In the health sector, major incidents correspond to “any incidents where the location, number, severity or type of lives of casualties requires extraordinary resources” [1]. Major incidents are rare events that require efficient and effective management to save as many lives as possible [1,2]. The scarcity of these events and the impact they can have both on the victims and on society, as the impact such stressful situations may have on rescuers clinical performance [3] has led to the development of many different training modalities [4]. Thus, most prehospital providers are trained to manage major incidents through tabletop or computer-based simulations, practical exercises without casualties, or full-scale simulations [5].

These simulations all present limitations, some of which can often be addressed by using other forms of training. However, 1 important limitation is common to almost all major incident simulations and resides in the fact that simulated patients do not evolve over time [6]. Indeed, in most simulations, victims do not evolve at all regardless of the time elapsed and despite treatment and transport decisions. For instance, a simulated victim experiencing a massive internal hemorrhage (such as splenic rupture) will present identical vital signs throughout the simulation [7]. Consequently, most current simulations fail to incorporate the temporal effects of decision-making, which is a major characteristic of major incident management [8].

A simplified mathematical model of human physiology could be a solution for simulating victims’ evolution over time [9]. However, most currently available mathematical models of human physiology present limitations preventing their use in real-time major incident simulations. In 2005, Sacco et al [10] proposed a model using respiratory rate, pulse rate, and motor response (RPM) to predict victims’ survival probabilities. In this model, a single discrete modification of RPM parameters is made depending on whether the victims remains at the incident scene or is transported to a hospital. This change of RPM parameters is based on expert consensus. While insightful for predicting survival probabilities, this model does not enable to calculate the continuous time-based evolution of multiple physiological and clinical parameters, based on treatment decisions. Many others models are complex and ask for high processing power, making them incompatible with their use to simulate the synchronous evolution of several victims[11]. Furthermore, most models are not open source and cannot be embedded into third-party applications [12].

Our objective was to develop and validate a simplified physiological mathematical model capable of simulating the evolution of several casualties, in real time, while remaining realistic in the context of major events training.

A modified version of the user-centered design (UCD) framework proposed by Farao et al [13] was used to define the development process of the physiological model. The UCD framework is based on the 5 modes of design thinking (empathize, define, ideate, prototype, and test) [14], with the addition of the knowledge base mode of the information systems research framework [15].

These 6 modes were sorted in 2 phases, the relevance phase and an iterative development phase. To validate the developed model, a validation phase was added. The overall resulting framework is summarized in Figure 1.

The objective of the relevance phase was to identify the problem to be solved and define the overall goal for the development phase.

The objective of the development phase was to develop a conceptual physiological model, implement it into a software application, and test its ability to simulate the evolution of victims during a major incident.

During the validation phase, relevant elements of Eysenbach’s Checklist for Reporting Results of Internet E-Surveys (CHERRIES) [19] have been used to report the results.

Since the participants who took part in the validation phase were all professionals working in the prehospital field, their participation did not fall within the scope of the Swiss Federal Act on Research involving Human Beings (HRA, RS 810.30). All the participants agreed to take part in this study without any form of compensation, and no incentive other than pure scientific interest was used to promote participation. The results of the evaluation phase were anonymized to ensure the confidentiality of participants’ assessments.

All the participants agreed to take part in this study without any form of compensation, and no incentive other than pure scientific interest was used to promote participation. The results of the evaluation phase were anonymized to ensure the confidentiality of participants’ assessments.

In total, 15 experienced prehospital professionals (5 advanced paramedics and 10 senior registrars working in ground and helicopter emergency medical services) were invited to participate in the validation study. All of them agreed to participate. Their characteristics are detailed in Table 4.

In order to test the different systems and their interactions, 9 simulated patients were automatically generated, with different personal characteristics. We then ascribed a specific pathology to each patient. No patient was ascribed more than 1 pathology. The characteristics of patients and ascribed pathologies are detailed in Table 5.

The validation team assessed each patient clinical and physiological parameters from T0 to T4. The initial and final states of all patients had a median rating of 5 (absolutely realistic) for both clinical (Table 6) and physiological (Table 7) parameters. The overall median rating was also 5 for each individual patient. Detailed assessments, by patient and time period, are displayed in Tables 6 and 7. Ratings as mean (SD) are available in Multimedia Appendix 1.

Overall clinical parameters include oxygen saturation, respiratory rate, heart rate, mean arterial pressure, Glasgow coma scale, and ability to walk. T4 corresponds to the time of death of the simulated patient (T4 equals 4 hours if the survival time was more than 4 hours). T1 to T4 correspond respectively to 25%, 50%, and 75% of T4. Results are presented as “median (IQR).”

Overall physiological parameters include alveolar volume, PaO2, PaCO2, blood volume, stroke volume, and intracranial pressure. T4 corresponds to the time of death of the simulated patient (T4 equals 4 hours if the survival time was more than 4 hours). T1 to T4 correspond respectively to 25%, 50%, and 75% of T4. Results are presented as “median (IQR).”

The overall median rating for each specific clinical (Table 8) and physiological (Table 9) parameter was 5 (absolutely realistic). Further results are available in Multimedia Appendix 1.

A simplified model of trauma patient evolution was successfully created using the modified version of the user-centered design framework and was deemed clinically realistic by experienced prehospital physicians and paramedics. The diversity of simulated clinical cases, which included nervous, cardiovascular and respiratory pathologies, and its suitability for real-time calculations makes this model a good candidate for inclusion in major incident simulations. Many kinds of simulations could be considered, computer-based table-tops and serious games, practical exercises without casualties, and even full-scale simulations. In this latter case, tablets could be used to display the vital parameters of simulated patients.

All clinical and physiological parameters had an overall median rating of 5 (absolutely realistic). Nevertheless, some parameters had lower median ratings when specific patients were analyzed. For instance, the oxygen saturation values of patient 4, who presented a complete left pneumothorax, had a median rating of 4 (rather realistic). Several hypotheses could explain these findings. First, even though an iterative process was followed to progressively refine the evolution of specific parameters, the current model is still an approximation of actual human physiology. Thus, further iterations may enable further improvements. Second, prehospital clinicians quickly manage airway and breathing issues when responding to critical situations. Therefore, these professionals have never witnessed the evolution of patients left untreated and could find it harder to conceive how vital and physiological parameters would develop in such cases. Finally, given the inherent variability in injury presentations and the subjective nature of textual descriptions, there is potential for variation in interpretation among prehospital providers. This inter-rater variability may contribute to the differing assessments of injury realism.

Other models have previously been developed [31], none of which were however deemed suitable for our purpose, that is, real-time simulation of multiple trauma patients with realistic heart-lung-brain interactions. While models such as HumMod and Biogears provide realistic, scientifically grounded models of patient evolution [32,33], their complexity usually requires considerable computational power, making them hardly suitable for major incident simulations. In addition, and although many physiological processes are supported by clear scientific literature and rules, the lack of data regarding most prehospital injuries had to be compensated for by clinical experience. The inclusion of senior prehospital physicians and paramedics since inception has thus enabled the creation of a model realistic enough to simulate the evolution of trauma patients during a major incident.

The model developed through this study presents the advantage of being based on an anatomically realistic model made of several blocks simulating the different body compartments which could be involved in case of injury. Neural and vascular lesions are thus easy to simulate, and these blocks could be divided up further if needed. Furthermore, other organ systems could be added according to the end-user’s requirements: for instance, adding clearance mechanisms could prove useful to the management of major incidents involving hazardous materials. Conversely, disabling or removing systems uninvolved in the simulated situation could save processing power.

Simulating the outcome of patients involved in major incidents could prove useful in several ways. Including an evolutive model in computer-based simulations could help ambulance and medical commanders better understand the impact of certain decisions, such as the use of human resources for treatment rather than for triage. The importance of accurate triage cannot be overstated and methods based on patient evolution should be promoted [9]. In this regard, this physiological model could be used to compare different triage algorithms among different populations. Indeed, some professional categories may be more efficient than others depending on the tool provided [34].

This model is not without limitations. The most important and most obvious limitation is the lack of pediatric-specific formulae, which prevents the generation of patients aged 16 years or younger. Another significant limitation is the absence of treatment-specific effects, which would be necessary to assess the effect of human and material resources through computer-based simulations. Since physiological and clinical parameters of simulated patients were assessed at relative intermediate time points (25%, 50%, and 75%) of death time, the absolute times (hours: minutes: seconds) of those intermediate points varied across simulated patients. This may complicate assessing patients’ conditions over time. However, to enhance clarity, the absolute value of each intermediate time point was displayed on the interface. In addition, experts were informed about the calculation method for intermediate time points before starting the evaluation of parameters. While the overall scores, both for clinical and physiological parameters, show the high realism of the validation study, overall scores could potentially hide specific inconsistencies. However, Q1 and Q3 values are mostly close to median values. The plausibility of simulated patients’ parameters has been assessed by only 15 experts. Given the overall score used in the assessment, the relatively small number of experts who evaluated the model may not have highlighted individual inconsistencies. Finally, the current version of this simplified physiological model was only developed to simulate traumas. Thus, other causes of major incidents, such as those linked to hazardous materials, cannot be simulated without further developments.

Another significant limitation is the absence of treatment-specific effects, which would be necessary to assess the effect of human and material resources through computer-based simulations.

Since physiological and clinical parameters of simulated patients were assessed at relative intermediate time points (25%, 50%, 75%) of death time, the absolute times (hours: minutes: seconds) of those intermediate points varied across simulated patients. This may complicate assessing patients’ conditions over time. However, to enhance clarity, the absolute value of each intermediate time point was displayed on the interface. Additionally, experts were informed about the calculation method for intermediate time points before starting the evaluation of parameters.

While the overall scores, both for clinical and physiological parameters, show the high realism of the validation study, overall scores could potentially hide specific inconsistencies. However, Q1 and Q3 values are mostly close to median values.

The plausibility of simulated patients’ parameters has been assessed by only 15 experts. Given the overall score used in the assessment, the relatively small number of experts who evaluated the model may not have highlighted individual inconsistencies.

Finally, the current version of this simplified physiological model was only developed to simulate traumas. Thus, other causes of major incidents, such as those linked to hazardous materials, cannot be simulated without further developments.

Embedding this model in computer-based simulations and serious games designed to teach major incident management should now be considered. The inclusion of evolutive pathologies could help determine the most efficient ways of dealing with specific major incident situations and prepare health professionals to deal with the stress of managing such events [3]. Further developments should also be contemplated. These should help address several limitations, including the unavailability of pediatric patients and the inclusion of treatment specific effects.

A simplified model of trauma patient evolution was successfully created and deemed clinically realistic by experienced clinicians. This model should now be included in computer-based simulations and its impact on the teaching of major incident management assessed through randomized trials.