, we compared their networks independently within diagnostic and intervention contexts. Expertise differences were considered supported if expert and trainee centroids differed significantly within a given context and not supported if no significant differences between groups were observed.\u003c/p\u003e \u003c/div\u003e"},{"header":"4 Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Distribution of Diagnostic and Intervention Activities\u003c/h2\u003e \u003cp\u003eThe analysis of the total activity distribution reveals distinct differences in reasoning focus between teams (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Expert-led teams dedicated a significantly larger proportion of their communication to IAs (66.2%) compared to DAs (33.8%). In contrast, Trainees exhibited a more balanced split but a comparatively higher reliance on diagnosis, with 54.2% of activities coded as IAs and 45.8% as DAs. This pattern is consistent with Proposition D2, indicating that expert-led teams engaged in a greater proportion of explicit intervention-oriented activity, whereas trainee-led teams showed a comparatively higher proportion of diagnosis-oriented communication.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eFrequency and percentage of communicative activities (diagnostic and intervention) by scenario stage and team\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSimulation Stage\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eAcross both Types of Teams\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eExpert-led Teams\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTrainee-led Teams\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"1\" nameend=\"c6\" namest=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eStage 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.93%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13.59%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e24.04%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eStage 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15.61%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15.04%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e15.84%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eStage 3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.36%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22.48%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e20.88%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eStage 4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9.67%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10.94%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e9.14%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eStage 5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.24%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21.79%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e19.58%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eStage 6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12.19%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16.15%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e10.51%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003e\u003cem\u003eNote.\u003c/em\u003e Stages: Stage 1\u0026thinsp;=\u0026thinsp;ventricular tachycardia with pulse (initial presentation ), Stage 2\u0026ndash;3\u0026thinsp;=\u0026thinsp;deterioration to pulselessness, Stage 4\u0026thinsp;=\u0026thinsp;asystole, Stage 5\u0026thinsp;=\u0026thinsp;ventricular fibrillation, Stage 6\u0026thinsp;=\u0026thinsp;return of spontaneous circulation (ROSC).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eExamining the volume of activity across the six stages of the simulation highlights differences in team momentum and focus (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Trainees spent the highest percentage of their communicative effort in Stage 1 (24.04%). Experts, conversely, peaked in Stage 3 (22.48%), aligning their maximum communicative output with the scenario's active intervention phase. A notable divergence occurred at the Return of Spontaneous Circulation (ROSC; Stage 6). Trainees showed a resurgence of DAs at this stage, while experts maintained a steady focus on IA. Across all stages, experts engaged in a higher percentage of IA-based communication than trainees, who maintained higher DA ratios throughout the scenario.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Lag Sequence Analysis (LSA)\u003c/h2\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the transition dynamics for the entire dataset at Lag-1. The overall pattern reveals a significant tendency for teams to maintain their collective attention on a single reasoning process. The strongly positive z-scores for self-transition (DA\u0026rarr;DA and IA\u0026rarr;IA) indicate that teams are highly likely to maintain their current state rather than rapidly shifting between modes. This highlights the clustered nature of team interactions and the general stability of reasoning blocks during the simulation.\u003c/p\u003e \u003cp\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhen stratified by experience level (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e), distinct transition patterns emerged. Experts were significantly more likely to transition from DAs to IAs (\u003cem\u003eP\u003c/em\u003e(DA\u0026rarr;IA) = .37) compared to trainees (\u003cem\u003eP\u003c/em\u003e(DA\u0026rarr;IA) = .28). This reflects a more direct progression from problem formulation to solution implementation in expert-led teams. Conversely, experts were less likely to switch back from an IA to a DA (\u003cem\u003eP\u003c/em\u003e(IA\u0026rarr;DA) = .19) compared to trainees (\u003cem\u003eP\u003c/em\u003e(IA\u0026rarr;DA = .23).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAnalyzing transitions across multiple lags (Lags 1\u0026ndash;5) revealed distinct patterns of cognitive persistence versus fluidity (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003ea and \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003eb). Trainees exhibited a higher probability of remaining in a Diagnostic Loop (DA\u0026rarr;DA) across all observed lags (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003eb). For example, at Lag-1, the self-transition probability for trainees was 0.72 compared to 0.63 for experts. As the lag increased, trainees remained more likely to be \u0026ldquo;stuck\u0026rdquo; in specific reasoning loops (both DA and IA), indicating that they tend to spend more time iterating within a single reasoning category when communicating.\u003c/p\u003e \u003cp\u003eIn contrast, the lower self-transition probabilities for experts across Lags 1\u0026ndash;5 suggest a more linear progression. Experts transitioned from diagnostic to intervention sequences more quickly, showed less fragmented, cyclical reasoning patterns observed in trainee-led teams, thereby supporting the model's prediction regarding expertise-driven transition fluidity (Proposition D1).\u003c/p\u003e \u003cp\u003eFurther insight into the persistence of prior states across lags is provided by the decay of adjusted residuals (z-scores). For expert-led teams, the z-scores for self-transitions decayed rapidly, dropping below the significance threshold (|z| \u0026lt; 1.96) starting at Lag-5. This suggests that an expert-led team's current reasoning is relatively independent of actions taken five or more steps earlier, enabling fluid adaptation to new stimuli. In contrast, trainee z-scores remained significantly above the threshold up to Lag-8, indicating a strong, prolonged influence from the past that constrained their ability to switch gears.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Epistemic Network Analysis (ENA)\u003c/h2\u003e \u003cp\u003eBeyond temporal sequences, the Epistemic Network Analysis (ENA) models reveal that DAs and IAs form fundamentally different structural networks of communication (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e8\u003c/span\u003e). The projection of the network centroids confirmed a statistically significant separation between the two reasoning modes along the primary dimension (Mean\u003csub\u003eIA\u003c/sub\u003e = -0.14, SD\u003csub\u003eIA\u003c/sub\u003e = 0.12 vs. Mean\u003csub\u003eDA\u003c/sub\u003e = 0.14, SD\u003csub\u003eDA\u003c/sub\u003e = 0.16; t(52.47)= -7.13, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, Cohen's d\u0026thinsp;=\u0026thinsp;1.8). This strong effect size indicates that IA and DA represent distinct reasoning modes with systematically different communicative states (Proposition A).\u003c/p\u003e \u003cp\u003eVisual inspection of the network nodes (centroids) reveals clear thematic clustering by reasoning goal. In the intervention cluster, the codes \u0026ldquo;Allocating Tasks\u0026rdquo; and \u0026ldquo;CLC: Checkback Statement\u0026rdquo; are heavily positioned toward the IA side, serving as the primary mechanisms for executing and verifying care. The subtraction network, highlighting differences in connection strength between IA and DA segments, shows strong, cohesive links between these two codes on the IA side, suggesting a tightly coordinated execution pattern.\u003c/p\u003e \u003cp\u003eConversely, the diagnostic cluster on the DA side shows a more distributed pattern of connections among \u0026ldquo;Questioning / Eliciting Information\u0026rdquo;, \u0026ldquo;Sharing Information\u0026rdquo;, \u0026ldquo;Sharing Hypothesis\u0026rdquo;, and \u0026ldquo;Evaluating Shared Information\u0026rdquo;, reflecting the integration of multiple information-related activities during diagnostic reasoning, which tend to co-occur in close temporal proximity.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe subsequently examined how experts and trainee teams differ in their communicative structure, specifically during DAs (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e9\u003c/span\u003e). Expert-led teams (Mean = -0.17, SD\u0026thinsp;=\u0026thinsp;0.16) were statistically distinct from trainee-led teams (Mean\u0026thinsp;=\u0026thinsp;0.08, SD\u0026thinsp;=\u0026thinsp;0.14) along the primary dimension (t(13.83)= -4.04, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, Cohen's d\u0026thinsp;=\u0026thinsp;1.71). This large effect size confirms that even when expert and trainee teams are ostensibly \u0026ldquo;diagnosing\u0026rdquo;, they are engaging in fundamentally different communicative processes. The positioning of nodes in the projected space illustrates a divergence in functional differentiation. Trainee networks clustered closely with \u0026ldquo;Questioning/Eliciting Information\u0026rdquo;, \u0026ldquo;Expressing Uncertainty\u0026rdquo;, and \u0026ldquo;Assigning Roles\u0026rdquo;, characterizing diagnosis as a homogeneous pattern of information seeking and role clarification. Expert networks, however, clustered with \u0026ldquo;Sharing Hypothesis\u0026rdquo;, \u0026ldquo;Allocating Tasks\u0026rdquo;, and \u0026ldquo;CLC: Checkback Statement\u0026rdquo;. Notably, experts utilized these action-oriented activities even within diagnostic segments, possibly \u0026ldquo;pre-loading\u0026rdquo; the intervention phase. The difference network is in support of these strategies: trainees showed stronger unique connections linking \u0026ldquo;Questioning / Eliciting Information\u0026rdquo; to \u0026ldquo;Sharing Information\u0026rdquo;, whereas experts displayed stronger connections linking \u0026ldquo;Sharing Hypothesis\u0026rdquo; to \u0026ldquo;Allocating Tasks\u0026rdquo;. Similar structural differences emerged when teams engaged in IAs (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e10\u003c/span\u003e). Expert-led teams (Mean = -0.13, SD\u0026thinsp;=\u0026thinsp;0.15) were significantly distinct from trainee-led teams (Mean\u0026thinsp;=\u0026thinsp;0.06, SD\u0026thinsp;=\u0026thinsp;0.16; t(15.90)= -3.04, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.01, Cohen's d\u0026thinsp;=\u0026thinsp;1.21). While the effect size is slightly smaller than in the diagnostic context, it remains robust, indicating that \u0026ldquo;intervening\u0026rdquo; means something structurally different to experts than it does to trainees. Trainee networks demonstrated continued use of diagnostic-oriented activities in the intervention phase, maintaining strong associations with \u0026ldquo;Questioning/Eliciting Information\u0026rdquo; and \u0026ldquo;Sharing Information\u0026rdquo; rather than switching purely to execution. Expert networks were tightly clustered around \u0026ldquo;Allocating Tasks\u0026rdquo; and \u0026ldquo;CLC: Checkback Statement\u0026rdquo;, prioritizing tightly coordinated execution patterns. The expert difference network was dominated by a strong triangle of connections between \u0026ldquo;Allocating Tasks\u0026rdquo;, \u0026ldquo;CLC: Checkback Statement\u0026rdquo;, and \u0026ldquo;Sharing Information\u0026rdquo;, reflecting a confirm-assign-verify protocol. In contrast, the strongest unique connections for trainees linked \u0026ldquo;Questioning / Eliciting Information\u0026rdquo; to \u0026ldquo;Allocating Tasks\u0026rdquo; and \u0026ldquo;Sharing Information\u0026rdquo;, likely reflecting hesitation and interrupted instructions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eComparing the results across both contexts reveals a fundamental difference in how the teams led by experts and those led by trainees organize their reasoning. Trainees exhibited a relatively homogenous communicative structure across both contexts. Whether diagnosing or intervening, their networks were consistently pulled toward \u0026ldquo;Questioning/Eliciting Information\u0026rdquo; and \u0026ldquo;Sharing Information\u0026rdquo;. This suggests they tend to lack a distinct \u0026ldquo;execution mode\u0026rdquo; and instead rely on a continuous, undifferentiated process of information gathering to manage uncertainty. Experts, however, demonstrated high functional differentiation. In the DA context, their structure bridged hypothesis generation with preparatory assignments, acting as a planning mode. In the IA context, they shifted entirely to a tight loop of assignment and verification, acting primarily in an execution mode.\u003c/p\u003e \u003c/div\u003e"},{"header":"5 Discussion","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Summary of findings\u003c/h2\u003e \u003cp\u003eThis study provides an initial empirical examination of the CDI-R process model in a high-stakes acute care simulation. All observable team communication was coded as CAs and additionally tagged as DA- or IA-oriented, allowing us to analyze distributional, temporal, and structural characteristics of DAs and IAs and their variation by expertise.\u003c/p\u003e \u003cp\u003eAt the distributional level, expert-led teams devoted a larger proportion of their communication to IA-oriented segments, whereas trainee-led teams showed a comparatively higher proportion of DA-oriented communication. Experts completed the simulation in less time while showing a higher proportion of IA-oriented activity. Across stages, trainees showed greater DA resurgence in later segments, whereas experts maintained IA dominance. These findings align with Proposition D2 and indicate that explicit engagement in IAs increases with expertise.\u003c/p\u003e \u003cp\u003eAt the temporal level, LSA revealed significant maintenance probabilities for both DA\u0026rarr;DA and IA\u0026rarr;IA transitions, indicating recurrent reasoning phases and supporting Proposition B. Expertise shaped transition dynamics (Proposition D1). Expert-led teams showed higher DA\u0026rarr;IA transition probabilities and lower IA\u0026rarr;DA reversions at Lag-1. Across multiple lags, expert self-transition effects decayed more rapidly, whereas trainee-led teams showed sustained DA\u0026rarr;DA and IA\u0026rarr;IA persistence across lags, indicating less dynamic progression.\u003c/p\u003e \u003cp\u003eAt the structural level, pooled ENA models demonstrated clear dissociation between DA- and IA-oriented segments (Proposition A). DA-oriented segments showed stronger co-occurrences among Sharing Information, Questioning / Eliciting Information, Evaluating Shared Information, Sharing Hypothesis, and Expressing Uncertainty. IA-oriented segments were characterized by stronger connections between Allocating Tasks and CLC: Checkback Statement. This phase-dependent variation in CA configuration supports Proposition C.\u003c/p\u003e \u003cp\u003eStratified ENA further revealed expertise-related differences within each reasoning orientation (Proposition D1). In DA-oriented segments, expert-led teams showed stronger links between Sharing Hypothesis and Allocating Tasks, whereas trainee-led teams showed denser connections among Sharing Information, Questioning / Eliciting Information, and Expressing Uncertainty. In IA-oriented segments, expert-led teams demonstrated tightly connected Allocating Tasks and CLC: Checkback Statement configurations, while trainee-led teams maintained stronger co-occurrences among joint information processing activities. Across contexts, expert networks showed greater differentiation between DA- and IA-oriented configurations, whereas trainee networks exhibited more homogeneous structures.\u003c/p\u003e \u003cp\u003eTaken together, the validation findings provide converging support for all core propositions of the CDI-R model, while also revealing \u003cem\u003ehow\u003c/em\u003e expertise modulates their interaction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Implications for CDI-R Model\u003c/h2\u003e \u003cp\u003eThe CDI-R model was introduced as a theory-driven process model for collaborative diagnostic and intervention reasoning. The present findings provide initial empirical support for its core assumptions and contribute to refining its theoretical propositions. Building on prior work on the CDR model (Radkowitsch et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Brandl et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and extending efforts to integrate contextual and interprofessional factors into healthcare problem-solving frameworks (Witti et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), CDI-R advances a process-oriented perspective on team-based clinical reasoning that integrates diagnostic, intervention, and collaborative dynamics.\u003c/p\u003e \u003cp\u003eIn the following, we discuss the implications of the findings for each proposition (A\u0026ndash;D), while explicitly addressing how the propositions interrelate.\u003c/p\u003e \u003cp\u003e \u003cem\u003eProposition A: Structural Distinction of DAs and IAs\u003c/em\u003e \u003c/p\u003e \u003cp\u003eProposition A states that DAs and IAs constitute conceptually and functionally distinct reasoning processes.\u003c/p\u003e \u003cp\u003eThe ENA findings provide structural support for this assumption. Across pooled analyses, DA-oriented segments were characterized by stronger connections among information sharing, questioning / eliciting information, evaluating shared information, sharing hypotheses, and expressing uncertainty. In contrast, IA-oriented segments showed structurally stronger connections involving allocating tasks and CLC (i.e., checkback statements).\u003c/p\u003e \u003cp\u003eThese results indicate that diagnostic and intervention reasoning are associated with distinct configurations of CAs. This supports the theoretical extension of the CDR model (Radkowitsch et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), which primarily focused on diagnostic knowledge coordination. By demonstrating that intervention-oriented segments are structurally dominated by coordination and verification processes, the findings support the view that intervention reasoning constitutes a distinct process rather than a direct extension of diagnostic reasoning.\u003c/p\u003e \u003cp\u003eAt the same time, the distinction observed under Proposition A is closely linked to Proposition C, which posits that CAs vary systematically depending on reasoning orientation. The distinction between DAs and IAs is reflected in systematic differences in the CAs associated with each reasoning mode. In this sense, the empirical support for Proposition A is closely linked to Proposition C, which specifies how these differences in collaborative activities emerge.\u003c/p\u003e \u003cp\u003e \u003cem\u003eProposition B: Non-Linear and Recurrent Interaction of DAs and IAs\u003c/em\u003e \u003c/p\u003e \u003cp\u003eProposition B assumes that DAs and IAs unfold in non-linear, recurrent patterns rather than in a strictly staged sequence. The LSA findings confirm this dynamic organization. In line with our analytic definitions, significant maintenance probabilities (DA\u0026rarr;DA; IA\u0026rarr;IA) indicate sustained reasoning phases, while switch transitions (DA\u0026rarr;IA; IA\u0026rarr;DA) show that reasoning modes interact and alternate over time.\u003c/p\u003e \u003cp\u003eThe multi-lag analyses further clarified that recurrence does not imply rigidity. Although both expertise levels exhibited sustained reasoning loops (Proposition B), expert-led teams showed faster decay of maintenance effects across lags, indicating a faster decline in the influence of prior states. This suggests that while recurrent patterns characterize collaborative reasoning in general, their temporal dynamics vary as a function of expertise (Proposition D1).\u003c/p\u003e \u003cp\u003eThese findings align with theoretical perspectives emphasizing that reasoning in complex environments is recursive and adaptive rather than linear (Hetmanek et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Radkowitsch et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Within the CDI-R framework, non-linearity is not noise but a structural feature of collaborative reasoning under dynamic task conditions. Thus, non-linear recurrence is a general feature of collaborative reasoning (Proposition B), but its temporal dynamics vary systematically with expertise (Proposition D1).\u003c/p\u003e \u003cp\u003e \u003cem\u003eProposition C: Phase-Dependent Modulation of Collaborative Activities\u003c/em\u003e \u003c/p\u003e \u003cp\u003eProposition C posits that CAs vary in relevance and configuration depending on whether teams are engaged in DAs or IAs. The ENA results provide clear empirical support for this. DA-oriented segments were structurally centered on joint information processing and monitoring activities, including sharing information, questioning / eliciting information, evaluating shared information, sharing hypothesis, and expressing uncertainty. IA-oriented segments were structurally centered on coordination and structured communication, particularly allocating tasks and CLC (i.e., checkback statement).\u003c/p\u003e \u003cp\u003eThese findings are consistent with theoretical work on team coordination and structured communication in high-risk domains (Salas et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Liu et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), which highlights the role of verification and role clarity during action implementation. They thus support the extension of collaborative reasoning models toward coordinated action, as proposed in CDI-R.\u003c/p\u003e \u003cp\u003eImportantly, expertise again modulated these phase-dependent configurations (Proposition D1). Expert-led teams showed more differentiated CA structures across DA- and IA-oriented segments, whereas trainee-led teams exhibited more homogeneous CA configurations across reasoning modes. This suggests that expertise shapes how clearly collaborative structures align with reasoning orientation.\u003c/p\u003e \u003cp\u003eThus, while CAs vary systematically with reasoning orientation (Proposition C), the degree of differentiation between these configurations depends on expertise (Proposition D1).\u003c/p\u003e \u003cp\u003e \u003cem\u003ePropositions D1 and D2: Expertise as Coordination and Explicit Intervention Engagement\u003c/em\u003e \u003c/p\u003e \u003cp\u003eProposition D1 states that expertise becomes visible in how teams coordinate and transition between DAs and IAs under time pressure. Proposition D2 states that experienced clinicians may engage in a higher proportion of explicitly enacted IAs. The present findings support both propositions while also refining them.\u003c/p\u003e \u003cp\u003eFirst, expert-led teams engaged in a higher proportion of IA-oriented segments while completing the simulation in less time. This finding is theoretically significant. Classical expertise theory would predict fewer observable steps to reach the same solution due to pattern recognition and knowledge encapsulation (Boshuizen \u0026amp; Schmidt, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1992\u003c/span\u003e). In the context of intervention reasoning, however, this assumption appears insufficient. The present findings suggest that expertise may involve more, not fewer, explicitly enacted intervention actions when these are strategically timed and tightly coordinated. This aligns with evidence that intervention-focused reasoning rather increases than decreases with clinical experience (Monajemi et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and that experienced clinicians invest substantial effort in dynamic interventions (Hartjes et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). These results therefore challenge the assumption, derived from classical expertise theory, that expert performance is characterized by fewer observable steps, and refine Proposition D2. Expertise in intervention-rich environments may manifest not as reduced observable activity, but as structured, anticipatory, and verification-driven coordination of action. This apparent deviation from minimalist expertise accounts invites further interpretation. One explanation for the increased IA frequency is that certain intervention steps serve deliberate safety and verification functions. Experts may enact additional IAs as structured double checks (e.g., confirming device functioning or task execution), reflecting monitoring and error-prevention strategies rather than inefficiency.\u003c/p\u003e \u003cp\u003eSecond, expertise shaped the temporal and structural organization of reasoning. Expert-led teams showed higher DA\u0026rarr;IA transition probabilities, reduced persistence within single reasoning modes, and more differentiated CA configurations across reasoning orientations. These findings directly support Proposition D1 while also reinforcing Propositions B and C. A similar expertise-related pattern was visible in how teams responded to the onset of the ROSC phase of the simulation. Trainee-led teams often appeared to treat ROSC as signaling completion of the scenario and shifted toward diagnostic discussion and reflection on the patient\u0026rsquo;s condition. In contrast, expert-led teams tended to interpret ROSC as a trigger for the next intervention phase, immediately transitioning toward post-resuscitation care. This pattern further illustrates how expertise shapes the timing and coordination of intervention-oriented reasoning under dynamic task conditions.\u003c/p\u003e \u003cp\u003eBeyond these differences in frequency (Proposition D2), the structural findings suggest a pattern of deeper integration. In expert-led teams, links between Sharing Hypothesis and Allocating Tasks in DA-oriented segments indicate that diagnostic formulation is already aligned with anticipated intervention. This suggests that intervention reasoning may function, in part, as a diagnostic probe, for example by systematically disconfirming alternative causal explanations (e.g., H\u0026rsquo;s and T\u0026rsquo;s reversible causes of cardiac arrest; Reddy \u0026amp; Hanmandlu, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), consistent with iterative models of generation, selection, and evaluation (Richters et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Expertise may therefore involve a closer alignment between diagnostic formulation (DAs) and subsequent intervention-oriented actions (IAs).\u003c/p\u003e \u003cp\u003eThis pattern suggests that, with increasing expertise, the functional boundary between DAs and IAs becomes more permeable.\u003c/p\u003e \u003cp\u003eThis pattern is consistent with the CDI-R model, in which Propositions A\u0026ndash;C describe structural properties of collaborative reasoning, whereas Proposition D specifies how these structures are more coherently integrated and adaptively coordinated with increasing expertise.\u003c/p\u003e \u003cp\u003e \u003cem\u003eSimulation Context and Model Generalizability\u003c/em\u003e \u003c/p\u003e \u003cp\u003eAlthough the present empirical test was conducted in a VR-based cardiac arrest simulation, the CDI-R model itself is not restricted to this specific context. Rather, it conceptualizes how collaborative diagnostic and intervention reasoning unfold in dynamic, time-pressured clinical environments. The present study therefore provides an empirical test case within acute care simulation rather than delimiting the model to that context.\u003c/p\u003e \u003cp\u003eFuture research in live clinical settings and across different domains of care will be necessary to examine the generalizability of these dynamics. Building on prior applications of the CDR model in clinical contexts (Brandl et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and efforts to incorporate contextual and interprofessional factors into healthcare problem-solving frameworks (Witti et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), CDI-R is positioned as a flexible process model adaptable across clinical environments.\u003c/p\u003e \u003cp\u003e \u003cem\u003eInterrelation of Propositions and Conceptual Refinement\u003c/em\u003e \u003c/p\u003e \u003cp\u003eAlthough Propositions A\u0026ndash;D were introduced as analytically distinct components of the model, their interrelations follow directly from the theoretical structure of CDI-R. Specifically, Propositions A\u0026ndash;C describe structural and temporal properties of collaborative reasoning, whereas Proposition D specifies how these processes are shaped by expertise. The empirical support for the individual propositions therefore also supports the coherence of the overall model (see Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEmpirical support for CDI-R propositions\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProposition\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTheoretical claim\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEmpirical support (this study)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDAs and IAs are structurally distinct reasoning processes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eENA showed significant centroid separation and distinct CAs configurations\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDAs and IAs unfold in non-linear, recurrent patterns\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLSA revealed significant self-transitions and switching patterns\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCAs vary depending on reasoning mode\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eENA showed systematic differences in CAs configurations between DAs and IAs segments\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eD1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExpertise shapes coordination and transition dynamics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eExperts showed higher DA\u0026rarr;IA transitions, lower stuckness, more differentiated structures\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eD2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExpertise linked to greater explicit IA frequency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eExperts showed higher proportion of IA-oriented activity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eBeyond supporting the individual propositions, the findings also point to a broader conceptual implication regarding the nature of collaborative reasoning. In team-based clinical settings, reasoning is not solely an individual cognitive process but is enacted through interaction. DAs and IAs involve both individual cognitive processing and collaboratively organized communicative activities. While the analytic separation of propositions supports theoretical clarity, the empirical patterns observed here underscore their tight coupling in practice. This should not be interpreted as a limitation of measurement but rather reflects the inherently interactive nature of collaborative reasoning. Within the CDI-R framework, collaboration is therefore not an external layer added to reasoning but constitutes the medium through which diagnostic and intervention processes are coordinated and realized.\u003c/p\u003e \u003cp\u003eThe present findings therefore support CDI-R as an integrated collaborative process model rather than as a set of isolated claims. They strengthen the theoretical necessity of intervention reasoning as a distinct construct within clinical reasoning research and contribute to refining how expert performance is conceptualized in dynamic, intervention-rich environments.\u003c/p\u003e \u003cp\u003eIntervention reasoning emerges not as a residual category beyond diagnosis, but as a coordinated, action-oriented reasoning mode that dynamically interacts with diagnostic processes and is systematically shaped by expertise.\u003c/p\u003e \u003cp\u003eFuture research across diverse clinical contexts will be required to further test and elaborate this integrated framework.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Limitations and further research\u003c/h2\u003e \u003cp\u003eWhile the present findings provide initial empirical support for the CDI-R model, several limitations qualify their interpretation.\u003c/p\u003e \u003cp\u003eFirst, the VR environment itself may introduce confounds related to navigating a novel interface. Prior research suggests that familiarity with VR reduces cognitive load during complex tasks (Lee et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Bruyne et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2026\u003c/span\u003e). We did not assess participants\u0026rsquo; previous VR experience. This could have influenced observed differences in reasoning patterns. Also, the VR headsets limited access to facial cues and reduced co-presence compared to bedside practice. While the simulation was designed to authentically represent acute care interactions, the technologically mediated context may remain a constraint on validity.\u003c/p\u003e \u003cp\u003eSecond, this initial validation of the CDI-R model focused on expert- versus trainee-led teams. However, we did not systematically vary team composition or role distribution. A more controlled design would need to systematically vary group configurations\u0026ndash;e.g., mixed (dyadic, larger team) expertise levels, rotating leadership, prior team familiarity, interprofessional team composition\u0026ndash;to isolate their independent and interactive effects on collaborative reasoning patterns.\u003c/p\u003e \u003cp\u003eIn addition, in five out of nine expert-led conditions, non-leader roles were filled by research staff knowledgeable with ACLS procedures and familiar with the VR environment and performing chest compressions and/or airway management, whereas trainee-led teams consisted exclusively of residents. This structural difference in team composition may have influenced collaborative dynamics independently of leadership expertise. To mitigate potential bias, research staff participating in supporting roles were instructed to perform their tasks in a standardized manner, avoiding behaviors that could unduly advantage or disadvantage any particular study condition. Although the leadership role was the primary experimental manipulation, future research should employ fully comparable team compositions to isolate the effects of expertise more precisely.\u003c/p\u003e \u003cp\u003eThird, although the CDI-R model proposes general principles for collaborative diagnostic and intervention reasoning, we tested it within a single acute care context. Cardiac arrest has unique task characteristics, such as highly protocolized interventions (ACLS guidelines), compressed time frames, and relatively clear diagnostic decision points, that may not reflect the full range of acute care scenarios. Other contexts, such as multi-trauma or sepsis, involve greater diagnostic ambiguity, longer timeframes, and less standardized intervention pathways, which may substantially alter the relative distribution and sequencing of DAs and IAs. Beyond acute care, it would further be valuable to investigate whether similar collaborative reasoning dynamics emerge in other complex team-based decision-making domains, thereby testing the broader domain-generality of the CDI-R framework.\u003c/p\u003e \u003cp\u003eFourth, we analyzed teams\u0026rsquo; DAs, IAs, and CAs patterns but did not directly link these patterns to patient outcomes or performance quality. The simulation concluded at ROSC for all teams, but we did not measure time-to-ROSC, adherence to ACLS protocols, or omission and commission errors. Future research should examine whether the patterns proposed in the CDI-R model predict measurable differences in care quality, patient safety, or treatment efficiency.\u003c/p\u003e \u003cp\u003eFifth, our analysis segments communication into discrete utterance-level codes, but clinical reasoning may involve cognitive processes that unfold between utterances or remain covert/internalized. Silent coordination, parallel processing by team members, or internal decision-making before verbalization are not captured. Experts may engage in more efficient covert reasoning that reduces their need for certain types of verbal collaboration. Multimodal methods such as eye-tracking, physiological measures, freeze probe recall, or post-simulation subjective rating approaches could provide insight into these latent cognitive processes (Salmon et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSixth, it remains unclear to what extent the patterns observed in VR-based cardiac arrest management transfer to real clinical settings, which involve higher stakes, greater stress, and additional environmental constraints. Simulation-induced inquiry may partly inflate the observable frequency of IAs while also enabling controlled investigation of intervention reasoning. Future research should disentangle simulation-specific exploration from clinically driven intervention behavior.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e5.5 Practical Implications\u003c/h2\u003e \u003cp\u003eThe present findings have implications for how intervention processes are conceptually framed within medical education. In many acute care curricula, intervention procedures, task allocation, and structured communication are already central training components (e.g., ACLS and crisis resource management). However, these elements are often approached from a procedural or protocol-driven perspective. The present results suggest that coordinated intervention under time pressure can also be understood as a structured form of collaborative reasoning that unfolds in dynamic interplay with diagnostic reasoning, rather than as a discrete downstream step(s).\u003c/p\u003e \u003cp\u003eOur findings indicate that effective IA-oriented phases involve systematic task allocation, closed-loop communication, and deliberate monitoring, and that these processes are dynamically coupled with diagnostic formulation. Instructional approaches may therefore benefit from explicitly framing team-based intervention not only as execution of protocols but as a coordinated reasoning process that integrates hypothesis generation, action planning, and real-time evaluation.\u003c/p\u003e \u003cp\u003e In simulation-based training, debriefings could emphasize more process analysis in addition to outcome evaluation and guideline adherence. By mapping the co-occurrence and sequencing transitions between DA- and IA-oriented phases and how CAs are configured within each, the ENA visualizations reveal precisely where and how reasoning transitions falter across expertise levels. These network structures can inform targeted feedback during post-simulation debriefs. For example, analyses may pinpoint when trainee-led teams remain anchored in diagnostic loops rather than transitioning to intervention in a timely manner. Such process-oriented analysis may also help identify \u003cem\u003esilent failures\u003c/em\u003e, situations in which teams reach apparently successful outcomes (e.g., ROSC) while still demonstrating fragile reasoning patterns, such as missed intervention steps, errors of omission, or limited engagement in systematic cause evaluation (e.g., H\u0026rsquo;s and T\u0026rsquo;s). By revealing these patterns, ENA and LSA visualizations can support more targeted feedback during debriefings. However, realizing this potential requires adequate faculty development to equip instructors with the skills to interpret process-oriented analytics and translate them into effective feedback, particularly as prior work highlights the importance of aligning learning analytics tools with users\u0026rsquo; needs and supporting stakeholders in understanding and using such data in practice (Alfredo et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eUsing the CDI-R framework along with ENA visualizations as a reflective lens may help make these collaborative reasoning structures visible and discussable.\u003c/p\u003e \u003cp\u003eBeyond individual debriefs, the CDI-R framework may enable systematic comparison of collaborative clinical reasoning patterns across teams and scenarios. If applied longitudinally, it could allow educators to trace how collaborative clinical reasoning develops over the course of training.\u003c/p\u003e \u003cp\u003eTogether, these implications do not argue for more intervention training per se, but for a more explicit conceptualization of intervention reasoning as a collaborative, action-oriented reasoning mode within existing team-based training programs.\u003c/p\u003e \u003c/div\u003e"},{"header":"6 Conclusion","content":"\u003cp\u003eClinical reasoning research has predominantly emphasized diagnostic reasoning, while intervention reasoning has received comparatively limited conceptual and empirical attention. Yet, in complex, high-risk team-based situations, such as cardiac arrest, reasoning cannot be reduced to uncertainty reduction alone. It unfolds as a dynamic interplay between DAs, IAs and CAs under time pressure.\u003c/p\u003e \u003cp\u003eThis paper responds to the conceptual and empirical gap in research on intervention reasoning by introducing CDI-R as a theory-driven process model that systematically integrates DAs, IAs, and CAs within a unified framework. Using VR-based acute care simulations as an initial empirical instantiation, we demonstrate that intervention reasoning constitutes a structurally distinct yet dynamically intertwined mode of collaborative reasoning. Moreover, expertise shapes not only how often interventions are enacted but also how diagnostic and intervention processes are temporally orchestrated and structurally integrated.\u003c/p\u003e \u003cp\u003eBy turning clinical reasoning research toward intervention, this paper expands the theoretical scope of collaborative reasoning models beyond diagnosis-centered accounts. CDI-R positions intervention reasoning as a coordinated, action-oriented component of team-based clinical reasoning that interacts dynamically with diagnostic processes. While further research is required to systematically test the model across diverse clinical reasoning contexts and link reasoning process patterns to performance outcomes, the present findings provide a conceptual and analytic foundation for studying how teams turn evolving situation assessments into coordinated action.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was funded by the U.S. National Science Foundation under Grant Nos. 2202451 and 2506865. In addition, the work of Constanze Richters was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under TRR 419/1 \u0026ndash; 542251251 and FOR 2385; FI 792/11-2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no competing interests to declare.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study protocol was approved by the Institutional Review Board (HUM00188482, HUM00193383).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate\u003cbr\u003e\u003c/strong\u003eInformed consent was obtained from all individual participants included in the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDue to the privacy of the video and sensitive nature of the simulation training, trainees were assured raw data would remain confidential and would not be shared.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to thank Dr. Michael Cole and Dr. James Cooke for sharing their expertise on ACLS, VR, and clinical simulation. We also thank Benjamin Root, Cami Trendy, Umair Syed, Morgan Carpenter, and Nikolas Grotewold for their data annotation efforts.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConstanze Richters:\u0026nbsp;\u003c/strong\u003eConceptualization; Methodology; Visualization; Validation; Project administration; Writing \u0026ndash; original draft; Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eKapotaksha Das:\u003c/strong\u003e Methodology; Software; Formal analysis; Visualization; Data curation; Writing \u0026ndash; original draft; Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFrank Fischer:\u0026nbsp;\u003c/strong\u003eFunding acquisition; Conceptualization; Supervision; Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMartin R. Fischer:\u003c/strong\u003e Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMatthias Stadler:\u003c/strong\u003e Funding acquisition; Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVitaliy Popov:\u0026nbsp;\u003c/strong\u003eConceptualization; Methodology; Formal analysis; Investigation; Data curation; Resources; Supervision; Funding acquisition; Writing \u0026ndash; original draft; Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbdoler, E. A., Parsons, A. S., \u0026amp; Wijesekera, T. P. (2023). The future of teaching management reasoning: Important questions and potential solutions. \u003cem\u003eDiagnosis\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1), 19\u0026ndash;23. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1515/dx-2022-0048\u003c/span\u003e\u003cspan address=\"10.1515/dx-2022-0048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlfredo, R., Milesi, M., Echeverria, V., Gašević, D., Buckingham Shum, S., Zhao, L., Yan, L., Jin, Y., Fan, J. X., Pammer-Schindler, V., Swiecki, Z., \u0026amp; Martinez-Maldonado, R. (2025). Co-designing AI-powered learning analytics: Bringing students and teachers together. \u003cem\u003eInternational Journal of Educational Technology in Higher Education\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(1), 78. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41239-025-00572-8\u003c/span\u003e\u003cspan address=\"10.1186/s41239-025-00572-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBakeman, R., \u0026amp; Quera, V. (2011). \u003cem\u003eSequential analysis and observational methods for the behavioral sciences\u003c/em\u003e. Cambridge University Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrandl, L., Stadler, M., Richters, C., Radkowitsch, A., Fischer, M. R., Schmidmaier, R., \u0026amp; Fischer, F. (2024). Collaborative Problem-Solving in Knowledge-Rich Domains: A Multi-Study Structural Equation Model. \u003cem\u003eInternational Journal of Computer Supported Collaborative Learning\u003c/em\u003e 19, 341\u0026ndash;368 (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11412-024-09425-4\u003c/span\u003e\u003cspan address=\"10.1007/s11412-024-09425-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoshuizen, H. P. A., \u0026amp; Schmidt, H. G. (1992). On the Role of Biomedical Knowledge in Clinical Reasoning by Experts, Intermediates and Novices. \u003cem\u003eCognitive Science\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(2), 153\u0026ndash;184. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1207/s15516709cog1602_1\u003c/span\u003e\u003cspan address=\"10.1207/s15516709cog1602_1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCharlin, B., Boshuizen, H. P. A., Custers, E. J., \u0026amp; Feltovich, P. J. (2007). Scripts and clinical reasoning: Clinical expertise. \u003cem\u003eMedical Education\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(12), 1178\u0026ndash;1184. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1365-2923.2007.02924.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1365-2923.2007.02924.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCook, D. A., Durning, S. J., Sherbino, J., \u0026amp; Gruppen, L. D. (2019). Management Reasoning: Implications for Health Professions Educators and a Research Agenda. \u003cem\u003eAcademic Medicine\u003c/em\u003e, \u003cem\u003e94\u003c/em\u003e(9), 1310\u0026ndash;1316. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/ACM.0000000000002768\u003c/span\u003e\u003cspan address=\"10.1097/ACM.0000000000002768\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Bruyne, J., Larmuseau, C., De Marez, L., Talsma, D., \u0026amp; Bombeke, K. (2026). Timing matters! Using delayed signaling to improve experiential learning in procedural VR training. \u003cem\u003eComputers \u0026amp; Education\u003c/em\u003e, 105553. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compedu.2025.105553\u003c/span\u003e\u003cspan address=\"10.1016/j.compedu.2025.105553\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDepartment of Defense, \u0026amp; Agency for Healthcare Research and Quality. (2010). \u003cem\u003eTeamSTEPPS\u0026reg;: Strategies and tools to enhance performance and patient safety\u003c/em\u003e. [Manual]. Department of Defense, TRICARE; Agency for Healthcare Research and Quality.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuong, Q. H., Pham, T. N., Reynolds, L., Yeap, Y., Walker, S., \u0026amp; Lyons, K. (2023). A scoping review of therapeutic reasoning process research. \u003cem\u003eAdvances in Health Sciences Education\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(4), 1289\u0026ndash;1310. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10459-022-10187-7\u003c/span\u003e\u003cspan address=\"10.1007/s10459-022-10187-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGruppen, L. (2017). Clinical Reasoning: De ning It, Teaching It, Assessing It, Studying It. \u003cem\u003eWestern Journal of Emergency Medicine\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(1), 4\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5811/westjem.2016.11.33191\u003c/span\u003e\u003cspan address=\"10.5811/westjem.2016.11.33191\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEricsson, K. A. (Ed.). (1991). \u003cem\u003eToward a general theory of expertise: Prospects and limits\u003c/em\u003e. Cambridge University Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFink, M. C., Reitmeier, V., Stadler, M., Siebeck, M., Fischer, F., \u0026amp; Fischer, M. R. (2021). Assessment of Diagnostic Competences With Standardized Patients Versus Virtual Patients: Experimental Study in the Context of History Taking. \u003cem\u003eJournal of Medical Internet Research\u003c/em\u003e, \u003cem\u003e23\u003c/em\u003e(3), e21196. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2196/21196\u003c/span\u003e\u003cspan address=\"10.2196/21196\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFischer, F., Kollar, I., Stegmann, K., \u0026amp; Wecker, C. (2013). Toward a script theory of guidance in computer-supported collaborative learning. \u003cem\u003eEducational Psychologist\u003c/em\u003e, \u003cem\u003e48\u003c/em\u003e(1), 56\u0026ndash;66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/00461520.2012.748005\u003c/span\u003e\u003cspan address=\"10.1080/00461520.2012.748005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFischer, F., Kollar, I., Ufer, S., Sodian, B., Hussmann, H., Pekrun, R., Neuhaus, B., Dorner, B., Pankofer, S., Fischer, M. R., Strijbos, J. W., Heene, M., \u0026amp; Eberle, J. (2014). Scientific Reasoning and Argumentation: Advancing an Interdisciplinary Research Agenda in Education. \u003cem\u003eFrontline Learning Research\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(3), 28\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.14786/flr.v2i2.96\u003c/span\u003e\u003cspan address=\"10.14786/flr.v2i2.96\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHartjes, M. G., Richir, M. C., Cazaubon, Y., Donker, E. M., Van Leeuwen, E., Likic, R., Pers, Y. M., Pi\u0026euml;t, J. D., De Ponti, F., Raasch, W., Van Rosse, F., Rychl\u0026iacute;ckov\u0026aacute;, J., Sanz, E. J., Schwaninger, M., Wallerstedt, S. M., De Vries, T. P. G. M., Van Agtmael, M. A., Tichelaar, J., \u0026amp; On behalf of the EACPT Education Working group. (2024). Enhancing therapeutic reasoning: Key insights and recommendations for education in prescribing. \u003cem\u003eBMC Medical Education\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(1), 1360. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12909-024-06310-4\u003c/span\u003e\u003cspan address=\"10.1186/s12909-024-06310-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeitzmann, N., Seidel, T., Opitz, A., Hetmanek, A., Wecker, C., Fischer, M. R., Ufer, S., Schmidmaier, R., Neuhaus, B., Siebeck, M., St\u0026uuml;rmer, K., Obersteiner, A., Reiss, K., Girwidz, R., \u0026amp; Fischer, F. (2019). Facilitating Diagnostic Competences in Simulations in Higher Education: A Framework and a Research Agenda. \u003cem\u003eFrontline Learning Research\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(4), 1\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.14786/flr.v7i4.384\u003c/span\u003e\u003cspan address=\"10.14786/flr.v7i4.384\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHetmanek, A., Engelmann, K., Opitz, A., \u0026amp; Fischer, F. (2018). Beyond Intelligence and Domain Knowledge. In F. Fischer, C. A. Chinn, K. Engelmann, \u0026amp; J. Osborne (Eds.), \u003cem\u003eScientific Reasoning and Argumentation\u003c/em\u003e (1st ed., pp. 203\u0026ndash;226). Routledge. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4324/9780203731826-12\u003c/span\u003e\u003cspan address=\"10.4324/9780203731826-12\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJanssen, J., \u0026amp; Kirschner, P. A. (2020). Applying collaborative cognitive load theory to computer-supported collaborative learning: Towards a research agenda. \u003cem\u003eEducational Technology Research and Development\u003c/em\u003e, \u003cem\u003e68\u003c/em\u003e(2), 783\u0026ndash;805. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11423-019-09729-5\u003c/span\u003e\u003cspan address=\"10.1007/s11423-019-09729-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality. \u003cem\u003eAmerican Psychologist\u003c/em\u003e, \u003cem\u003e58\u003c/em\u003e(9), 697\u0026ndash;720. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/0003-066X.58.9.697\u003c/span\u003e\u003cspan address=\"10.1037/0003-066X.58.9.697\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKentros, M., Kamarianakis, M., Cole, M., Popov, V., Protopsaltis, A., \u0026amp; Papagiannakis, G. (2025). Non-linear, Team-based VR Training for Cardiac Arrest Care with enhanced CRM Toolkit. \u003cem\u003eEurographics 2025 - Short Papers\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2312/EGS.20251054\u003c/span\u003e\u003cspan address=\"10.2312/EGS.20251054\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKumar, D., \u0026amp; Basheer, A. (2021). Does management reasoning constitute the backbone of the clinical learning environment? Conceptual analysis of the existing notions. \u003cem\u003eJournal of Advances in Medical Education \u0026amp; Professionalism\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1), 54\u0026ndash;58. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.30476/jamp.2020.84431.1138\u003c/span\u003e\u003cspan address=\"10.30476/jamp.2020.84431.1138\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, J. Y., de Jong, N., Donkers, J., Jarodzka, H., \u0026amp; van Merri\u0026euml;nboer, J. J. (2023). Measuring cognitive load in virtual reality training via pupillometry. \u003cem\u003eIEEE Transactions on Learning Technologies\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e, 704\u0026ndash;710. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TLT.2023.3326473\u003c/span\u003e\u003cspan address=\"10.1109/TLT.2023.3326473\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiaw, S. Y., Chen, F. G., Klainin, P., Brammer, J., O\u0026rsquo;Brien, A., \u0026amp; Samarasekera, D. D. (2010). Developing clinical competency in crisis event management: An integrated simulation problem-based learning activity. \u003cem\u003eAdvances in Health Sciences Education\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(3), 403\u0026ndash;413. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10459-009-9208-9\u003c/span\u003e\u003cspan address=\"10.1007/s10459-009-9208-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, L., Hao, J., von Davier, A. A., Kyllonen, P., \u0026amp; Zapata-Rivera, D. (2016). A tough nut to crack: Measuring collaborative problem solving. In Y. Rosen, S. Ferrara, \u0026amp; M. Mosharraf (Eds.), \u003cem\u003eHandbook of Research on Technology Tools for Real-World Skill Development\u003c/em\u003e (pp. 344\u0026ndash;359). IGI Global. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/978-1-4666-9441-5\u003c/span\u003e\u003cspan address=\"10.4018/978-1-4666-9441-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeier, A., Spada, H., \u0026amp; Rummel, N. (2007). A rating scheme for assessing the quality of computer-supported collaboration processes. \u003cem\u003eInternational Journal of Computer-Supported Collaborative Learning\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 63\u0026ndash;86. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11412-006-9005-x\u003c/span\u003e\u003cspan address=\"10.1007/s11412-006-9005-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMonajemi, A., \u0026amp; Rikers, R. M. J. P. (2011). The role of patient management in medical expertise development: Extending the contemporary theory. \u003cem\u003eThe International Journal of Person Centered Medicine\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(1), 161\u0026ndash;166. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5750/ijpcm.v1i1.42\u003c/span\u003e\u003cspan address=\"10.5750/ijpcm.v1i1.42\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMonajemi, A., Rostami, E., Savaj, S., \u0026amp; Rikers, R. (2012). How does Patient Management Knowledge Integrate into an Illness Script? \u003cem\u003eEducation for Health\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(3), 153. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4103/1357-6283.109791\u003c/span\u003e\u003cspan address=\"10.4103/1357-6283.109791\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNorman, G., Pelaccia, T., Wyer, P., \u0026amp; Sherbino, J. (2024). Dual process models of clinical reasoning: The central role of knowledge in diagnostic expertise. \u003cem\u003eJournal of Evaluation in Clinical Practice\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(5), 788\u0026ndash;796. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/jep.13998\u003c/span\u003e\u003cspan address=\"10.1111/jep.13998\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePenner, J. C., Minter, D. J., Abdoler, E. A., \u0026amp; Parsons, A. S. (2025). Reasoning on Rounds 2: A Framework for Teaching Management Reasoning in the Inpatient Setting. \u003cem\u003eJournal of General Internal Medicine\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e(6), 1424\u0026ndash;1429. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11606-024-09039-4\u003c/span\u003e\u003cspan address=\"10.1007/s11606-024-09039-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePopov, V., Harmer, B., Raphael, S., Scott, I., Sample, A. P., Cooke, J. M., \u0026amp; Cole, M. (2025). Elucidating cognitive processes in cardiac arrest team leaders: A virtual reality-based cued-recall study of experts and novices. \u003cem\u003eAnnals of Medicine\u003c/em\u003e, \u003cem\u003e57\u003c/em\u003e(1), 2470976. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/07853890.2025.2470976\u003c/span\u003e\u003cspan address=\"10.1080/07853890.2025.2470976\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePopov, V., Li, Y., Cooke, J., Sample, A., \u0026amp; Cole, M. (2023). \u003cem\u003eTowards Supporting Technical and Non-Technical Skills Development by Using Multimodal Debriefing System After Multi-User VR-Based Simulation Training\u003c/em\u003e. 368\u0026ndash;369. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.22318/cscl2023.644173\u003c/span\u003e\u003cspan address=\"10.22318/cscl2023.644173\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadkowitsch, A., Sailer, M., Fischer, M. R., Schmidmaier, R., \u0026amp; Fischer, F. (2022). Diagnosing Collaboratively: A Theoretical Model and a Simulation-Based Learning Environment. In F. Fischer \u0026amp; A. Opitz (Eds.), \u003cem\u003eLearning to Diagnose with Simulations\u003c/em\u003e (pp. 123\u0026ndash;141). Springer International Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-89147-3_10\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-89147-3_10\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReddy, D. R. S., \u0026amp; Hanmandlu, A. (2022). The 4 H\u0026rsquo;s and T\u0026rsquo;s: How reliable is this mnemonic in classifying etiologies of in-hospital cardiac arrests? \u003cem\u003eResuscitation\u003c/em\u003e, \u003cem\u003e175\u003c/em\u003e, 3\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.resuscitation.2022.03.035\u003c/span\u003e\u003cspan address=\"10.1016/j.resuscitation.2022.03.035\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichters, C., Schmidmaier, R., Popov, V., Schredelseker, J., Fischer, F., \u0026amp; Fischer, M. R. (2024). Intervention skills \u0026ndash; a neglected field of research in medical education and beyond. \u003cem\u003eGMS Journal for Medical Education\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(4). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3205/ZMA001703\u003c/span\u003e\u003cspan address=\"10.3205/ZMA001703\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichters, C., Stadler, M., Radkowitsch, A., Schmidmaier, R., Fischer, M. R., \u0026amp; Fischer, F. (2023). Who is on the right track? Behavior-based prediction of diagnostic success in a collaborative diagnostic reasoning simulation. \u003cem\u003eLarge-Scale Assessments in Education\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(1), 3. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40536-023-00151-1\u003c/span\u003e\u003cspan address=\"10.1186/s40536-023-00151-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosen, M. A., DiazGranados, D., Dietz, A. S., Benishek, L. E., Thompson, D., Pronovost, P. J., \u0026amp; Weaver, S. J. (2018). Teamwork in healthcare: Key discoveries enabling safer, high-quality care. \u003cem\u003eAmerican Psychologist\u003c/em\u003e, \u003cem\u003e73\u003c/em\u003e(4), 433\u0026ndash;450. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/amp0000298\u003c/span\u003e\u003cspan address=\"10.1037/amp0000298\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSackett, G. P. (1979). The lag sequential analysis of contingency and cyclicity in behavioral interaction research. \u003cem\u003eHandbook of infant development\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e, 623\u0026ndash;649.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalas, E., Sims, D. E., \u0026amp; Burke, C. S. (2005). Is there a Big Five in Teamwork? \u003cem\u003eSmall Group Research\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(5), 555\u0026ndash;599. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1046496405277134\u003c/span\u003e\u003cspan address=\"10.1177/1046496405277134\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalas, E., Wilson, K. A., Murphy, C. E., King, H., \u0026amp; Salisbury, M. (2008). Communicating, Coordinating, and Cooperating When Lives Depend on It: Tips for Teamwork. \u003cem\u003eThe Joint Commission Journal on Quality and Patient Safety\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(6), 333\u0026ndash;341. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1553-7250(08)34042-2\u003c/span\u003e\u003cspan address=\"10.1016/S1553-7250(08)34042-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalmon, P. M., Stanton, N. A., Walker, G. H., Jenkins, D., Ladva, D., Rafferty, L., \u0026amp; Young, M. (2009). Measuring Situation Awareness in complex systems: Comparison of measures study. \u003cem\u003eInternational journal of industrial ergonomics\u003c/em\u003e, \u003cem\u003e39\u003c/em\u003e(3), 490\u0026ndash;500. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ergon.2008.10.010\u003c/span\u003e\u003cspan address=\"10.1016/j.ergon.2008.10.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShaffer, D. W., Collier, W., \u0026amp; Ruis, A. R. (2016). A Tutorial on Epistemic Network Analysis: Analyzing the Structure of Connections in Cognitive, Social, and Interaction Data. \u003cem\u003eJournal of Learning Analytics\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(3), 9\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18608/jla.2016.33.3\u003c/span\u003e\u003cspan address=\"10.18608/jla.2016.33.3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimon, H. A., \u0026amp; Newell, A. (1971). Human problem solving: The state of the theory in 1970. \u003cem\u003eAmerican Psychologist\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(2), 145\u0026ndash;159. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/h0030806\u003c/span\u003e\u003cspan address=\"10.1037/h0030806\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStadler, M., Niepel, C., \u0026amp; Greiff, S. (2019). Differentiating between static and complex problems: A theoretical framework and its empirical validation. \u003cem\u003eIntelligence\u003c/em\u003e, \u003cem\u003e72\u003c/em\u003e, 1\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.intell.2018.11.003\u003c/span\u003e\u003cspan address=\"10.1016/j.intell.2018.11.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTay, S. W., Ryan, P. M., \u0026amp; Ryan, C. A. (2016). Systems 1 and 2 thinking processes and cognitive reflection testing in medical students. \u003cem\u003eCanadian Medical Education Journal\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(2), e97\u0026ndash;103. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.36834/cmej.36777\u003c/span\u003e\u003cspan address=\"10.36834/cmej.36777\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilcox, G., \u0026amp; Heudes, A. (2017). Clinical Reasoning in the Assessment and Planning for Intervention for Oppositional Defiant Disorder. \u003cem\u003eCanadian Journal of School Psychology\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(1), 46\u0026ndash;58. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0829573516658996\u003c/span\u003e\u003cspan address=\"10.1177/0829573516658996\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWittenburg, P., Brugman, H., Russel, A., Klassmann, A., \u0026amp; Sloetjes, H. (2006). ELAN: A professional framework for multimodality research. In 5th international conference on language resources and evaluation (LREC 2006) (pp. 1556\u0026ndash;1559).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWitti, M. J., Zottmann, J. M., Wershofen, B., Thistlethwaite, J. E., Fischer, F., \u0026amp; Fischer, M. R. (2023). FINCA \u0026ndash; a conceptual framework to improve interprofessional collaboration in health education and care. \u003cem\u003eFrontiers in Medicine\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 1213300. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmed.2023.1213300\u003c/span\u003e\u003cspan address=\"10.3389/fmed.2023.1213300\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZ\u0026ouml;rgő, S., \u0026Aacute;rva, D., \u0026amp; Eagan, B. (2024, November). Making sense of the model: Interpreting Epistemic networks and their projection space. In International Conference on Quantitative Ethnography (pp. 119\u0026ndash;135). Cham: Springer Nature Switzerland.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"advances-in-health-sciences-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ahse","sideBox":"Learn more about [Advances in Health Sciences Education](http://link.springer.com/journal/10459)","snPcode":"10459","submissionUrl":"https://submission.nature.com/new-submission/10459/3","title":"Advances in Health Sciences Education","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Clinical reasoning, Intervention reasoning, Team-based reasoning, Acute care, Clinical expertise, Simulation-based training","lastPublishedDoi":"10.21203/rs.3.rs-9497414/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9497414/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eClinical reasoning in acute care unfolds under time pressure as teams must continuously interpret evolving patient information while coordinating treatment. While research has predominantly focused on diagnostic reasoning, it remains insufficiently understood how care teams generate, coordinate, and enact interventions, and how these processes are organized.\u003c/p\u003e \u003cp\u003eTo address this gap, we introduce the Collaborative Diagnostic\u0026ndash;Intervention Reasoning (CDI-R) model, a theory-driven process model that conceptualizes clinical reasoning as the interplay of diagnostic activities (DAs), intervention activities (IAs), and collaborative activities (CAs).\u003c/p\u003e \u003cp\u003eWe provide an initial empirical examination of the model using a virtual reality cardiac arrest simulation with 29 teams (\u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;116 participants). Team interactions were coded at the utterance level. Lag Sequential Analysis (LSA) was used to examine transitions between DAs and IAs, and Epistemic Network Analysis (ENA) was used to examine the structural organization of CAs.\u003c/p\u003e \u003cp\u003eFindings reveal three key patterns. First, DAs and IAs constitute structurally distinct reasoning modes, characterized by different configurations of CAs. Second, reasoning unfolds in non-linear and recurrent sequences, with sustained engagement within and transitions between reasoning modes. Third, expertise shapes the organization of reasoning: expert-led teams engaged in a higher proportion of IA-oriented activity, transitioned more frequently from diagnosis to intervention, and showed more differentiated collaborative structures, whereas trainee-led teams exhibited more loops within reasoning modes.\u003c/p\u003e \u003cp\u003eBy explicitly integrating intervention reasoning into models of clinical reasoning, this study advances a process-oriented account of team-based reasoning. The CDI-R model provides a framework for examining how teams coordinate diagnosis and intervention in dynamic clinical settings.\u003c/p\u003e","manuscriptTitle":"Turning clinical reasoning research toward intervention: Validating a process model using VR simulation data","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-12 10:33:15","doi":"10.21203/rs.3.rs-9497414/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"96214491631362824287885225074604896582","date":"2026-05-11T01:25:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"239698533491751345960655787300434437850","date":"2026-05-03T15:30:21+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-01T15:29:40+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-25T01:56:36+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-25T01:56:29+00:00","index":"","fulltext":""},{"type":"submitted","content":"Advances in Health Sciences Education","date":"2026-04-22T13:58:31+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"advances-in-health-sciences-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ahse","sideBox":"Learn more about [Advances in Health Sciences Education](http://link.springer.com/journal/10459)","snPcode":"10459","submissionUrl":"https://submission.nature.com/new-submission/10459/3","title":"Advances in Health Sciences Education","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"17f1aa9b-8f91-41e8-97b9-0bf8f443aa07","owner":[],"postedDate":"May 12th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"96214491631362824287885225074604896582","date":"2026-05-11T01:25:08+00:00","index":10,"fulltext":""},{"type":"reviewerAgreed","content":"239698533491751345960655787300434437850","date":"2026-05-03T15:30:21+00:00","index":9,"fulltext":""},{"type":"reviewersInvited","content":"3","date":"2026-05-01T15:29:40+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-12T10:33:18+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-12 10:33:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9497414","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9497414","identity":"rs-9497414","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}