Resources

Bringing models to newborns

Bringing models to newborns – Apps to predict weight change and risk of hyperbilirubenima during first days of life



http://neoweight.rudraya.cloud


Dr. Gilbert Koch, University Children’s Hospital Basel, Switzerland (presenter at ACoP)

Prof. Dr. Julia E. Vogt, ETH Zurich, Switzerland

Prof. Dr. Sven Wellmann MD, University Children’s Hospital Regensburg, Germany

Prof. Dr. Marc Pfister MD, University Children’s Hospital Basel, Switzerland


Introduction

Duration of hospitalization of newborns and their mothers was reduced to 2-3 days in the last couple of years. As a result of shortened in-house monitoring of bilirubin and weight, the number of readmissions is dramatically increasing. This indicates the clear need to reliably predict the risk for complications such as clinically relevant weight loss or hyperbilirubinemia during the first week life.


Decision support tools

Pharmacometrics methods (PMX) based on dynamical systems and non-linear mixed effect modeling, as well as machine learning techniques such as random forest (ML) were applied to develop our decision support tools.


Tool to predict and manage weight change Every newborn is experiencing a weight loss during the first 2-4 days of life due to fluid loss and/or negative energy balance. It should be noted that a weight loss greater than 8-10% is considered to be dangerous in newborns as such weight loss can be associated with serious short- and long-term complications. We developed a PMX based tool utilizing clinical data from 1335 neonates to predict weight changes over 7 days based on individual key risk factors (newborn and mother). The tool was validated based data from additional 300 neonates.


Tool to predict and manage bilirubin change Neonatal jaundice due to hyperbilirubinemia affects 60% of all newborns and is the most frequent reason for readmissions in the first month of life. Almost 10% of newborns develop critical hyperbilirubinemia, which can lead to serious long-term developmental complications. We developed a PMX and ML based tool utilizing clinical data from 362 neonates to predict the probability of the onset of clinically relevant hyperbilirubenima in the next 48 hours based on the identified key risk factors (newborn).


Application of developed decision support tools

NeoWeight App: Current users: Used in selected Hospitals in Germany and Switzerland. Qualification process: Validation with external data set, prospective study and certification planned.


NeoBilirubin App: Current users: Used in our hospital. Qualification process: Validation with external data set; prospective study planned; CE certification process initiated.