Abstract

It is difficult to distinguish ADR from reactions associated with organ dysfunction/immaturity or genetic mutations. In this prospective cohort study, clinical pharmacists assessed each ADR using neonatal-specific severity and probability scales. A machine learning-based risk score was then developed to predict the presence of ADR in neonates. In 98/412 (23.8%) (56.3%; male) neonates, 187 her ADRs (0.42 ADR/patient) associated with 49 different drugs (37.12%) were identified. Enoxaparin, dexmedetomidine, vinblastine, dornase alfa, etoposide/carboplatin, and prednisolone were identified as high-risk drugs. The independent variables included in the risk score for predicting the presence of ADR follow the random forest importance criteria. Systemic hormones (2 points), cardiovascular drugs (3 points), cardiovascular diseases (1 points), nervous system drugs (1 point), parenteral nutrition therapy (1 point) , (cut-off value). A neonatal-specific high-performance risk score is expected to identify and prevent high-risk ADRs in neonates before they occur. In addition, this web tool can be used to increase clinician awareness of these drugs, allowing mitigation strategies (drug changes, doses, duration of treatment, etc.) to be suggested based on the benefit-harm relationship between neonates and suspected drugs can consider central approach.