The Government, through the Secretariat of Digital Policies of the Department of Business and Employment, has today presented an advanced artificial intelligence (AI) model for the prediction of incidents that has been developed by the Medical Emergency System of Catalonia (SEM) and the Centre of Innovation for Data tech and Artificial Intelligence (CIDAI), within the framework of the Artificial Intelligence Strategy of Catalonia, Catalonia.AI.

The development of this predictive model aims to adapt the SEM’s response and management capacity to possible growth or decline in activity that can be foreseen. This initiative was born with the aim of making an accurate prediction of activity based on the detection of patterns derived from historical data generated by the SEM.

In the words of the Secretary of Digital Policies of the Generalitat de Catalunya, Maria Galindo, “High Impact Projects like this, which CIDAI develops through public-private collaboration within the framework of the Catalonia.AI strategy, contribute to promoting the development and adoption of innovation with AI, demonstrating the value of this technology to solve high-impact challenges, with real use cases”. “In this way, not only are innovative solutions validated, but this advanced knowledge in AI is also transferred to the ecosystem”, she adds.

Although currently activity prediction is carried out based on the average of historical data, the commitment to AI allows the creation of a predictive model with self-learning thanks to the introduction of variables. This predictive vision allows the calculation of the SEM’s resource sizing to be systematized.

“The SEM has a variable volume of activity that requires constant adaptation of the available resources, both in the Health Coordination Center and in the units strategically distributed throughout the territory,” explains the director of the SEM, Anna Fontquerni . In 2023, the SEM responded to 2,211,161 incidents , of which 1,171,291 did not require the activation of resources and were attended to through non-face-to-face assistance. On the other hand, 1,039,870 incidents did require the mobilization of resources, such as an ambulance. “The development of this predictive model using AI is the first step towards having a tool that allows us to anticipate face-to-face and non-face-to-face activity. Thanks to all the data generated by the SEM since 2014 and the preparation of predictive models by CIDAI, this advance is of great use in decision-making”, assures the director of the SEM.

The head of the Information Systems, ICT and Data Area of ​​the SEM, Raimon Dalmau , highlights that “this proof of concept is a first step for the SEM in the use of AI and machine learning techniques, which has given a very good result as a predictive model of activity and we propose to expand its scope in the short term to other aspects of the service.” To generate this tool, a massive data dump has been carried out, collected since 2014. Specifically, information from 27 million incidents of various types has been used.

In this sense, Albert Gual, deputy head of CECOS in Reus , places special emphasis on the predictive capacity offered by the model on different time scales. He highlights that with the tool “you have an approximation of incidents that can be received daily in each work shift, both in a week’s time and in the next few hours”. Thanks to the precision of the data, this tool helps in service planning and improves efficiency without using unnecessary resources, and “would allow the team of professionals to be sized correctly depending on the activity”, assures Gual.

“The impact of this project goes beyond its technical implementation, given that the high reliability of incident prediction demonstrated with the project facilitates more proactive and efficient management of the Emergency System’s resources,” adds the director of CIDAI and Scientific Director of the Digital Area of ​​Eurecat, Joan Mas i Albaigès .

Artificial intelligence models that can predict the volume of incidents with a high degree of accuracy

The SEM receives a large number of calls related to incidents, although their volume varies depending on the time of day, day of the week or time of year, among other factors. In this scenario, the challenge that arose was to implement an AI-based system that could predict the number of incidents in a short time horizon, allowing for early and efficient management of resources.

The project has developed AI models trained on historical data generated by the SEM that can predict future activity volumes with a high degree of accuracy. The project also incorporates a dashboard that has been made available to the SEM, in order to complement its usual operations.

The manager of CIDAI, Marco Orellana, led the presentation of the project, in which the head of Innovation of the i2CAT Foundation, Karla Trejo, also participated; the director of Development of the Artificial Intelligence Ecosystem and Data Spaces at Huawei, Roi Rodríguez; and the researcher in machine learning at Eurecat Arnau Berenguer Jiménez.

The initiative is part of the High Impact Projects (PAI) carried out by CIDAI within the framework of the Catalonia.AI strategy. This PAI, which has had the participation of SEM as promoter of the challenge, has been led by Eurecat, which in collaboration with the i2CAT Foundation and Huawei, has developed the innovative solution based on advanced data analytics and AI tools.