The primary result of this study is the creation of a verified and validated FRAM model of how the paediatric major trauma system functioned during the mass casualty event as shown in Fig. 2.

A FRAM model of the essential functions of a paediatric major trauma centre responding to a mass casualty event

The functions and how they are “coupled” (inter-connected) are represented by the cloud of hexagons and interconnected lines. The model commences with a function outside of the system studied “To manage the on-scene Emergency”, (at the bottom left of the model). This function is termed a background function, shown as a grey box, this function is outside the boundary of the system of interest and activates the model to begin. Other background functions, also coloured grey include “To Provide Paediatric Hospital Care”, referring to ensure care is given to all the other none-mass casualty incident children and “To interface with Hospital Command”, the linkage to the hospital command and control structure during the event. The boundary of study of the model is then completed by a series of final functions, also coloured grey, “To provide PICU post op care”, “To send to Theatre”, “To continue PICU care”, “To continue HDU (High Dependency Unit) care”, “To continue ward care” “To provide ongoing non-incident care” and “to observe in outpatients”.

Each of the stages of the care pathway is given a different colour, notification functions are purple, staff call in is yellow, creation of a surgical commander silver, hospital preparation functions are green, the reception, resuscitation, and radiology functions are red, to manage in theatre or Paediatric Critical Care (PCC) or the wards are blue and the multi-disciplinary ward round is orange.

Table 1 maps the key functions against the resilience potentials of to respond, monitor, anticipate and learn [6]. Most of the key functions of the complex system are responsive in nature, with staff being called in, then a one-way flow of stabilising patients in resuscitation bays, transferring the patients to the Computer Tomography (CT) scanner, then deciding where the patient should go to next, either directly to theatre, to the PICU or HDU or the wards.

Monitoring functions within the system included “To be the Surgical Commander” and “To review (patients) as part of a multi-disciplinary ward round”. The “To be Surgical Commander” was also one of two anticipatory functions alongside “To decide” the (patient) destination. The only learning resilience function identified in the system was the “To review as part of a multi-disciplinary team”.

With confidence developed in the model, actual timings during the MCI were compared with those produced by the model using expected timings for functions. These expected WAI findings were Function Process Time (Tp) the time it took for a function to go from input to output, the WAI Function Output Lag Time (To) the time it took to move from one function ending to starting another function and WAI Total Time of Functions (Tt) the total time for functions in the system. These expected timings were constructed on discussion with subject matter experts, for example discussion with senior PED nurse regarding how many minutes it takes to triage a severely injured child. The exception was the function “To stabilise in Resus” which was theoretically derived from a series of simulated resuscitations suggesting an average resuscitation time of thirty minutes for trauma patients published previously [15]. Work As Done (WAD) in FRAM models represents the actual work done within the system of interest, as opposed to how it is imagined to work (WAI). Mean WAD Function Start Times and Function Process Times are presented. Table 2 shows the expected mean timings produced by the model of the MCI and timings recorded during the MI for the first eight patients, three of whom went to theatre.

The final stage of the modelling was to focus the WAI Trauma system studied (Fig. 1) into a model that captures all the functions of interest for P1 (most severely injured patients), from reception through resuscitation to definitive care. This focussed WAI model (Fig. 3) will allow future “what if” analyses to test the system, with respect to P1 patients. Such “what if” scenarios include what if the number of patients presenting to the hospital exceeds the number of resuscitation bays.

A simplified FRAM model of a paediatric major trauma centre responding to a mass casualty event

The in-depth interviews supported by the FRAM model provided further insights into how the functions supported resilient operations during the management of the MCI. The function “To be surgical commander”, can be observed to provide the resilience potential to monitor and anticipate throughout the system during the incident. This key function was itself a spontaneous adaptation to practice by a single surgeon, which was not detailed in the major incident plan in advance of the incident. In practice the function was achieved by having a senior surgeon on the “shopfloor,” directly observing how care was being provided, as opposed to being sited in a command centre. A practical sequelae of FRAM modelling of the response is that this function has now been established in the major incident plan for a group of hospitals that includes the paediatric hospital studied. A second key function can be identified with many outputs, that of “To review as part of a multi-disciplinary team”. In addition to monitoring patients and ensuring holistic care to children and families in the ward setting, this function also enhanced resilience by capturing learning during the incident, in terms of extent of potential injuries, occult injuries to hunt for, understanding of potential human cross contamination due to shrapnel and other factors [4]. Noticeable by its absence in the model is a function to anticipate the number of incoming patients, which could have been achieved by establishing close contact with the scene of the incident. Review of all interview transcripts highlighted the lack of communication from the scene into the hospital.

A further key function is that of “To decide the patient destination”, this enhanced the response resilience potential of the system, in terms of damage control surgery, further damage control resuscitation or normal critical care / ward care. Further examination of how and where this function was achieved is warranted; particularly when one considers the one-way system of resuscitation-radiology-decision-making to Theatres or PICU, where it was imperative that patients did not return to the resuscitation bay after leaving for radiology.

The modelling has also highlighted the central and rate-limiting role of CT scanning has on the response of the system during the major incident. Due to the high likelihood of blast injuries, a high number of children were CT scanned. For the most seriously injured this was after approximately thirty minutes of resuscitation (including intubation, ventilation, and sedation) prior to CT in one of the three staffed resuscitation bays initially available. The model highlights the key function of “To perform CT scan”, particularly when the hospital has only one scanner available. If, unlike on the night of the event, more than three patients had arrived contemporaneously, each requiring resuscitation, then a second function of “To re-triage for CT during resuscitation” would be required to ensure the finite resource of the CT scanner was not targeted to the wrong patients during the one-way resuscitation flow described above. At the research site hospital this would now entail Trauma Team Leaders in the resuscitation bays having a structured communication huddle, possibly mid-stabilisations, to determine the appropriate order of patients to go to CT scan. This could mean that the first child stabilised for CT may not be the first to go, for example, if a patient with a time-critical head injury was “about” to be stable for CT scan, they would take priority.

Analysis of the expected mean process timings from FRAM model and mean actual process timings of the MCI (Table 2) also provided some other valuable insights for future MIP development. The model predicted that on average a patient would arrive in the CT scanner every 37 min. The actual average time was 38 min from arrival to commencing scanning, providing some construct validity to the modelling process. However, based on a damage control operative time of sixty minutes, analysis of the model highlights that actual times in theatre were more than twice this. It has been recognised that improvements in damage control timing is required, with a second anaesthetist in theatre now monitoring this in the current MIP. Comparison of the model and actual timings also shows a significant overshoot on predicted timings for post-operative PICU bed availability after the end of surgery, which is also being optimised.

A team experienced in creating and analysing FRAM models is required [16] and the structure and output of a FRAM model is dependent on the information provided to this team [17]. Reflexivity is a state of continual awareness and understanding on the part of research team members that their prior experiences and/or assumptions may influence all aspects of the study [18]. One researcher (RM) worked within the system modelled during the mass casualty event. Several steps were taken to foster a reflexive research study design including a continued reflexive dialogue between the international researchers with differing research backgrounds and understandings of the study phenomenon. While the model in this study was verified using the FRAM Model Interpreter and by comparing actual timings with model predictions, these methods may not capture all aspects of system performance, particularly under extreme stress conditions like MCIs. Also, the transferability of the FRAM model itself is limited since it is built on a local system. However, modern MCI management is built on similar principles in different settings and at that level the model might provide insights also in other health care settings.