Predictive Modeling in the Intensive Care Unit

Project Description

In the Netherlands, sepsis is the most common cause of mortality in the intensive care unit (ICU). Severe sepsis is present in around 30% of patients admitted to the ICU and can be caused by various processes. Central line-associated bloodstream infections (CLABSI) are among the most common hospital acquired infections (HAI) and constitute a high burden by increased morbidity, prolonged hospitalization, and increased health care expenditure. Central venous catheters (CVCs) are indispensable in the ICU for administration of high risk medication and parenteral nutrition. Even if optimal precautions (such as sterile CVC insertion, daily control of the CVC insertion site) are taken, CLABSI is still common, with an incidence of 1.1 per 1000 catheter days in the Netherlands. Different risk factors for development of CLABSI have been established, however it is not clearly understood which other factors may predispose to CLABSI. While there is debate about the most appropriate definition of a CVC infection, a straightforward registry of CLABSI with clear diagnostic criteria has been established in the Netherlands since 2010.

Early prediction of ICU patients developing CLABSI could assist healthcare professionals in decision making in daily practice. More precise prediction could reduce unnecessary removal of CVC, or in contrast take early precautionary measures by removing the CVC before bloodstream infection occurs, reduce the number of blood cultures taken and reduce the amount of unnecessary antibiotics started (which has a major impact in the context of antimicrobial resistance). Therefore, we aim to develop and implement a real-time algorithm for the prediction of CLABSI in ICU patients based on data that is readily available in the electronic health record (EHR).

To identify all common predictive factors concerning CLABSI in the ICU, a systematic review of the literature will be performed. The resulting factors will be combined with factors derived from expert opinion and discussed in a consensus group. Through a Delphi procedure, experts from the departments of Intensive Care, Infectious Diseases, Microbiology and Hygiene and Infection Prevention will reach consensus of the most important determinants predicting CLABSI in the ICU. These determinants will be used in the modelling phase, where we aim to develop a model for the day-to-day prediction of CLABSI in the ICU.

People

Ruud van Kaam

Ruud van Kaam

PhD student

Intensive Care Medicine, Radboudumc

Tim Frenzel

Tim Frenzel

Intensivist

Intensive Care Medicine, Radboudumc

Jeroen Schouten

Jeroen Schouten

Intensivist

Intensive Care Medicine, Radboudumc

Astrid Hoedemaekers

Astrid Hoedemaekers

Intensivist

Intensive Care Medicine, Radboudumc

Marcel van Gerven

Marcel van Gerven

Professor

Artificial Intelligence, Radboud University

Hans van der Hoeven

Hans van der Hoeven

Professor

Intensive Care Medicine, Radboudumc

Luca Ambrogioni

Luca Ambrogioni

Assistant professor

Artificial Intelligence, Radboud University