Investigative Analysis of Hospital Module In MIMIC-IV Database for Neonatal Patients

Abstract

The paper aims to investigate and perform mortality analysis of different neonatal data trends present in the ‘‘hospital’’ module of MIMIC-IV dataset. MIMIC-IV is an openly available medical dataset consisting of around 60000 neonatal patients. The hospital module stores digital health records of patients like laboratory tests performed, procedures or services provided by the hospital etc. Google Big query is used to access and filter the MIMIC-IV database. The data visualization is done by using Google Looker Studio. The results show that 98.6% of the admitted neonates were advised for blood tests. 40% of neonates could not survive as anticipated in laboratory tests by ‘‘abnormal’’ flag. The topmost tested lab item in case of neonates was pH’’. 47% of the neonates belonged to group‘‘neonates with birth weight greater than 2.49 kg having other problems’’ followed by group ‘‘premature newborns with significant problems The highest microbiological specimen tested for neonates was ‘‘Blood Culture’’ accounting for 45% of all specimens. It was seen from the analysis that ESCHERICHIA COLI’’ is the microorganism affecting neonatal mortality highest out of all. It was interesting to acknowledge from the analysis that 80% of the antibiotics given to non-surviving neonates fall into sensitive category. Hence, this analysis has highly contributed in finding the correlative features with respect to mortality from hospital module of MIMIC-IV neonatal dataset and will be useful for AI and medical researchers. It will also be supportive in the process of building a machine learning model for neonatal mortality prediction.