Exploratory Data Analytics for Healthcare

R. Lakshmana Kumar, R. Indrakumari, B. Balamurugan, Achyut Shankar, R. Lakshmana Kumar, R. Indrakumari, B. Balamurugan, Achyut Shankar

• Éditeur: CRC Press
• Année: 2021
• ISBN:
  9781000527056
• Édition: 1
• Sujet: Ciencia de la computación
• Sous-sujet: Visión y reconocimiento de patrones computacionales

Exploratory data analysis helps to recognize natural patterns hidden in the data. This book describes the tools for hypothesis generation by visualizing data through graphical representation and provides insight into advanced analytics concepts in an easy way.

The book addresses the complete data visualization technologies workflow, explores basic and high-level concepts of computer science and engineering in medical science, and provides an overview of the clinical scientific research areas that enables smart diagnosis equipment. It will discuss techniques and tools used to explore large volumes of medical data and offers case studies that focus on the innovative technological upgradation and challenges faced today.

The primary audience for the book includes specialists, researchers, graduates, designers, experts, physicians, and engineers who are doing research in this domain.

1 Visual Analytics Scopes and Challenges

Kalpana Hazarika, A. Ambikapathy, and Shobana R.

Galgotias College of Engineering & Technology

Amit Agrawal

ABES Engineering College

DOI: 10.1201/9781003050827-1

CONTENTS

1. 1.1 Introduction: Concept of Visual Analytics
   1. 1.1.1 Process of Visual Analytics
   2. 1.1.2 Need and Benefits of VA in Healthcare
2. 1.2 VA Technologies and Tools
   1. 1.2.1 General Features of Visual Analytics Tools
      1. 1.2.1.1 Data Visualization
      2. 1.2.1.2 Dashboards
      3. 1.2.1.3 Integration with Multiple Data Sources
      4. 1.2.1.4 Collaboration
   2. 1.2.2 Visual Analytics Tools
3. 1.3 Scope of VA in Different Sectors of Medical Science
4. 1.4 Challenges to Face
5. 1.5 Conclusion
6. References

1.1 Introduction: Concept of Visual Analytics

The volume of information exchange is growing exponentially in almost all sectors, and even in healthcare. This is a sector that deals with human beings, where the cost of a life is not replenished. It is a domain that certainly cannot be excluded and needs more accuracy in its data interpretation for accurate prediction of future or immediate course of actions. Hence, it has to work efficiently for its exact prediction and conclusion to generate the next step of actions during normal or critical situations.

Visual analytics (VA) is the solution to these complex data and gives a visual interactive information clustering-relevant statistics as per end-users’ needs. Every sector of healthcare needs a different set of information for its analysis and interpretation corresponding to its objectives. This demands nonhomogeneous data combination for analytical or visual analysis to represent the statistical model of the original data. VA merges computer-aided evaluation methods with communicative visualizations for better understanding and decision-making based on logical inferencing with the high volume of complex datasets [1]. In simple words, this is a methodological study supported by communicative visual interfaces. It integrates cognitive intellect with synthesizing efficacy of the enhanced methods to gain hidden understanding into multifaceted problems for decision-making. Hence, the key objectives of VA are to design operational tools and procedures to combine meaningful data and develop comprehension from huge, vigorous, confusing, and sometimes contradictory information. Also, its due responsibility is to deliver judicious, strong, and logical assessment from the discovered insights and communicate assessment effectively for action [1].

1.1.1 Process of Visual Analytics

As discussed, VA combines intuitive and graphical evaluation techniques with a close humanoid interaction to discover facts from unstructured information. Here, the automatic data evaluation performs a systematic process of scrutinizing, cleaning, transmuting, and demonstrating data for further analysis, whereas graphical evaluation techniques will have a communicative panel on the monitor for experts to choose data sources and display patterns for visual display, simple data interpretation, and accessibility to competent authorities [1].

But identifying the best-automated algorithm for healthcare data analysis is another challenge. Another important requirement for all related examinations is developing a strongly integrated solution for the successful integration of computerized evaluation procedures with suitable imagining and communication techniques. These analyses require a huge data set; hence, yet another prerequisite of visual or automatic analysis lies in integrating heterogeneous data sources. Automated Analysis through data mining algorithms plays an important role in data cleaning, normalization, grouping, and integration. The primary goal of knowledge discovery in healthcare is to pull out meaningful facts from large databases. The knowledge discovery in databases is nothing but a programmed, exploratory analysis and modeling of vast data storehouses.

The VA control panel has various engines involving visual graphics to represent the analytical results. The results are displayed on the screen after computational algorithms work. VA discovers actionable commercial facts for business intelligence support. To do so, it takes advantage of deep learning technologies to detect uncovered tendencies and characteristics of large data.

1.1.2 Need and Benefits of VA in Healthcare

For a decade, VA has played an instrumental role in healthcare data analysis to give a big momentum to the overall performances of the sector. The governance of health systems is now open to adopting new tools and technologies. It has also recognized that for the survival of healthcare institutions, data analytics is an integrated part of the sector.

The factors for which the acceptance pace of analytics is accelerating in healthcare are stated under the following:

• Using Electronic Health Record (EHR), a complicated set of structured and unstructured data that VA can use to quickly demonstrate the results in a sensible way for further course of action. The data may include a patient’s medical history maintained over time, including demographic data, development notes, complications, and treatments to provide need-based services.
• If technology needs future improvement in the healthcare system, then the EHR data used by doctors should be empowered by the emerging power of analytics and machine learning. Using modern analytical techniques, superior information can be provided to doctors for patient care. Meaningful inferences and predictions can be pulled out from it, like the probability of a patient’s successive risk to vital organ failure or catastrophic failure based on blood pressure reading, significant signs of illness, test results, genetic background, and recent clinical research data [2].
• Cloud computing infrastructure allows dynamic data processing with massive data. The extensive application of remote server accessibility in healthcare endorses policymakers to efficiently optimize operational management cost of treatment and offers individualistic care plans for better outcomes. It holds the potential to produce vast data compared to warehouses a decade ago [3].

1.2 VA Technologies and Tools

These tools and technologies are playing a key role in the sustainable progressive development of small and big enterprises. It helps in exploring and visualizing the current status of any business operations for long-term strategic planning and policymaking. By identifying important patterns and deriving key insights, it helps to take actionable decisions for accomplishing set goals. To perform these analyses, VA plays a significant role in a large volume of data handling from various sources and processing to give a need-based visual presentation [1].

1.2.1 General Features of Visual Analytics Tools

1.2.1.1 Data Visualization

It encompasses scientific visualization – a multidimensional scientific data representation to support clinical experts to identify the pattern from biological data, information visualization – the study of communicative pictorial representations of descriptive data to improve human intellectual, and visual analytics – which combines them both with an emphasis on analytical reasoning through an interactive visual interface [4].

The professional data visualization engineers must have the following technology expertise for developing interactive visual interface tools:

• Basic mathematics: understanding of linear equations, geometric procedures, and trigonometric function.
• Graphics: knowledge on canvas, vector-based graphics, Web Graphics Library, computer graphics, and network theory for creating or developing rich 2D and 3D interactive graphics.
• Engineering algorithms: arithmetical, basic, and general layout algorithms.
• Data analysis: scrutinizing, cleaning, transmuting, and demonstrating data for inherent knowledge discovery, drawing conclusions, and reviewing operational strategies.
• Design aesthetics: designing ethics, visual and color judgement, collaboration, thoughts, etc.
• Visual basis: visual coding (visual programing languages such as R, Scala, and Python Programming Languages), visual analysis, and graphical interaction.
• Visualization solutions: timely decisions through an in-depth analysis of raw data that helps in visualization of common business scenarios.

1.2.1.2 Dashboards

These are control panels that show the major implementation of data visualization. It generates a dynamic pictorial graphical presentation of specific data for uninterrupted updates to keep us informed. These panels are incredible in providing a quick view on a need-based analysis if one needs to see it at a glance. By using dynamic panels, the status can be observed as and when required. This feature allows the decision-makers to immediately address the issue with a suitable solution. Apart from inventory control, it is used for run-time KPI monitoring, steering other responsibilities, and ensuring effective utilization of personnel and resources. The major benefits are summarised as quick and easy access, forming of better decisions, meeting growing business demands, improving an inefficient reporting system, and displaying reliable data.

However, healthcare professional dashboards are as follows:

• Clinical dashboards: This dashboard integrates correlated data into a single point of display or “dashboard” to optimize worker efficiency, accelerate medical judgements, rationalize workflow, and minimize human error in clinical practice. Here, clinical data are regularly recovered, recorded, examined, and shared to intended users for better care of patients. This also helps in tracking public health eruptions due to any flu epidemic to plan employment needs effectively by displaying the current loadable capacity of patients for emergency rooms, bed occupancy rates, nursing unit tallies, etc. [5].
• Hospital dashboards: It assists hospitals to keep a close and vigilant eye on significant functions of all departments for workflow and patient care quality improvements. The main objective of this dashboard is to follow patient satisfaction, optimum doctor staff-to-patient ratio, optimize the cost of care, and take care of required logistics. For this, it maintains a KPI structure that conducts surveys on the following parameters: Patient Satisfaction, Operational Workload, Operation Costs, Patient Volume, and Patient Experience. This includes surveys like patients admission, average waiting time, doctors’ treatment plan, confidence in treatment, the average cost per patient and satisfaction, cost incurred by the department for patients’ care, patients per doctor, number of doctors, etc. [6].
• Patient dashboards: This healthcare dashboard displays key information relating to patients’ overall health. With this health summary dashboard, viewers can observe the following information: patients’ vital signs, i.e., heart rate, temperature, and blood pressure, patients sleeping duration and the quality of sleep, daily step goals. It also helps to follow the aggregate duration of patient stay and laboratory waiting hours. Further, it may help physicians decide where to focus their treatment and preventive care efforts [7].
• Physician dashboards: This assists medical professionals to follow patient satisfaction, patient count, and new patient registration updates as well as track the metrics like the volume of cases treated, probable mortality, and reporting of adverse events, along with quality and process like significant parameters. To provide on-time treatment, appropriate appointment setup, lesser waiting time, increase patient safety, and reduce patient dissatisfaction, it is important to analyse healthcare plans for physician’s allocation. By improving the ratio of patients per physician to 100:1, healthcare institutions can deliver better care by streamlining physician allocation through this dashboard.
• Quality and risk management dashboards: It supports in keeping a close eye on complex compliance issues, managing patient welfare information, hospitals recommended treatment count, and comparing hospital standards with in-house performance [8].

1.2.1.3 Integration with Multiple Data Sources

The act of merging data from several sources into a unified set is called data integration. During this process, data is cleaned and transformed to enable accurate analysis. Through data integration, dissimilar data sources are reviewed together to provide invaluable business intelligence to support better business decision-making. The common sources of the medical database are Healthcare Data Warehouse, NoSQL, Cloud Services, etc. Healthcare data is collected to improve the cost and quality of healthcare and to generate meaningful insights for patients and researchers [9]. In this case, the visual analysis tool plays a vital role in communicating with all of the different data sources and extracting the desired information through visual presentation.

1.2.1.4 Collaboration

This feature allows multiple healthcare professionals to view and work with the same analysed data set, which helps to select evaluation methodology to work in a group. It facilitates sharing one’s analysis with others, enables to give inputs on their dissimilar concepts, and helps holds conferences in the VA platform for better understanding of the case and summarisation. Collaboration leads to the following benefits in the world of healthcare [10].

• Enhances confidence in data quality, reliability, and balance: collaboration increases the sample size; sharing helps in enhancing data quality. Inferencing the pattern of a particular case from a large data sample gives a more accurate prediction than using a small data sample. If a large scale of data is indicating the same issue, naturally, it enhances the dependability on data to achieve more accuracy in analysis.
• Enables better healthcare products: Accurate analysis from massive data gives better product development. It assists de...