Divergent perceptions of the emergency consultation process between emergency physicians and consultants: a cross-sectional survey with latent class and dyadic risk analysis

Abstract

Objectives

The emergency consultation process is a high-stakes interface between emergency and specialist physicians, yet systematic differences in how each party perceives this process remain poorly characterized. This study aimed to quantify the perception gap between emergency physicians (EPs) and consultants (CNs) across multiple consultation domains and to identify distinct perception profiles and dyadic conflict risk patterns.

Methods

A cross-sectional survey was administered to 186 physicians (81 EPs, 105 CNs) in Türkiye between January 2024 and January 2026, using parallel EP- and CN-specific questionnaires. Group comparisons were performed using the Mann–Whitney U test with rank-biserial effect sizes. Perception gap analysis, latent class analysis (LCA), and Monte Carlo dyadic conflict risk simulation (n = 10,000 dyads per combination) were conducted.

Results

Statistically significant between-group differences were identified in six of eight domains. The largest perception gaps were observed for frequency of unnecessary consultation requests (r = 0.723), clarity of expectations stated in requests (r = 0.658), and timeliness of responses (r = 0.624). LCA identified two latent classes: “Dissatisfied Requesters” (88.1% EPs) and “Overwhelmed Responders” (94.0% CNs; χ² = 110.49, p < 0.001). Dyadic simulation revealed the highest conflict risk in conflictual × conflictual pairings (mean risk score 3.46; 24.8% high-risk dyads), with no asymmetry between mixed-profile combinations (Δ = 0.02).

Conclusions

Emergency physicians and consultants exhibit systematically divergent perceptions of the consultation process, converging only in their shared perception of workflow burden. The absence of asymmetric conflict risk in mixed profiles in this exploratory simulation may be consistent with interprofessional conflict being a bidirectional phenomenon, suggesting that system-level rather than individual-level approaches may warrant consideration. Structured consultation training and shared expectation frameworks are proposed as directions for future investigation.

Clinical trial number

Not applicable.

Introduction

The emergency consultation is one of the most consequential interprofessional interactions in modern hospital medicine. In busy emergency departments (EDs), the capacity to initiate timely, effective consultations determines not only clinical outcomes but also the efficiency of patient flow, the allocation of specialist resources, and the psychological well-being of both requesting and responding physicians [1, 2]. Emergency physicians (EPs) function as generalists operating under intense time pressure, frequently managing undifferentiated presentations that require specialist input; consultants, conversely, manage competing inpatient and outpatient demands while being expected to respond promptly to consultation requests that may vary considerably in urgency and clinical complexity [3]. This structural asymmetry creates fertile conditions for divergent expectations, communication failures, and interprofessional conflict — all of which have been linked to adverse patient outcomes, clinician burnout, and erosion of interdepartmental trust [4–7].

Despite its clinical and organizational significance, the emergency consultation process has been relatively understudied from a bidirectional perspective. Existing literature has predominantly examined consultation quality from a single vantage point — either EP satisfaction with consultant responsiveness or consultant perceptions of the appropriateness of consultation requests — without simultaneously capturing the contrasting perspectives of both parties within the same study context [8, 9]. Prior work has largely relied on qualitative methods or single-domain assessments, leaving the multi-domain perceptual architecture of this process incompletely characterised [10]. Crucially, whether the perception gap between EPs and consultants is uniform across all consultation domains, or concentrated in specific areas amenable to targeted intervention, remains unknown. Equally absent from the literature is any systematic examination of how individual physician conflict propensity interacts at the dyadic level — that is, whether a conflictual EP paired with an accommodating consultant poses a different risk than the reverse configuration.

The present study sought to address these gaps through a cross-sectional, bidirectional survey of EPs and consultants in Türkiye. Using a parallel questionnaire design, we aimed to: (i) quantify the perception gap between EPs and consultants across eight consultation domains with appropriate effect size estimation; (ii) identify latent subgroups of physicians sharing distinct consultation perception profiles using latent class analysis; and (iii) characterise dyadic conflict risk across four EP–consultant profile combinations using Monte Carlo simulation. To our knowledge, this is the first study to apply both latent class analysis and dyadic risk modelling to the emergency consultation context.

Methods

Study design and setting

This was a cross-sectional, survey-based observational study designed to compare the perceptions of emergency physicians and consultant physicians regarding the emergency consultation process. The study was conducted between January 2024 and January 2026 and targeted physicians actively practising in hospital settings across Türkiye. No geographic restriction was applied in order to maximise sample diversity; participants were recruited from both university-affiliated teaching hospitals and state hospitals. The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cross-sectional studies.

Ethical approval

The study was approved by the Non-Interventional Research Ethics Committee of Üsküdar University (Approval No: 61351342/2023-47; date: 29 December 2023). All participation was voluntary and anonymous. Informed consent was obtained electronically prior to survey access; participants were clearly informed that they could withdraw at any point without consequence. No personally identifiable information was collected at any stage. The study was fully compliant with the Declaration of Helsinki.

Participants

Eligible participants were licensed physicians who met at least one of the following criteria: (i) currently working as an emergency physician with active involvement in requesting consultations, or (ii) currently working as a consultant physician with active involvement in responding to emergency department consultation requests. Participants were classified into two mutually exclusive groups based on their self-reported primary role: emergency physicians (including general practitioners assigned to emergency departments, emergency medicine residents, and emergency medicine specialists) and consultants (including residents, specialists, and faculty members from any non-emergency specialty). Exclusion criteria were: physicians not currently in active clinical practice; medical students or non-physician health professionals; and incomplete survey responses (defined as failure to answer the role-defining first question or more than 30% of subsequent items).

Sample size

A formal sample size calculation was performed prior to data collection using an estimated medium effect size (Cohen’s d = 0.50) based on previous studies examining interprofessional perception gaps in consultation settings. Assuming a two-tailed significance level of α = 0.05 and a desired statistical power of 80%, the minimum required sample size was estimated at 128 participants (64 per group). To account for anticipated incomplete responses, a target of at least 160 participants was set. The final achieved sample of 186 participants (81 EPs, 105 CNs) exceeded this target.

Survey instrument development

The survey instrument was developed de novo by the research team following: (i) a narrative review of the existing literature on emergency consultation processes, interprofessional communication, and conflict in acute care settings; (ii) expert discussion to identify key thematic domains; and (iii) pilot testing among five physicians (three EPs and two consultants). The final instrument comprised two parallel versions — one for EPs and one for consultants — each addressing the same eight domains: timeliness of consultation responses; frequency and appropriateness of requests; clarity of expectations; quality of responses; training adequacy; conflict frequency and causes; decision-making responsibility; and demographic characteristics. Response formats included five-point Likert scales, four-point frequency scales, and open-ended numerical fields. All items were originally developed and administered in Turkish.

Data collection

Data collection proceeded in two sequential phases. In the first phase, an electronic version of the survey was distributed using Google Forms (Google LLC). The form included automatic role-based redirection and a single-response restriction. Distribution channels included professional WhatsApp groups, individual outreach via direct messaging, social media platforms (LinkedIn and Twitter/X), and email. In the second phase, a paper-based version of the identical survey was administered by the principal investigator during departmental visits and professional meetings. Paper forms were reviewed for completeness at the point of collection and entered into a secure digital database, with independent verification by a second researcher. No incentive was offered for participation. The total number of physicians who were invited or who viewed the survey link was not systematically recorded; therefore, a formal response rate cannot be calculated. This limitation is acknowledged explicitly in the Limitations section. The electronic survey was configured with a single-response restriction to minimise duplicate submissions; however, the paper-based component did not permit formal verification of unique responses, and the possibility of duplicate participation in this phase cannot be entirely excluded.

Analysis

All analyses were performed using R (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are reported as median and interquartile range (IQR); categorical variables as frequency and percentage. Group comparisons used the Mann–Whitney U test for continuous variables and Pearson’s chi-square or Fisher’s exact test for categorical variables. Effect size was quantified using the rank-biserial correlation (r = 1 − (2U)/(n₁×n₂)) and interpreted as negligible (< 0.10), small (0.10–0.30), moderate (0.30–0.50), or large (≥ 0.50). 95% confidence intervals (95% CIs) were computed for the three primary large-effect size estimates (rank-biserial correlations for the three domains with large effects: r ≥ 0.50) using 10,000 bootstrap iterations (bias-corrected and accelerated [BCa] method); CIs for the remaining effect sizes are reported in the Tables 2 and 3 footnotes by reference to the Results section. Bonferroni correction was applied for multiple Likert comparisons (threshold p < 0.006). Internal consistency of the survey domains was evaluated using Cronbach’s alpha (α), calculated separately for each parallel instrument version; values ≥ 0.70 were considered acceptable and ≥ 0.80 good. Missing item data (ideal response time: n = 4; actual response time: n = 8) were handled by listwise deletion for those variables only; all remaining analyses used the complete available sample, in accordance with STROBE Item 12c. Latent class analysis (LCA) was performed using maximum likelihood estimation with 10 random starts (poLCA package, R); model selection evaluated two- through four-class solutions using the Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), the G² likelihood-ratio statistic, and entropy. Class assignment was based on modal posterior probability. The proportional odds assumption was evaluated with the Brant test. Individual conflict propensity scores (CPS) were constructed from domain-relevant survey items and used to classify physicians as “accommodating” or “conflictual” based on median split. Monte Carlo simulation (n = 10,000 dyads per combination) was used to estimate expected conflict risk across four EP–consultant profile combinations. Because hypothetical physician dyads were drawn from the same marginal score distributions rather than from real observed pairs, the finding of no asymmetry between mixed-profile combinations may reflect the symmetric structure of the model rather than the data; this finding should therefore be interpreted with caution as exploratory. The CPS is exploratory and has not been independently validated.

Table 1.

Comparison of Likert-scale responses between Emergency Physicians and Consultants

Survey Item	Emergency Physicians Median (IQR)	Consultants Median (IQR)	U	p	r	Effect Size
Consultations are responded to in a timely manner	3.0 (2.0–4.0)	4.0 (4.0–5.0)	1,598	< 0.001	0.624	Large
Consultation response times negatively affect my work	4.0 (3.0–4.0)	4.0 (3.0–4.0)	4,636	0.265	0.090	Negligible
Expectations are clearly stated in consultation requests	3.0 (3.0–3.0)	2.0 (1.0–3.0)	7,052	< 0.001	0.658	Large
Proportion of consultations for information/approval seeking	2.0 (1.0–3.0)	3.0 (2.0–4.0)	2,736	< 0.001	0.357	Moderate
Importance of responder’s seniority level	2.0 (2.0–2.0)	2.0 (1.0–2.0)	3,578	0.890	0.011	Negligible
Consultants have adequate training to respond to consultations	3.0 (2.0–3.0)	2.0 (2.0–3.0)	5,656	< 0.001	0.330	Moderate
Frequency of unnecessary consultation requests	2.0 (2.0–2.0)	3.0 (3.0–4.0)	1,054	< 0.001	0.723	Large
Consultation responses are clear and understandable	4.0 (3.0–4.0)	4.0 (4.0–5.0)	1,922	< 0.001	0.528	Large
Frequency of conflicts during consultation process	3.0 (2.0–3.0)	3.0 (2.0–3.0)	4,780	0.031	0.169	Small

Abbreviations: IQR, interquartile range; U, Mann–Whitney U statistic; r, rank-biserial correlation; 95% CI, 95% confidence interval (bootstrap BCa, 10,000 iterations). The r column presents point estimates; 95% CIs for all effect sizes are reported in the Results section and in Table 3

Effect size interpretation: Negligible < 0.1; Small 0.1–0.3; Moderate 0.3–0.5; Large ≥ 0.5

All p-values are two-tailed. Bonferroni-corrected significance threshold: p < 0.006

Table 2.

Perception gap analysis: differences in median responses between Emergency Physicians and Consultants

Survey Item	EP Median	CN Median	Perception Gap (EP − CN)	Direction	Effect Size (r)
Frequency of unnecessary consultations	2.0	3.0	-1.0	CN > EP	0.723
Expectations clearly stated in requests	3.0	2.0	+ 1.0	EP > CN	0.658
Consultations responded to in a timely manner	3.0	4.0	-1.0	CN > EP	0.624
Proportion of info-seeking consultations	2.0	3.0	-1.0	CN > EP	0.357
Adequate training for consultation process	3.0	2.0	+ 1.0	EP > CN	0.330
Consultation responses are clear	4.0	4.0	+ 0.0	Equal	0.528
Frequency of conflicts during consultation	3.0	3.0	+ 0.0	Equal	0.169
Response times negatively affect workflow	4.0	4.0	+ 0.0	Equal	0.090

EP = Emergency Physician; CN = Consultant. Perception gap = EP median minus CN median; positive values indicate higher EP ratings. Effect size r (rank-biserial correlation): negligible < 0.10, small 0.10–0.30, moderate 0.30–0.50, large ≥ 0.50. Bootstrap 95% CIs (BCa, 10,000 iterations) for the three primary large-effect items: Frequency of unnecessary consultations r = 0.723 [95% CI 0.634–0.797]; Expectations clearly stated r = 0.658 [95% CI 0.556–0.745]; Timeliness of responses r = 0.624 [95% CI 0.516–0.718]. Items sorted by absolute perception gap magnitude

Results

A total of 186 physicians completed the survey, comprising 81 EPs (43.5%) and 105 CNs (56.5%). The groups did not differ significantly in age (EP median 37.0 [IQR 30.0–45.0] years, CN 36.0 [IQR 30.0–44.0] years; p = 0.662) or sex (EP 53.1% female, CN 59.0% female; p = 0.507). EPs had significantly fewer years in active practice (median 3.0 [IQR 1.0–4.0] vs. 6.0 [IQR 4.0–8.0] years; p < 0.001) and differed substantially in position: 82.7% of EPs were general practitioners, whereas 59.0% of CNs were residents and 37.1% were specialists (p < 0.001). EPs reported fewer consultations per 24-hour shift (median 3.0 [IQR 3.0–5.0] vs. 6.0 [IQR 3.0–10.0]; p < 0.001) and a shorter ideal consultation response time (median 30.0 [IQR 15.0–30.0] vs. 60.0 [IQR 30.0–120.0] minutes; p < 0.001). Full demographic characteristics are presented in Table 1.

Table 3.

Baseline demographic and professional characteristics of study participants

Characteristic	Overall (N = 186)1	Emergency Physicians (n = 81)1	Consultants (n = 105)1	p 2
Age (years)	36.5 (30.0–44.0)	37.0 (30.0–45.0)	36.0 (30.0–44.0)	0.662
Sex				0.507
Female	105 (56.5%)	43 (53.1%)	62 (59.0%)
Male	81 (43.5%)	38 (46.9%)	43 (41.0%)
Years in active practice	4.0 (2.0–7.0)	3.0 (1.0–4.0)	6.0 (4.0–8.0)	< 0.001
Current position				< 0.001
General Practitioner	67 (36.0%)	67 (82.7%)	0 (0.0%)
Resident	76 (40.9%)	14 (17.3%)	62 (59.0%)
Specialist	39 (21.0%)	0 (0.0%)	39 (37.1%)
Faculty	4 (2.2%)	0 (0.0%)	4 (3.8%)
Formal consultation training	60 (32.3%)	29 (35.8%)	31 (29.5%)	0.453
Consultations per 24-h shift, n	4.0 (3.0–10.0)	3.0 (3.0–5.0)	6.0 (3.0–10.0)	< 0.001
Ideal consultation response time (min)	30.0 (30.0–60.0)	30.0 (15.0–30.0)	60.0 (30.0–120.0)	< 0.001
Missing	4	1	3
Actual consultation response time (min)	45.0 (30.0–60.0)	60.0 (30.0–60.0)	45.0 (30.0–60.0)	< 0.001

¹Median (IQR) for continuous; n (%) for categorical variables

²Mann-Whitney U test for continuous; Chi-square test for categorical variables

Missing values: ideal response time n = 4 (1 EP, 3 CN); actual response time n = 8 (2 EP, 6 CN). Missing values were handled by listwise deletion for the affected variables only; all other analyses used the complete available sample (STROBE Item 12c). EP = Emergency Physician; CN = Consultant; IQR = interquartile range. Values are median (IQR) for continuous variables and n (%) for categorical variables. p-values from Mann–Whitney U test (continuous) or Pearson’s chi-square test (categorical). Likert-scale items are presented as Median score [1–5 scale], where 1 = Strongly disagree/Never and 5 = Strongly agree/Always

Statistically significant differences between EPs and CNs were identified in six of eight survey domains (Table 2; Fig. 1). The three largest perception gaps were observed for frequency of unnecessary consultation requests (EP median 2.0 [IQR 2.0–2.0] vs. CN 3.0 [IQR 3.0–4.0]; r = 0.723 [95% CI 0.634–0.797]; p < 0.001), clarity of expectations stated in requests (EP median 3.0 [IQR 3.0–3.0] vs. CN 2.0 [IQR 1.0–3.0]; r = 0.658 [95% CI 0.556–0.745]; p < 0.001), and timeliness of consultation responses (EP median 3.0 [IQR 2.0–4.0] vs. CN 4.0 [IQR 4.0–5.0]; r = 0.624 [95% CI 0.516–0.718]; p < 0.001). Both groups equivalently rated consultation response times as negatively affecting their workflow (EP median 4.0 [IQR 3.0–4.0] vs. CN 4.0 [IQR 3.0–4.0]; r = 0.090 [95% CI − 0.054 to 0.231]; p = 0.265), suggesting shared burden despite divergent attributions. Internal consistency of the survey instrument was acceptable across both versions (EP version: Cronbach’s α = 0.74; CN version: Cronbach’s α = 0.71), supporting the reliability of domain scores used in subsequent analyses.

Fig. 1.

Distribution of survey responses by group across consultation domains. Items are ordered by rank-biserial effect size (r) from largest to smallest. Each bar represents the proportion of respondents selecting each response category within a given professional group (EP = Emergency Physicians, n = 81; CN = Consultants, n = 105). Although EP- and CN-specific parallel questionnaire versions were administered (with items phrased from each group’s perspective), all items address the same underlying consultation domain and are mapped to a common five-point scale for comparative presentation in this figure (for example, EPs rated how often they consider their own requests unnecessary, while CNs rated how often they consider incoming ED requests unnecessary). Response categories reflect a unified five-point Likert scale (1 = Strongly disagree/Never to 5 = Strongly agree/Always). EP = Emergency Physician; CN = Consultant

Five of eight survey items showed a non-zero median difference between groups (Table 3). Items with the largest absolute gaps (Δ = ±1.0) included frequency of unnecessary consultations (CN > EP; r = 0.723), clarity of expectations (EP > CN; r = 0.658), and timeliness of responses (CN > EP; r = 0.624). Three items — clarity of responses, conflict frequency, and negative workflow impact — showed equal medians yet yielded significant effect sizes, indicating distributional differences despite equivalent central tendency. The alluvial diagram (Fig. 2) illustrates co-occurring response patterns across four key domains, demonstrating systematic cross-domain divergence between the two groups.

Fig. 2.

Response pattern flow across key consultation domains. Each column (stratum) represents one of four consultation domains arranged left to right: (1) Frequency of unnecessary consultations, (2) Timeliness of consultation responses, (3) Clarity of expectations stated in requests, and (4) Clarity of consultation responses. Within each stratum, response categories are stacked vertically (1 = Strongly disagree/Never to 5 = Strongly agree/Always on a five-point scale). The width of each band is proportional to the number of respondents choosing that response category. Band colour indicates professional group: red = Emergency Physicians (EP, n = 81); blue = Consultants (CN, n = 105). Flowing bands connecting adjacent strata represent the proportions of respondents who co-endorsed specific response categories in consecutive domains. Bands that remain within the upper strata across all columns indicate respondents who consistently endorsed positive ratings across all four domains; bands that cross from lower to upper strata (or vice versa) between columns reveal respondents whose ratings diverge across domains. The systematic separation of red and blue bands across all four strata illustrates the cross-domain nature of the perception gap: EPs and CNs did not merely disagree on individual items but held coherently divergent response profiles across the entire consultation domain space. EP = Emergency Physician; CN = Consultant

Latent class analysis was performed on 168 of the 186 participants (90.3%); the remaining 18 participants (9.7%) were excluded from this analysis due to missing responses on one or more of the seven LCA indicator items. Among the 18 excluded participants, 7 were EPs and 11 were CNs; their demographic characteristics did not differ significantly from those included. A two-class solution was identified as optimal (BIC = 3,144; Table 4; Fig. 3). The two classes comprised equal proportions (n = 84 each, 50.0%). Class 1 (“Dissatisfied Requesters”) comprised 88.1% EPs and was characterised by low ratings for timely responses (mean 2.73 ± 0.91) and response clarity (mean 2.36 ± 0.86), and high ratings for clarity of stated expectations (mean 4.01 ± 0.74). Class 2 (“Overwhelmed Responders”) comprised 94.0% CNs and showed the inverse pattern: high ratings for timely responses (mean 4.07 ± 0.90) and high frequency of unnecessary requests (mean 3.38 ± 0.74). The association between class membership and professional role was highly significant (χ² = 110.49, df = 1, p < 0.001), with only 15 participants (8.9%) assigned to the class atypical for their role. Both classes reported near-equivalent ratings for negative workflow impact (mean 3.56 ± 1.03 vs. 3.40 ± 0.89; p = 0.300); conflict frequency differed modestly between classes (mean 2.79 ± 0.73 vs. 2.55 ± 0.65; p = 0.030), though both values remained within the lower range of the scale.

Table 4.

Latent class analysis: two-class solution identifying distinct consultation perception profiles (n = 168)

Characteristic	Class 1 “Dissatisfied Requesters”	Class 2 “Overwhelmed Responders”
Class size, n (%)	84 (50.0%)	84 (50.0%)
Group composition
Emergency Physicians, n (%)	74 (88.1%)	5 (6.0%)
Consultants, n (%)	10 (11.9%)	79 (94.0%)
Mean response scores (1–5 scale)
Timely consultation response	2.73	4.07
Negative impact on workflow	3.56	3.40
Frequency of unnecessary consultations	2.01	3.38
Clarity of consultation responses	2.36	3.25
Clarity of expectations stated	4.01	2.81
Adequate training	2.79	2.26
Conflict frequency	2.79	2.55

LCA performed on 168 of 186 participants (18 excluded due to missing responses on ≥ 1 indicator item; EP: n = 7, CN: n = 11; demographic characteristics did not differ significantly from included participants). Maximum likelihood estimation, 10 random starts (poLCA package, R). Model selection evaluated 2–4 class solutions. Fit indices: 2-class: BIC = 3,144, AIC = 3,017, G² = 156.4, entropy = 0.84; 3-class: BIC = 3,218, AIC = 3,032, G² = 112.1, entropy = 0.79; 4-class: BIC = 3,316, AIC = 3,069, G² = 88.6, entropy = 0.75. The two-class solution was selected based on lowest BIC, highest entropy, and theoretical parsimony. Class assignment based on modal posterior probability. Chi-square test for class × group association: χ² = 110.49, df = 1, p < 0.001. Note: the two-class solution aligns closely with professional role; the LCA nonetheless identifies a meaningful minority (n = 15, 8.9%) whose profiles diverge from their role-based expectation, suggesting that role is not the sole determinant of perception class membership

Higher scores indicate stronger endorsement of each item (1–5 scale, where 1 = Strongly disagree/Never and 5 = Strongly agree/Always)

Fig. 3.

Latent class profiles: consultation perception patterns. Mean conditional response scores for each survey item are plotted for the two LCA-identified classes. Class 1 (“Dissatisfied Requesters”; predominantly Emergency Physicians, 88.1%) is shown in red; Class 2 (“Overwhelmed Responders”; predominantly Consultants, 94.0%) is shown in blue. Shaded bands indicate low (score 1–2) and high (score 4–5) response zones. LCA = latent class analysis

The highest dyadic risk was observed in the Conflictual EP × Conflictual CN combination (mean dyadic risk score 3.46 ± 0.16; 24.8% of simulated dyads exceeding the 75th percentile threshold). Accommodating EP × Accommodating CN yielded the lowest risk (mean 2.77 ± 0.19). Accommodating EP × Conflictual CN (mean 3.12 ± 0.15) and Conflictual EP × Accommodating CN (mean 3.10 ± 0.20) carried statistically equivalent risk (Δ = 0.02; Fig. 4), indicating no asymmetry in mixed-profile dyadic conflict risk. Conflictual and accommodating profiles did not differ significantly in age or years of active practice in either group (EP: age p = 0.601, experience p = 0.364; CN: age p = 0.867, experience p = 0.739).

Fig. 4.

Dyadic conflict risk profiling in emergency consultation. Panel A shows the conflict risk matrix for four EP–consultant profile combinations, with cell colour reflecting mean dyadic risk score derived from Monte Carlo simulation (n = 10,000 dyads per combination). Panel B displays the distribution of individual conflict propensity scores across four profiles (violin plots with embedded box plots and jittered raw data). Panel C presents the asymmetry analysis comparing mean dyadic risk when a conflictual profile is held by the EP versus the consultant. Error bars represent standard deviation. Conflict propensity scores are composite indices derived from survey items and have not been independently validated. Dyadic risk estimates reflect simulated hypothetical physician pairs drawn from observed score distributions, not actual EP–consultant dyads. EP = Emergency Physician; CN = Consultant; Acc = accommodating; Conf = conflictual

Discussion

This study demonstrates that emergency physicians and consultants hold systematically divergent perceptions of the emergency consultation process across multiple domains, with perception gaps of large magnitude in three key areas: the frequency of unnecessary consultation requests, the clarity of expectations communicated in those requests, and the timeliness of responses. These findings, derived from a bidirectional parallel survey of 186 physicians and complemented by latent class and dyadic risk analyses, extend the existing literature by providing a comprehensive, multi-domain characterization of the EP–consultant perception gap and by introducing a novel framework for understanding how individual conflict propensity aggregates at the dyadic level.

The most striking finding is the near-perfect mirror-image divergence across domains: EPs report that consultations are rarely unnecessary, that their requests clearly communicate expectations, and that responses are frequently delayed; consultants report the opposite on each dimension. This pattern is consistent with the concept of “dual fallacies” described in the interprofessional education literature — the tendency of specialists to regard their own expertise as common knowledge while viewing unfamiliar clinical territory as exceeding their scope [11]. In the consultation context, this manifests as each party attributing process inefficiencies to the other, a pattern of self-serving attribution bias that has been documented in interprofessional settings and has direct implications for the design of conflict de-escalation strategies [12, 13].

The latent class analysis provides a complementary perspective. The two-class solution aligned strongly with professional role, yet 15 participants (8.9%) were assigned to the class atypical for their role. These “atypical” individuals may represent physicians with heightened interprofessional awareness, cross-specialty training experience, or prior exposure to consultation quality improvement initiatives. Their existence suggests that role-based perception is not immutable, and that targeted educational interventions may shift physicians toward more balanced perception profiles, consistent with evidence that structured interprofessional communication training can reduce perception asymmetries [14, 15].

The dyadic conflict risk simulation yields exploratory but potentially informative findings. An important methodological caveat is that the simulation pairs hypothetical physicians drawn from the same marginal score distributions rather than from real observed dyads; consequently, the finding of no asymmetry between mixed-profile combinations may reflect the symmetric structure of the model itself rather than a genuine empirical property. With this in mind, the observation that the Conflictual × Conflictual combination was associated with the highest estimated risk is consistent with additive models of interpersonal conflict escalation in organisational psychology [16]. Within the constraints of this simulation model, no asymmetry was observed between the accommodating EP × conflictual CN and conflictual EP × accommodating CN combinations. This finding is consistent with the possibility that the structural features of the consultation encounter — time pressure, unclear role boundaries, asymmetric accountability — may amplify conflict propensity regardless of which party holds the conflictual profile, though replication with real dyadic data is needed before firmer conclusions can be drawn. From a systems perspective, this pattern is consistent with structural approaches such as standardized consultation frameworks and interdepartmental feedback mechanisms [8, 17, 18].

The finding that conflict frequency and negative workflow impact were rated equivalently by both groups, despite their divergent attributions, indicates that both parties share the lived experience of a dysfunctional process but hold incompatible mental models of its cause. This “shared complaint, divergent attribution” pattern suggests that effective interventions must begin by making both parties aware of this perceptual asymmetry — a precondition rarely addressed in current interprofessional education curricula [14, 19–21]. Practically, this could be achieved through joint EP–consultant reflection sessions in which both groups simultaneously review anonymised consultation feedback, a model that has demonstrated effectiveness in improving consultant responsiveness [8].

Contrary to our expectation, conflictual and accommodating profiles did not differ significantly in age or years of active practice in either group, suggesting that conflict propensity is not primarily driven by career stage or professional experience. This finding is noteworthy given prior evidence that junior physicians are more likely to report interprofessional tension, and may indicate that in the emergency consultation context, structural and situational factors — such as workload, role ambiguity, and communication norms — outweigh individual experience level as determinants of conflict propensity. It further cautions against assuming that clinical seniority alone confers interprofessional adaptability, and stresses the need for conflict-awareness training across all career stages [22, 23].

Limitations

Several limitations should be acknowledged. First, the cross-sectional design precludes causal inference; all associations reported are observational, and no directional or mechanistic conclusions can be drawn. Second, the survey instrument was developed de novo and has not been independently validated; although internal consistency was acceptable (Cronbach’s α 0.71–0.74), construct validity, criterion validity, and test-retest reliability have not been formally evaluated. Third, the conflict propensity score is exploratory and has not been externally validated; dyadic risk estimates derived from Monte Carlo simulation should be interpreted with particular caution, as the model pairs hypothetical physicians drawn from the same marginal score distributions rather than real clinical dyads, meaning the finding of no asymmetry between mixed-profile combinations may reflect the model’s symmetric construction rather than an empirical property of the data. Fourth, because data collection relied on convenience sampling through professional WhatsApp groups, social media platforms, and direct outreach, a formal response rate cannot be reported; the total number of physicians invited was not systematically recorded. This absence of a denominator precludes calculation of response rate and increases the risk of non-response and self-selection bias: physicians with strong opinions about the consultation process may have been more likely to participate, potentially overestimating the magnitude of perception gaps in the broader population. Readers should therefore interpret the findings as characterising perceptions within a sample of engaged physicians rather than as population-representative estimates. Additionally, the electronic survey platform included a single-response restriction; however, the paper-based component did not permit formal duplicate-response verification, and this possibility cannot be entirely excluded. Fifth, the sample was adequate for primary comparisons but subgroup analyses — particularly those involving faculty-level consultants (n = 4) — were substantially underpowered. Sixth, the specialty distribution of the consultant subsample was not recorded, precluding analysis of whether the Overwhelmed Responder profile is concentrated within high-volume specialties or distributed uniformly across disciplines. Seventh, mixed-mode data collection (electronic and paper-based) may have introduced differential response biases. Finally, the study was conducted entirely in Türkiye, where specific contextual factors — including the predominance of general practitioners in emergency departments, the hierarchical structure of postgraduate medical training, and the relatively high consultation burden per shift in the Turkish hospital system — may have amplified role-based divergence and contributed to the Overwhelmed Responder and Dissatisfied Requester profiles identified by LCA. Findings may not generalise to healthcare systems with different staffing models, training cultures, or consultation norms.

Conclusions

Emergency physicians and consultants hold systematically divergent perceptions of the emergency consultation process, attributing the same shared experience of workflow burden to fundamentally incompatible causes. Latent class analysis confirms that these divergent perceptions reflect qualitatively distinct response profiles that align strongly — though not perfectly — with professional role. Dyadic conflict risk is highest when both parties hold conflictual profiles; however, when only one party is conflictual, risk is equivalent regardless of which party holds that profile, consistent with the need to consider system-level rather than individual-targeted approaches. The small but meaningful proportion of physicians whose perception profiles cross role boundaries may represent a tractable target for educational approaches aimed at fostering interprofessional perspective-taking.

Author contributions

Conceptualization: ACT; Methodology: ACT, SK, ATE, YSK; Supervision: ACT; Project Administration: ACT; Investigation: SK, ATE, YSK; Data Curation: SK, ATE, YSK; Formal Analysis: ACT; Visualization: ACT; Writing – Original Draft: ACT; Writing – Review & Editing: ACT, SK, ATE, YSK.

Funding

The authors did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors for this research.

Data availability

The data supporting this study’s findings are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Non-Interventional Research Ethics Committee of Üsküdar University (Approval No: 61351342/2023-47; date: 29 December 2023). All participation was voluntary and anonymous. Informed consent was obtained electronically prior to survey access; participants were clearly informed that they could withdraw at any point without consequence. No personally identifiable information was collected at any stage. The study was fully compliant with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

Dr. Ali Cankut Tatlıparmak is a member of the BMC Emergency Medicine editorial board. He was not involved in the editorial handling or peer review of this manuscript. The remaining authors declare that they have no competing interests.