Yenidoğan Hastalarda İlaç Kaynaklı Sorunlar

Abstract

The higher rate of drug-related problems (DRPs) in hospitalized neonates is associated with the diversity of prescribed drugs and differences in age, weight and diagnosis. The aims of this study were (i) to perform risk (probability and severity) analysis of drug-related problems [adverse drug reactions (ADRs), drug-drug interactions (DDIs), and medication errors (MEs)] detected in all treatment steps of patients admitted to the Hacettepe Children’s Hospital neonatal intensive care unit (NICU) between February 2020 and June 2021 using the risk matrix, (ii) to develop models that predict these problems before they happen with a newborn-centered approach using artificial intelligence, (iii) to reduce the alert fatigue and workload of clinicians by implementing optimal pharmacotherapy with these models. At the same time, it was aimed to develop another model that predicts the presence of a malnutrition marker extrauterine growth restriction (EUGR) in patients discharged from the NICU. A total of 412 patients were included, 6 (12.25%) of 49 different ADRs were detected in 98 (23.8%) patients were found to be high-risk based on objective risk analysis. However, potential DDIs were observed in 125 (30.4%) patients and clinically significant DDIs were observed in 38 (30.3%) of these patients. In addition, 4 (5.33%) of 75 different clinically significant DDIs were found to be high risk category. Considering the MEs, at least one type of MEs was found in 174 (42.2%) patients orignating from physicians (prescribing and monitoring), and 235 (57.0%) patients originating from nurses (preparation and administration). Patients who were included in the study prospectively randomized to train and test sets (70:30) in all models using a 10-fold cross-validation method with R Program (caret package). The AUC values demonstrated accurate prediction performance of the obtained models of ADRs, DDIs, MEs, and EUGR were found to be 0.862, 0.929, 0.920, and 0.884, respectively. As a result, it is estimated that these problems will be minimized with the use of web-tools developed for use in clinical practice with the obtained risk analyzes and high-performance prediction models to predict the presence of EUGR (http://www.softmed.hacettepe.edu.tr/NEO-DEER/), ADRs (http://softmed.hacettepe.edu.tr/NEO-DEER_Adverse_Effect/), DDIs (http://www.softmed.hacettepe.edu.tr/NEO-DEER_Drug_Interaction/), MEs (http://softmed.hacettepe.edu.tr/NEO-DEER_Medication_Error/).