Skip to main content
conference lunch move company map contacts lindholmen lindholmen 2 travel info


PreSISe - Prehospital Decision Support for Identification of Sepsis Risk

PreSISe is a long-term project to develop and evaluate AI-based decision support for prehospital assessment of sepsis risk.

Sepsis (formerly known as blood poisoning) is a common, life-threatening condition — often with a rapid course of development — that affects between 25,000 and 40,000 people a year in Sweden, more than the three most common forms of cancer combined (prostate, breast and colon). The mortality rate is high, around 20%. The incidence is therefore higher and the mortality rate is percentage-wise on a level with more well-known illnesses such as stroke and heart attacks (approximately 25,000 cases of each annually). It is therefore high time, and a matter of some urgency, to put countermeasures in place to reduce the mortality rate and effects of sepsis. Artificial intelligence, or AI, provides us with new tools to do this.

Sepsis is defined as “an infection with life-threatening organ dysfunction caused by an improperly regulated systemic immune defense.” The condition emerges when the immune system attacks the body’s own tissues as a misdirected response to defend against invading bacteria. Survivors may need to undergo amputation or run the risk of other complications such as kidney problems. This results in a great deal of suffering, as well as high health care costs. Despite the fact that sepsis is a global problem and one of the leading causes of death, it is relatively unknown, even in health care. Early identification and rapid treatment are crucial to the outcome. Every delay in treatment with antibiotics lessens the chances for survival and increases the risk of complications.

Prehospital care is care administered by ambulance attendants, for example. Important decisions about treatment, the care process, and even destination are taken here, early on in the healthcare chain. In order to support staff, it is necessary, for example, to have a better exchange of information with other players, as well as more and better clinical decision-making support. More than half of patients with sepsis have their first contact with medical care in the prehospital phase. If patients at risk are identified early, treatment can begin in the ambulance, the receiving clinic can be warned ahead of time and the patient can be directed to the right place at the hospital more quickly. Early identification is therefore fundamental. Currently, this comes from clinical experience and simpler tools based on vital signs. One problem is that the vital signs are normal in one-third of the cases. In addition, the symptoms are numerous and diffuse, and can be similar to less serious conditions such as influenza. There are indications that ambulance personnel suspect sepsis in as few as 12% of cases. More reliable risk prediction can therefore have a major effect on mortality rates and complications with even moderate improvements. Investigating how AI methods can increase precision through such measures as including symptoms in the assessment is therefore justified. The PreSISe project has therefore been initiated within PICTA (Prehospital ICT Arena). The goal is within a broadly composed stakeholder group to develop, evaluate and put into use a national, “open” prehospital AI decision-making support system to assess sepsis risk. In addition, the objective is to also define a model for how this decision-making support can over the long term be made available, maintained, improved and administered by the “appropriate” stakeholders after the end of the project. To achieve the desired use and dissemination, the correct management of CE requirements and to ensure the continual supply of quality-assured input data for continuous improvement, these issues must be taken into consideration right from the start of the project.


The decision support is designed for ambulance care and a general IT platform (see Figure 1), an “ambulance journal.” In the structured work process, for example under AMLS (Advanced Medical Life Support), the ambulance journal is continually updated with information that is processed in the background and assessed by an AI algorithm based on sepsis risk. Personnel are notified in the event of heightened risk. Adaptation of the care process and adequate measures (for example, organ support therapy or supplementary tests) can be started and the receiving clinic notified in advance. This differs in several ways from the current situation. On the one hand, it does not require the staff to suspect sepsis at an initial stage; on the other, statements by the patient or their immediate family are taken into consideration in addition to vital signs, and finally the information can be processed by AI algorithms that are expected to be superior to current methods. The AI support does not make the diagnosis, however; that happens further along the healthcare chain, since this may require a bacterial culture, for example. Only risk is predicted.

PreSISe is being driven towards its goal along three parallel tracks that interact with each other: a broad, nationally available AI decision-making support that has garnered support (Figure 2). The three tracks are data collection and compilation, AI methods and visualization.


An initial step has been taken through the structure of PreSISe-1, for which funding is now being sought. The foundation for the decision-making support and its long-term national administrative model is being laid here. The three interacting tracks thus involve:

  1. Pooling and processing of the parties’ existing sepsis data
  2. Development and evaluation of selected AI methods on this sepsis data
  3. Through user tests, evaluating how the results are to be visualized and implemented so as to function in clinical operations, as well as their effects on the care process


In addition, a long-term administrative model in PreSISe-1 will be proposed. This includes, but is not limited to, data representation, interoperability and standards, maintenance and development, regulatory issues including CE and MDD/MDR, and availability through open source.

Partners in PreSISe-1

  • PICTA (Prehospital ICT Arena)
  • Chalmers
  • PreHospen (Högskolan i Borås)
  • Skaraborgs Sjukhus Skövde (VGR)
  • Karolinska Institutet Södersjukhuset
  • Aweria AB
  • InterSystems Sverige AB
  • MedITeQ AB

PICTA’s focus group in clinical decision support, as well as groups such as FLISA, are reference groups for the project.