Evaluation of CDSS in Detection of SIRS and Sepsis in Pediatric Patients

Overview

This trial aims to evaluate the accuracy of a Clinical Decision-Support System to support early recognition of SIRS in paediatric intensive care patients. This assessment will be rated by the primary goals, the sensitivity and specificity of the system. Two experienced paediatric intensivists, who are blinded for the CDSS results, will analyse the electronic patient file (EPF) for SIRS criteria and thus establish our Goldstandard. All SIRS events recognized by the CDSS during the patient's stay are taken into account and will be compared with the established Goldstandard. The secondary goal of this trial is to evaluate the CDSS-results with the assessment of SIRS by paediatric doctors during their routine work on the PICU.

Clinical decision-support systems (CDSS) are designed to solve knowledge-intensive tasks for supporting decision-making processes. Although many approaches for designing CDSS have been proposed, due to high implementation costs, as well as the lack of interoperability features, current solutions are not wellestablished across different institutions. Recently, the use of standardized formalisms for knowledge representation as terminologies as well as the integration of semantically enriched clinical information models, as openEHR Archetypes, and their reuse within CDSS are theoretically considered as key factors for reusable CDSS. The investigators already successfully transferred their concept into a prototype and evaluated the practicability on clinical data sets and in close cooperation between the clinicians and the technical experts. To the author's knowledge, currently, there are no openEHR based CDSS approaches which have been implemented and evaluated with such complex and important clinical contexts. Hence, the first clinically evaluated CDSS based on openEHR was successfully designed. When enhancing the described approach and implementing a live system, it might support clinicians to identify the patient's course of disease at an early stage, which can lead to better outcome for the patient. Furthermore, the system can serve as a basis for integrating (cross-institutional) machine learning components that could facilitate dealing with other high-complex decision problems or revealing yet unknown disease patterns.