Data Analytics for UMC Utrecht

Profile of our Client

The University Medical Center (UMC) Utrecht believes that everyone deserves the best possible healthcare and therefore continuously pushes itself to find the best and newest treatments. Constructing teams of professionals, patients, and partners, the UMC Utrecht wants not only to be able to account for the healthcare needs for now, but also for the future. They want to be able to provide innovative solutions for the medical issues of today and tomorrow.

 

The Challenge

UMC Utrecht launched a program ‘Applied Data Analytics in Medicine’ (ADAM) to explore the opportunities for data analytics in healthcare. One of the four challenges of the ADAM-project is at the Neonatal Intensive Care Unit (NICU) and more specifically, for babies with a gestational age less than 32 weeks. These babies have a high probability of obtaining sepsis during their stay at the NICU. Is it, therefore, possible to use big data as a solution to proactively treat or maybe even prevent sepsis? The NICU is in need of an algorithm that validates suspicions of sepsis and predicts the type of bacteria that will most likely be found in the blood.

 

The Solution

By combining multiple databases that are already present at the UMC Utrecht, such as the Research Data Platform, MetaVision (real-time data), UPOD (cell data) and HiX (patient information), a predictive algorithm is constructed by Finaps and SAS that takes all variables into account.  Finaps and SAS used SAS Enterprise Guide and SAS Enterprise Miner, to connect the databases and to construct an algorithm. This algorithm helps to predict whether the diagnosis of sepsis is justified and if so, which type of bacteria is likely to be present.

 

This algorithm will be the back-end of an embedded application developed by Finaps. This application will provide a detailed overview of the (values of the) predictive variables as well as the predictions itself for new suspicions of sepsis. The detailed overview of the variables gives the doctor insights in why that particular prediction is given. This is important since the algorithm should never be leading but rather used as a decision support. The doctor always makes the final decision in the treatment of sepsis.

 

The Result

The medical professionals now have a data-driven algorithm that can support their decision to start antibiotic treatment. The two significant results obtained by this solution are firstly, identifying un-justified suspicions which can lead to a reduction of the amount of unnecessary antibiotics given. Secondly, by predicting the type of bacteria, it is possible to directly apply the right antibiotic treatment. This greatly improves the long-term outcome of the preterm baby.