Tool: NEO-DEER: A Web-Tool for Machine Learning-Based Risk Score Prediction

Version: 1.0

Published: 2022-01-15

Contributors: Nadir Yalcin, PhD
                      Merve Kasikci, MSc
                      Hasan Tolga Celik, Assoc Prof, MD
                      Karel Allegaert, Prof, MD
                      Kutay Demirkan, Prof, PharmD
                      Sule Yigit, Prof, MD
                      Murat Yurdakok, Prof, MD

Logo Designer: Nuri Beydemir

Description: Prospectively 11,908 medication orders from 412 NICU patients over 17 months were comprehensively analyzed by a pediatric clinical pharmacist. A machine learning-based adverse effect risk score was developed for these patients, with a total of 187 adverse effects determined objectively by a risk analysis (matrix). The positive predictive value and C index for this risk score are 0.893 and 0.918, respectively. It is estimated that adverse effects can be prevented before they occur, with the use of this free, user-friendly, online, non-registered, and high-performance web-tool, which predicts adverse effects in each patient admitted to the NICUs. According to this risk score, Youden Index obtained with the parameters of cardiovascular system drugs (3 points), endocrine system drugs (2 points), nervous system drugs (1 point), parenteral nutrition treatment (1 point), and circulatory system diseases (1 point). The optimal cut-off value was determined as 3 points.