Automated partial premature infant pain profile scoring using big data analytics

Abstract

Lack of valid and reliable pain assessment in the neonatal population has become a significant challenge in the Neonatal Intensive Care Unit (NICU). Current practice in the NICU involves the meticulous, time-consuming and potentially bias process of manual interpretation of pain scores. In an attempt to forego the manual scoring system, this thesis presents an initial framework to automate a partial pain score for newborn infants using big data analytics that automates the analysis of high speed physiological data. The design of the novel Artemis Premature Infant Pain Profile (APIPP) is proposed in this thesis. An ethically approved retrospective clinical research study was performed to calculate APIPP scores from premature infant data collected from the Artemis platform. Using the Premature Infant Pain Profile (PIPP) as the base gold standard scale, scoring techniques were automated to create data abstractions from the physiological streams of Heart Rate (HR) and Oxygen Saturation (SpO2). These were then brought together to compute an automated partial pain score (APIPP) that was based on gestational age, HR and SpO2. Through the retrospective clinical research study, and to evaluate the effectiveness and feasibility of automating the scale in the future, APIPP was retrospectively compared with the PIPP which was manually scored by nursing staff at The Hospital for Sick Children, Toronto. Furthermore, the characteristics in HR were also assessed in a thorough manner by preforming statistical tests to assess the resourcefulness of HR as a measure to identify a pain response. Future research will focus on the clinical validation of this work by carrying out prospective research to implement an algorithm based on the design proposed in this thesis that can be integrated into a clinical decision support system named Artemis.