Standardized data definitions for adult trauma video review: an expert consensus report

Abstract

Background

Trauma video review (TVR) provides granular data for quality improvement, education, and research but is hampered by a lack of standardized data definitions, which limits multicenter collaboration, benchmarking, and the generalizability of findings. We aimed to develop a universal data dictionary and collection tool through an expert consensus process to standardize adult TVR data capture across institutions.

Methods

The Trauma Video Review Collaborative convened a multidisciplinary panel of subject matter experts in trauma resuscitation from eight centers in the USA and Canada. Through a series of virtual meetings and discussions between May 2023 and July 2024, the group used a consensus-based process to identify and define core data elements. The development focused on core data elements directly observable from video, structured to mirror the clinical workflow of trauma resuscitation.

Results

The final standardized data dictionary includes 95 items (44 categorical and 51 numeric/timestamps) with detailed definitions to ensure uniform collection. Categorical variables can be marked "unable to determine" to distinguish from missing data, as are the numeric/timestamps with a predefined timestamp. All timestamps are in military time format. Data points are pragmatically organized into five categories: (1) pre-arrival preparation, (2) handoff, (3) resuscitation period, (4) procedures and interventions, and (5) non-technical skills, incorporating the Trauma Non-Technical Skills tool. The dictionary is designed to link with registry data for patient demographics and outcomes.

Conclusion

This work establishes the first expert consensus-driven, universal data dictionary for adult TVR. It provides a common lexicon to standardize data collection, facilitating robust multicenter research, enabling objective benchmarking, and lowering the barrier to entry for new TVR programs. This tool represents a critical foundational step toward creating a large-scale TVR data registry to advance the science and practice of trauma resuscitation.

Introduction

Trauma video review (TVR) is the practice of audiovisual recording of trauma resuscitations for the purposes of quality improvement, education, and research.1,5 It offers the granular data required to recreate complex interactions during trauma resuscitation in order to identify key interventions and uncover factors associated with improved outcomes.⁶ Other methods of data collection have been shown to suffer from recall bias and poor accuracy compared with TVR.⁷

Although previous studies using standardized data definition and collection provided substantial insights in adult trauma care,^{8 9} they are focused on specific study objectives. Establishing a universal data dictionary has the potential to increase the impact of TVR. Using a standardized dataset, data can be integrated across multiple institutions allowing for the inclusion of patients from different demographics and locales and facilitating benchmarking and collaboration among institutions. A standardized dataset also establishes a common lexicon by which variables are consistently defined and captured, minimizing any variations in data collection. Similar data definition tools for emergency airway management¹⁰ and cardiac arrest¹¹ have led to more uniform data collection and research.

Building on previous partnerships,^{1 2 8} the Trauma Video Review Collaborative (TVRC)⁹ members enlisted a group of subject matter experts to create a universal data dictionary and collection tool to improve the quality improvement process, delineate practice patterns, and identify areas for further research.

Methods

The TVRC approached centers with active adult TVR programs and formed a group of experts tasked with the creation of a standardized data dictionary and collection tool to facilitate collaboration among TVRC sites, as well as participation of new centers planning to establish adult TVR programs. Given no human subject research, no ethics approval was necessary.

Tool development

A multidisciplinary group of subject matter experts in adult trauma resuscitation and quality improvement, with representation from trauma surgery, burn surgery, emergency medicine, critical care medicine, and trauma nursing was recruited. All members had extensive experience in TVR use in quality improvement and research. Members coordinated via group emails and met virtually on four occasions between May 2023 and July 2024 to discuss and define the TVR universal data dictionary. Each meeting averaged 16 attendees representing eight centers in the USA and Canada. All meetings were held virtually, were 1 hour to 2 hours in duration, and were recorded for those who were unable to attend. Members communicated between meetings using group email as a discussion platform. Disagreements were discussed until a formal vote of consensus was achieved.

During these meetings, the working group examined key events recorded by TVR that could impact the care of adult critically injured patients and identified data points that were considered core to TVR. Data points considered for inclusion could occur at any time in the resuscitation from the preparation phase to transport out of the bay. The group then identified and removed redundant data points. To organize the data structure, categories were created to mirror the flow of care in real time and to make TVR more efficient. The group avoided the inclusion of data readily available in the patient record, such as patient demographics and outcomes.

Results

The final standardized data dictionary included 95 items: 44 categorical and 51 numeric. All categorical variables were assigned specific response values, such as yes/no, whereas numeric variables were all times at which key interventions occurred. Time was specified to be recorded in military time format (day/month/year hours:minutes:seconds) for accurate calculation of time to events/interventions. A detailed description was provided for all data points to aid in uniform data collection. In order to distinguish inability to abstract data from missing values, ‘unable to determine’ is a distinct category for categorical variables, and ‘01/01/2020 00:00:01’ signifies unable to be abstracted from video online supplemental file 1 contains the developed tool).

The data points were pragmatically categorized into five groups: (1) pre-arrival preparation, (2) handoff period between prehospital staff and trauma team members, (3) resuscitation period, (4) procedures and interventions phase, and (5) non-technical skills (figure 1).

Figure 1. Trauma video review data categories.

Hospital factors, patient face sheet, and outcomes data

The group determined that although patients will have center identifiers, center-specific variables were not recorded for each patient and instead be linked using a standard data registry tool. These include core measures such as trauma center verification level, hospital size, location, and urbanicity. Patient-specific variables such as demographic data and outcomes follow the same definitions as the National Trauma Database and will be collected from trauma registries maintained by each center online supplemental file 1.

Pre-arrival preparation

Group discussions focused on the variability of pre-arrival preparation between centers. Accordingly, the group agreed on three data points to be collected for this phase: whether the camera was on patient’s arrival, whether trauma team pre-arrival briefing was performed or not, and the trigger for TVR online supplemental file 1.

Patient arrival and handoff

Since time to event (eg, ‘time to access’ and ‘time to blood’) analysis requires a ‘time zero’; patient arrival time, defined as the time of appearance of the patient on the screen, was chosen. Emergency medical service (EMS) handoff times were defined and added. It was decided that factors such as interruptions to handoff, repetition of information, and availability of pre-hospital vascular access be captured in a binary fashion (table 1).

Table 1. Data definitions and values for pre-hospital and patient handoff phase.

Pre-hospital/Patient handoff	Variable values	Definition
Hand-off start time	Numeric (time)	When EMS starts hand-off, record as time of day, allows to calculate difference between arrival and start of hand-off
Hand-off end time	Numeric (time)	When EMS completes hand-off tool, or if tool not complete, when EMS stops speaking, record as time of day
Did patient arrive with pre-hospital access?	Categorical: yes/no/unable to determine	If the patient arrived with any type of access (intravenous/intraosseous)

Time is recorded in day/month/year hours:minutes:seDay/Month/Year Hours:Minutes:Seconds format. The full list is available in online supplemental file 1.

EMS, emergency medical service.

Resuscitation period

Resuscitation phase was defined as the time the primary survey was initiated to the time of completion. Completion was defined separately depending on blunt or penetrating mechanism of injury (table 2). Diagnostic adjuncts including radiography and focused assessment with sonography for trauma were also included within the resuscitation phase. Although individual components of the primary survey are to be carefully examined, the group determined that capturing discrete timestamps for each component falls outside the purview of universal data collection. The time of the first set of vital signs, defined as recording at minimum a systolic blood pressure and a heart rate, was identified as a crucial timestamp and was included. Similarly, the time of the first episode of hypotension (systolic blood pressure <90 mm Hg) was included.

Table 2. Data definitions and values for the resuscitation phase.

Resuscitation	Variable values	Definition
What parts of the primary survey were completed?	Check all that apply (airway/breathing/circulation/disability/exposure)	Based on vocalization, for example, ‘airway intact, bilateral breath sounds, bilateral DP pulses’, etc.
Time primary survey completed (once all boxes above complete)	Numeric (time)	Completion of primary will depend on blunt versus penetrating mechanism, in blunt mechanism completion of primary includes airway, breathing, circulatory, disability, and signs of anterior external injury, does not include log roll. In penetrating injury, include all the same as blunt injury; however, it also includes examination of the back and record as time of day.
Was the patient rolled?	Categorical: yes/no/unable to determine	Log roll of patient to complete exposure
Hemothorax or pneumothorax	Categorical: yes/no/present on video review but not identified by team/unable to determine	For a hemo/pneumothorax to be considered relevant on the primary survey, it must be of large enough size to require intervention
Significant external hemorrhage	Categorical: yes/no/present on video review but not identified by team/unable to determine	Ongoing external hemorrhage more than simple venous oozing
Time of FIRST intravenous start attempt	Numeric (time)	Time FIRST attempt of ANY TYPE (IO/intravenous/CVC) in trauma bay, defined as needle on skin if able; excluding pre-hospital access.
Access type?	Categorical: intravenous/IO/CVC	If CVC is the first attempt, will also auto-populate in the procedures module
FAST result	Categorical: positive abdomen/positive heart/positive both/negative/indeterminate/incomplete/unable to determine	The results stated by the team in the trauma bay, if only part of the FAST is done and no positive finding reported, mark as incomplete
Any episodes of hypotension?	Categorical: yes/no/prehospital/unable to determine	Including if patient arrives hypotensive, defined as systolic blood pressure <90 mm Hg

Time is recorded in day/month/year hours:minutes:seDay/Month/Year Hours:Minutes:Seconds format. The full list is available in online supplemental file 1.

CVC, central venous catheter; FAST, focused assessment with sonography for trauma; IO, intraosseous.

Cavitary triage, presence of unprotected airway, hemo/pneumothorax, unstable pelvis, and external hemorrhage were all included with definitions to help accurately record as binary variables (table 2). The group decided to include access and time to access as part of the resuscitation. Any type of access, whether established by EMS or placed in the emergency department, was included and clearly defined in online supplemental file 1.

Procedural evaluations

The working group established that the timing of all procedures could be conceptualized in three phases: (1) time at which the team leader requested the procedure; (2) the time at which the procedure physically began and (3) the time of completion. Consequently, three data points were included for each procedure decision, initiation, and completion, with definitions included to ensure reliability (table 3). Likewise, if the decision to perform a procedure was not verbalized by the team leader, the decision time was to be marked as unable to be determined.

Table 3. Data definitions and values for the procedures phase.

Procedures and interventions	Variable values	Definition
Was MTP activated pre-arrival?	Categorical: yes/no/unable to determine	MTP activation verbalized prior to arrival
Intubation—decision time	Numeric (time)	Time that individual leading the resuscitation verbalizes decision to intubate, record as time of day
Intubation—start time	Numeric (time)	When laryngoscope (or bronch if fiberoptic approach) inserted into mouth, record as time of day
Intubation—completion time	Numeric (time)	ETT confirmed to be in place by treating team and first breath delivered via BVM/vent, record as time of day
CPR—performed	Categorical (pre-hospital, in trauma bay, no)	was CPR initiated at any point
CPR start time/initiation	Numeric (time)	If prehospital use patient arrival time; if in trauma bay, record as time of day.
ROSC	Categorical: yes/no/unable to determine	If ROSC is achieved
ROSC time	Numeric (time)	Time of ROSC verbalized by team or present on video, record as time of day.
Time of death in trauma bay	Numeric (time)	Record as time of day if patient pronounced in the bay
Disposition	Categorical	OR, IR, ICU, morgue, to CT, to ED, other
End of resuscitation	Numeric (time)	Time trauma leader indicates the patient is ready to leave the trauma bay, record as time of day; if not explicitly stated then determined by reviewer

Intubation is included as an example, as all procedures and interventions have a variable to capture a failed attempt (a decision was made but the procedure was not performed or completed) as well as decision, start, and stop times. Time is recorded in day/month/year hours:minutes:seDay/Month/Year Hours:Minutes:Seconds format. If unable to determine the timestamp ‘01/01/2020 00:00:01’ will be entered. The full list is available in online supplemental file 1.

BVM, Bag valve ventilation; CPR, cardiopulmonary resuscitation; ED, emergency department; ETT, endotracheal tube; ICU, intensive care unit; IR, interventional radiology; MTP, massive transfusion protocol; OR, operating room; ROSC, return of spontaneous circulation.

In addition to key procedures, specific interventions and resuscitation-related events were also identified by the expert panel. Identified interventions included blood transfusions, activation of a massive transfusion protocol, and administration of transfusion adjuncts (intravenous calcium and tranexamic acid). Key significant events included post-intubation hypotension, cardiac arrest, return of spontaneous circulation, and time of death in the trauma bay or time of departure was also included (table 3).

Non-technical skills

The group recognized the importance of evaluating the non-technical aspects of care and agreed on the use of Trauma Non-Technical Skills teamwork rating tool¹² as a validated assessment tool. Given the current use of a modified version,^{13 14} it was agreed to adopt this simplified version.

Discussion

We developed a universal data definition and collection tool developed by a panel of experts from the TVRC. TVR is a widely adopted practice throughout the USA.⁶ It offers valuable insights into the initial phase of in-hospital care of adult trauma patients.15 8 9 13 15,17 Data definitions in previous studies were established individually for each study, which leads to some inconsistencies. TVR data collection is currently restricted to the study’s duration and focuses solely on predefined hypotheses. To unlock the full potential of TVR in exploring the ‘known unknowns’ of trauma care,³ a universal approach to data definition and collection is essential. This universal data collection tool represents the first step in unified data collection across TVR programs. All TVRC sites are committed to transitioning to this new methodology of data collection and are working on developing data sharing agreements between their respective institutions.

A universal data collection tool is crucial to the continued evolution of trauma care for several reasons. First, universal data definition and collection simplify collaboration among centers and facilitate higher-quality prospective multicenter research. A shared vocabulary helps to mitigate potential biases that are inherent in human observation. Second, by facilitating continuous enrollment of patients into TVR data registries, this tool could increase the inclusion of previously under-represented populations, strengthening the generalizability of derived insights to a wider population. Third, it opens new avenues of research such as an objective comparison of variability in resuscitation practices across many centers and potentially identifying factors associated with improved outcomes. Fourth, the tool’s modularity and ‘baseline’ data collection facilitate specialized, time-limited data collection tools for specific projects or sites, requiring minimal extra effort after the initial setup. Lastly, as prior research supports including TVR in performance improvement initiatives at Level I and II trauma centers,³ this tool lowers the barrier to entry for centers without an existing TVR program.

Data definitions were a challenge. Some data definitions such as tracheal intubation,¹⁰ tube thoracostomy,¹⁵ and thoracotomy¹ were adopted from the published literature, whereas, in others, the group opted for a more universally applicable definition than previously published definitions (eg, ‘Resuscitation end time’ was defined as the time the team leader declares the disposition).¹⁸ The advantage of collecting timestamps is that new time to events/interventions can be created, in addition to consensus-defined ones (table 4). Since all centers periodically delete videos as part of data management protocols,⁶ another challenge for the group was to include as much granular data as possible, avoiding overburdening data abstractors.

Table 4. Calculated time intervals for a selection of outcomes.

Calculated intervals	Formula
Handoff length	(Hand-off end time) − (hand-off start time)
Time to first set of vitals	(Time of first vitals) − (patient arrival time)
Time to intravenous access	(Time to access completion) − (patient arrival time)
Time to blood	(Blood transfusion start time) − (patient arrival time)
Time to first procedure	(Decision time for first procedure/intervention) − (patient arrival time)
CPR duration	(Time of ROSC) − (CPR initiation)
Time to the next phase of care	(Time of departure) − (end of resuscitation)
Patient trauma bay dwell time	(Time of departure) − (patient arrival time)

Time is recorded in day/month/year hours:minutes:seDay/Month/Year Hours:Minutes:Seconds format. The full list is available in online supplemental file 1.

CPR, cardiopulmonary resuscitation; ROSC, return of spontaneous circulation.

TVR is a labor-intensive task, and physicians and allied health staff mostly work on a volunteer basis. The rapidly evolving fields of context-aware multi-modal artificial intelligence (AI) and vision AI¹⁹ offer significant potential for automating TVR and removing the human reviewer bottleneck. This tool will be indispensable in benchmarking the performance of such technologies against human experts.

Future challenges and opportunities

As with any first attempt, there are limitations to consider. This data dictionary was established on a consensus basis by subject matter experts; however, it may not represent TVR researchers as a whole and is prone to human error in foresight. To maximize input from different perspectives, we opted for an open forum discussion instead of other methodologies.²⁰

An inherent limitation of definitions is that they are established in the context of the available technology at the time. Furthermore, as the science of trauma evolves, so will the standards and benchmarks. As such, this data dictionary will require periodic updates and revisions to maintain pace with current practice guidelines and standards of care. To give the audience easy access to the latest iteration, as well as the ability to submit potential issues with data collection and entry, we have created a dedicated webpage (https://tvrcollaborative.org/), and a data collection tool is also available on REDCap instrument library under TRVC Data Dictionary). Submitted issues will be screened and addressed either via quick updates for minor issues or during the TVRC annual meeting.

With each update, the threat that previously collected data may no longer seamlessly integrate with current data exists. One limitation of the current effort is its focus on adult patients; this is in part to avoid overlapping with ongoing efforts by already existing collaboratives among pediatric centers focusing on video review of pediatric trauma resuscitations.²¹ Another limitation is that it is developed based on practice norms primarily in the USA and Canada, and as such, may have limited generalizability to other unique trauma systems or circumstances.

Conclusions

This tool developed by an American and Canadian collaborative of TVR experts will facilitate standardization of the adult TVR data collection. It represents the first step in the development of a data registry of TVR that will be used to advance the care of trauma patients.