PHAS: An End-to-End, Open-Source, and Portable Healthcare Analytics Stack

Abstract

In today's binary world, digital health is of paramount importance, primarily due to the prevalence of IoT devices, upsurging health costs, growing elderly population, and shortage of clinical providers, to name a few. In light of these, providing efficient and smarter full-stack healthcare data analytics to manage and process healthcare data is a crucial topic from both academic and professional perspectives. The key goals of this full-stack healthcare data analytics are to detect health issues and promote human-being health proactively. The existing healthcare data analytics stacks are generally classified into commercial or open-source solutions. In designing a healthcare data stack, it is critical to offer the researchers a collaborative, modular, easy-to-use, cost/time-effective, reproducible, uniform, and shared- knowledge framework. Such healthcare stacks need to pave the way for researchers to focus on developing data analytics algorithms, and the underlying infrastructure should comply with longitudinal characteristics of healthcare data. Nonetheless, the existing healthcare data analytics stacks need holistically incorporate all the parameters mentioned. In this paper, we propose a novel healthcare data analytics stack called Open-Source Portable Healthcare Analytics Stack (PHAS) to address this issue. PHAS considers the mentioned features by fusing the merits of open-source and commercial solutions at the right place in its architecture. PHAS proposes a new shared-knowledge and time-series-aware framework, which enables researchers to perform health data collection, integration, storage, visualization, and analysis. Moreover, we demonstrate the capabilities of the PHAS framework by implementing an anomaly detection algorithm for heart rate and blood pressure anomaly detection using the Medical Information Mart for Intensive Care III (MIMIC III) dataset. We provide open-source PHAS for the community to integrate into their solutions.