Over the years, fire departments have been searching for methods to identify their operational disruptions and establish strategies that allow them to efficiently organize their resources. The present work develops a methodology for breakage calculation and another for predicting disruptions based on machine learning techniques. The main objective is to establish indicators to identify the failures due to the temporal state of the organization in the human and vehicular material. Likewise, by forecasting disruptions, to determine strategies for the deployment or acquisition of the necessary armament. This would allow improving operational resilience and increasing the efficiency of the firemen over time. The methodology was applied to the Departmental Fire and Rescue Doubs (SDIS25) in France. However, it is generic enough to be extended and adapted to other fire departments. Considering a historic of breakdowns of 2017 and 2018, the best predictions of public service breakdowns for the year 2019, presented a root mean squared error of 2.5602 and a mean absolute error of 2.0240 on average with the XGBoost technique.

No abstract

When ambulances’ turnaround time (TT) in emergency departments is prolonged, it not only affects the victim severely but also causes unavailability of resources in emergency medical services (EMSs) and, consequently, leaves a locality unprotected. This problem may worsen with abnormal situations, e.g., the current coronavirus disease 2019 (COVID-19) pandemic. Taking this into consideration, this paper presents a first study on the COVID-19 impact on ambulances’ TT by analyzing historical data from the Departmental Fire and Rescue Service of the Doubs (SDIS 25), in France, for three hospitals. Because the TTs of SDIS 25 ambulances increased, this paper also calculated and analyzed the number of breakdowns in services, which augmented due to shortage of ambulances that return on service in time. It is, therefore, vital to have a decision-support tool to better reallocate resources by knowing the time EMSs ambulances and personnel will be in use. Thus, this paper proposes a novel two-stage methodology based on machine learning (ML) models to forecast the TT of each ambulance in a given time and hospital. The first stage uses a multivariate model of regularly spaced time series to predict the average TT (AvTT) per hour, which considers temporal variables and external ones (e.g., COVID-19 statistics, weather data). The second stage utilizes a multivariate irregularly spaced time series model, which considers temporal variables of each ambulance departure, type of intervention, external variables, and the previously predicted AvTT as inputs. Four state-of-the-art ML models were considered in this paper, namely, Light Gradient Boosted Machine, Multilayer Perceptron, Long Short-Term Memory, and Prophet. As shown in the results, the proposed methodology provided remarkable results for practical purposes. The AvTT accuracies obtained for the three hospitals were 90.16%, 97.02%, and 93.09%. And the TT accuracies were 74.42%, 86.63%, and 76.67%, all with an error margin of 10 min.

Emergency medical services (EMS) provide crucial emergency assistance and ambulatory services. One key measurement of EMS’s quality of service is their ambulances’ response time (ART), which generally refers to the period between EMS notification and the moment an ambulance arrives on the scene. Due to many victims requiring care within adequate time (e.g., cardiac arrest), improving ARTs is vital. This paper proposes to predict ARTs using machine-learning (ML) techniques, which could be used as a decision-support system by EMS to allow a dynamic selection of ambulance dispatch centers. However, one well-known predictor of ART is the location of the emergency (e.g., if it is urban or rural areas), which is sensitive data because it can reveal who received care and for which reason. Thus, we considered the ‘input perturbation’ setting in the privacy-preserving ML literature, which allows EMS to sanitize each location data independently and, hence, ML models are trained only with sanitized data. In this paper, geo-indistinguishability was applied to sanitize each emergency location data, which is a state-of-the-art formal notion based on differential privacy. To validate our proposals, we used retrospective data of an EMS in France, namely Departmental Fire and Rescue Service of Doubs, and publicly available data (e.g., weather and traffic data). As shown in the results, the sanitization of location data and the perturbation of its associated features (e.g., city, distance) had no considerable impact on predicting ARTs. With these findings, EMSs may prefer using and/or sharing sanitized datasets to avoid possible data leakages, membership inference attacks, or data reconstructions, for example.