C3 - 03: ASSESSMENT OF TRAUMA TEAM COMMUNICATION USING AUTOMATED DISCOURSE CODING AND EPISTEMIC NETWORK ANALYSIS
Hee Soo Jung, MD, Alexandra Rosser, BS, Charles Warner-Hillard, MPH, Ryan Thompson, MD, Brooke Moungey, MD, Valerie Mack, BSN, Carla M Pugh, MD, PhD, David W Shaffer, PhD, Sarah A Sullivan, PhD; University of Wisconsin
Background: Epistemic Network Analysis (ENA) is a technique that can model and compare the structure of connections between the elements of team communication and their impact on team performance. We hypothesized that team communication elements could be coded with an automatic discourse coder and that their connections, as modeled by ENA, would predict the quality of team performance in trauma simulation.
Methods: 28 teams of five trainees - a trauma chief resident, surgery resident, emergency medicine resident, and two nurses - participated in simulated trauma team resuscitations. The Team Emergency Assessment Measure (TEAM) was used to rate teamwork during each resuscitation.
An interdisciplinary group performed an in-depth qualitative analysis of two transcripts. Eight communication codes were identified. Algorithms matching codes to key phrases and patterns of words were developed. These were validated by measuring agreement of automated coding algorithms with coding by a trained human rater on a subset of total transcribed discourse (kappa ≥ 0.86).
ENA was used to model the connections between communication codes. ENA sums the co-occurrences of codes in nearby lines of discourse and uses these sums to produce a network of connections between codes. ENA models of teams in the highest tertile of TEAM scores (n = 9, mean = 48.56±1.50) and lowest tertile (n = 10, mean = 40.90±2.34) were compared. Further qualitative analysis was used to identify the codes that accounted for significant and meaningful differences between groups: Asking about Actions, Asking for Information, Intentions, Reasoning, Pathological Descriptions, and Status Quo Information.
Results: ENA models of higher- and lower- performing teams were significantly different (p = 0.002). Higher scoring groups had stronger connections between Status Quo Information (describing completed actions or current status) and Reasoning (providing justification or weighing options based on known information). Reviewing the original data that produced these connections in the model suggests that this reflects when members provide their thoughts linked to information delivered about the current situation.
Conclusions: Connections between describing the current status and providing reasoning were associated with higher performing team performance. Trauma team communication performance can be assessed using automated discourse coding and ENA.