Development, validation and use of a medical risk analysis too for mass gatherings

Mass gathering medicine (MGM) forms a unique and challenging subspecialty of disaster medicine, which falls under emergency medicine. Because patients at a mass gathering event (MGE) can present a whole range of conditions and injuries, from minor injuries or common illnesses to life-threatening conditions and injuries, routine prehospital care providers like emergency medical technicians and paramedics, emergency nurses, and physicians are ideally trained to provide in-event healthcare. But there is more to MGM than the on-site treatment of patients during an event. Those who provide mass gathering medicine must also identify potential public health threats and try to mitigate, minimize, and respond to public health emergencies. Hence, it’s a unique and challenging character.
Moreover, MGE can put a severe burden on the host community. To mitigate the burden of the MGE on the host emergency medical services (EMS) and local EDs, adequate inevent health services (IEHS) are indispensable. However, the planning and prediction of medical resources (personnel, equipment, and finances) for any MGE is still inaccurate and sometimes even happens haphazardly. Even after four decades of mass gathering studies, there are no international standards for the numbers of medical personnel and the level of care that should be present at a MGE. If any regulations exist, they are fragmented and primarily based on local standards of care. Although several authors published models that predict the number of patient presentations, only some models have undergone cross or external validation.
This doctoral study aimed to validate a Belgian prediction model for patient encounters at
mass gatherings, i.e., Plan Risk Manifestations (PRIMA). To validate the model, we
compared the actual number of patient presentations at three types of MGE with the
predicted number of patient presentations of two previously published prediction models
and the PRIMA model. We chose three MGE that regularly occur in Belgium, i.e., music
mass gatherings, public cultural gatherings, and football mass gatherings.
First, we had to elaborate on the development of the PRIMA model by reporting on the two-step method we followed to develop, update, and support the model. The first step was a continuous, systematic literature review. The second step was running the calculation model using the data provided by the Belgian Federal Public Service Health, Food Chain Safety and Environment, event organizers, and municipalities. We decided that the starting point of our model was three medical axes: 1) isolation risk, 2) population risk, and 3) risk of illness. The three medical axes and the calculation model in the PRIMA model make it possible to predict the number of patient presentations and advise on allocating medical resources on-site at the MGE.
Second, we validated the model for music MGEs. We compared the actual number of patient presentations from 87 music MGEs with the predicted number of patient presentations by the Arbon, Hartman, and PRIMA models. We found that all three models had high under-prediction rates and moderate over-prediction rates. When we compared the mean deviations, the PRIMA model had the lowest mean deviation of all predicted patient presentation rates, the highest mean number of over-prediction, but the lowest mean number of under-predictions. In addition, the PRIMA model made two correct predictions on the number of patient presentations. 
Third, we validated the PRIMA model for public cultural MGEs. To do so, we compared actual patient presentation rates from the Ghent Festivities, Belgium’s largest public open-air cultural festival, collected between 2013 – 2019, with predictions by existing models. When we compared the Arbon, Hartman, and PRIMA predictions, we found significant differences between the predicted number of patient presentations and the actual number of patient presentations. The PRIMA model overestimated the number of actual patient presentations for each occasion, whereas the other models had a high underprediction rate.
Fourth, we validated the PRIMA model for football MGEs. To do so, we compared actual patient presentation rates from 41 football games played at Belgium’s largest football stadium (King Baudouin stadium, Brussels, Belgium) between 2010-2019, with predictions by the Arbon, Hartman, and PRIMA models. We found that the PRIMA and Hartman models overestimated the number of patient encounters for each game. On the other hand, the Arbon model was the only model to underestimate the number of patient presentations.
Finally, we wanted to see if the presence of in-event health services (based on the PRIMA model’s advice) could reduce the impact of Europe’s largest electronic dance music festival on the host community emergency medical services and local emergency departments. To do so, we conducted a retrospective analysis of all patients seeking medical care at Europe’s largest EDMF. We found that most patients only required first aid, but 150 patients needed transport by IEHS to local EDs. In total, 18 patients who were taken to the hospital by IEHS remained admitted to the local hospital for more than 24 hours. None of the EDs of the local hospitals exceeded their standard operational capacity, nor did any of their ICU departments (although one hospital had more admissions to their ICU wards during the MGE than average). The impact of the MGE on local EMS was also limited. The ambulance closest by had a slight increase in EMS calls, but the ambulance did not exceed its standard operational capacity. On average, the other local ambulance had fewer EMS calls during the MGE than during the weekend days when there was no MGE. To conclude, although we did encounter some discrepancies between the predictions of the PRIMA model and the actual patient presentations at the MGE included in our studies (e.g., the high number of underpredicted patient presentations at music MGE), we concluded that the PRIMA model is, within its domain of applicability, sufficiently valid for the intended application.