UPDEMIN: an efficient protocol for predicting patient upcoming situation, decision and intervention in wireless body sensor network

Abstract

In this article, we propose UPDEMIN, a novel approach for predicting UPcoming patient situation, DEcision Making and INtervention in wireless body sensor network. Inspired by the National Early Warning Score (NEWS) system, a new representation and usage of all the vital signs scores to create a diagnostic label (DL) representing the overall health situation. The formulation of the DL does not help only in accurate prediction of the upcoming health situation but provides a decoding mechanism to identify the reasons for deterioration and which vital signs are driving it. Simulations show high accuracy in predicting several upcoming situations ahead of time with 90% accuracy. Compared to other methods, UPDEMIN shows superior performance with less historical data or a smaller database, exceeding 91% prediction accuracy versus 66% for the closest method. The performance of the classification model has shown perfect precision and recall values, as demonstrated by the F1-score, which recorded 0.92. Moreover, the predicted situation can be decoded easily to identify which vital signs cause criticality, allowing the medical team to make the appropriate decision and prepare for the right intervention. The capability to decode the predicted health situation is unique and expected to have an insightful return on preventative intervention decisions.

Keywords

• Wireless body sensor network
• internet of things
• National Early Warning Score system
• machine learning
• decision tree
• prediction