Data Democracy

At the Nexus of Artificial Intelligence, Software Development, and Knowledge Engineering

Feras A. Batarseh, Ruixin Yang

• Publisher: Academic Press
• Year: 2020
• Print ISBN: 9780128183663
• eBook ISBN: 9780128189399
• Topic: Biological Sciences
• Subtopic: Biotechnology
• Index: Biological Sciences

Data Democracy: At the Nexus of Artificial Intelligence, Software Development, and Knowledge Engineering provides a manifesto to data democracy. After reading the chapters of this book, you are informed and suitably warned! You are already part of the data republic, and you (and all of us) need to ensure that our data fall in the right hands. Everything you click, buy, swipe, try, sell, drive, or fly is a data point. But who owns the data? At this point, not you! You do not even have access to most of it. The next best empire of our planet is one who owns and controls the world's best dataset. If you consume or create data, if you are a citizen of the data republic (willingly or grudgingly), and if you are interested in making a decision or finding the truth through data-driven analysis, this book is for you. A group of experts, academics, data science researchers, and industry practitioners gathered to write this manifesto about data democracy. - The future of the data republic, life within a data democracy, and our digital freedoms - An in-depth analysis of open science, open data, open source software, and their future challenges - A comprehensive review of data democracy's implications within domains such as: healthcare, space exploration, earth sciences, business, and psychology - The democratization of Artificial Intelligence (AI), and data issues such as: Bias, imbalance, context, and knowledge extraction - A systematic review of AI methods applied to software engineering problems

Section II

Implications of a data democracy

Outline

6. Data openness and democratization in healthcare: an evaluation of hospital ranking methods

7. Knowledge formulation in the health domain: a semiotics-powered approach to data analytics and democratization

8. Landsat's past paves the way for data democratization in earth science

9. Data democracy for psychology: how do people use contextual data to solve problems and why is that important for AI systems?

10. The application of artificial intelligence in software engineering: a review challenging conventional wisdom

6

Data openness and democratization in healthcare

an evaluation of hospital ranking methods

Kelly Lewis ¹ , Chau Pham ¹ , and Feras A. Batarseh ^{2 1} College of Science, George Mason University, Fairfax, VA, United States ²Graduate School of Arts & Sciences, Data Analytics Program, Georgetown University, Washington, D.C., United States

Abstract

The democratization of data in healthcare is subsequently becoming one of the most impacting hurdles in patient service, hospital operations, and the entire medical field. Controversy among healthcare data has brought challenges to many medical case studies including legal rights to data, data ownership, and data bias. The experiment performed in this chapter shows thorough hospital quality metrics and provides a new ranking system that compares hospitals and identifies which ones are best for a potential patient, based on their needs and medical status.

Keywords

Data; Healthcare; Hospitals; Patient; Quality

The past, like the future, is indefinite and exists only as a spectrum of possibilities.

Stephen Hawking

1. Introduction

Nowhere is scientific stagnation more apparent than the medical field. Abundant resources are devoted to medical research, yet increased spending has not resulted in tangible outcomes [1]. The National Institute of Health estimates that “researchers would find it hard to reproduce at-least three-quarters of all published biomedical findings” [2]. Many publications in the medical industry are now reliant on large and complex datasets, often yielding contradictory and “irreproducible results” [2]. As a result, “massive investments in basic biomedical and molecular research have resulted in negligible dividends for patients” [3].

The lack of quality publications is due to many factors, including increased administration costs, excess regulation, and an overbearing presence by the pharmaceutical industry. Bigger budgets are used to cover enlarged administration and bureaucratic costs, not scientists or lab gear [4]. Larger bureaucracies are symptoms of increasingly complex regulatory system, much of which emerged as a reaction against casualties in clinical trials [5]. Regulations, often self-imposed by institutions, have increased time costs for researchers. In 2000, the average cost to develop a new drug was $802 million—time costs accounted for half of all research and development expenses [4].

Investigators spend large sums of their time performing administrative and managerial tasks, partly attributed to a high turnover rate within clinical research organizations. Protracted time lengths also result in less evidence-based science as companies rush to get their drugs approved. Swelling costs have led to a reliance on pharmaceutical companies funding studies. Pharmaceutical companies are businesses with profit, not scientific advancement; subsequently, biases and underreporting are regularly underreported in industry-sponsored research. In a 2008 study of antidepressant drug trials submitted to the FDA, “97 percent of the trials…yielded positive results” [6]. The study reported that “39 percent of the studies were found to have negative or questionable results” [6]. The lack of negative trial outcomes is not due to superior research models; the number of FDA drug approvals from industry-sponsored research has steadily declined [3].

A dearth of quality publications cannot purely be blamed on any single institution or factor. Medical research is bound to face discrepancies because of its inherent complexities. Computer scientists, for example, can rely on the fact that a computer will function as programmed. Biologists and chemists, however, rely on cells and molecules that often change in unexpected ways. Errors most frequently occur in data analysis, either intentionally or unintentionally [7]. Measurement errors, unvalidated data, fraud, and misinterpretation of data can result in skewed analysis [2]. A gap in appropriate data analysis is one of the contributing factors to “a general slowdown in medical progress” [2].

Flaws in data analysis have been well-documented by researchers, yet institutions have been unable to propose long-term solutions. Publications can improve on studies by increasing transparency standards. Data and conclusions from government-sponsored trials can only be available to the public if they are published. Yet, “less than half of trials initiated by academic researchers are published.” [8]. One of the main obstacles to publication was “ongoing data analysis” or “manuscript preparation” [8]. Many of these obstacles can be aided with Open Data and Open Science within healthcare. How would healthcare transform within the data republic? This chapter aims to design and explore that.

2. Healthcare within a data democracy—thesis

While healthcare research is typically privatized to prevent plagiarism or release of sensitive information, experimental variables should be available to the public to catalyze the research cycle, allow for correct reproductions and alterations, and provide scientist and citizens of a data republic a complete understanding of the experiment or research.

3. Motivation

Data privacy and security measures include removing identifying attributes of patients. Removing this part of data to comply with the De-Identification Act can cause some negative implication on the outcomes during analysis [9]. Security a...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Dedication
• To: Aaron Swartz—the creator of the Open Access Manifesto.
• Contributors
• A note from the editors
• Foreword
• Preface
• Section I. The data republic
• Section II. Implications of a data democracy
• Index