Journal:Health informatics: Engaging modern healthcare units: A brief overview

Abstract

With a large amount of unstructured data finding its way into health systems, health informatics implementations are currently gaining traction, allowing healthcare units to leverage and make meaningful insights for doctors and decision makers using relevant information to scale operations and predict the future view of treatments via information systems communication. Now, around the world, massive amounts of data are being collected and analyzed for better patient diagnosis and treatment, improving public health systems and assisting government agencies in designing and implementing public health policies, while also instilling confidence in future generations who want to use better public health systems.

This article provides an overview of the |Health Level 7 FHIR architecture, including the workflow state, linkages, and various informatics approaches used in healthcare units. The article discusses future trends and directions in health informatics for successful application to provide public health safety. With the advancement of technology, healthcare units face new issues that must be addressed with appropriate adoption policies and standards.

Keywords: health informatics, public health, information systems, health policy, public health systems

Introduction

Machine learning is the fastest-growing topic in computer science today, and with it a health informatics implementation of ML is one of the more difficult problems to solve. [1, 2]

Emerging economies are increasing their investments in healthcare, which makes sense and encourages health professionals to adopt sound frameworks and regulatory standards, as well as health IT, to improve the quality and efficacy of care. [3] In this expanding field, new age occupations can be established. This new field has the potential to be a lucrative career path in the future. With a clear flow of information across many medical subsystems, adoption of electronic health record systems (EHRs) will improve the health care system going forward. [4]

Big data is frequently employed in the field of health informatics, as new data is constantly pouring into the system, requiring analysis and interpretation in order to make rational decisions. [5, 6] This big data has ushered in a new era for healthcare companies to improve decision-making through the comprehensive integration of data from a range of sources, allowing for much faster and more effective decision making. [7] As such, within and outside of the medical industry, computational health informatics has become an emerging field of study. [7–9]

In recent years, the healthcare industry has seen a rapid growth in medical and healthcare data, which can be used to improve facilities and public healthcare utilization and implementation using novel treatment and diagnosis methodologies. In turn, this more efficient use of healthcare data gives patients confidence in using the best public healthcare services available and aids governments in developing better healthcare policies. [10]

In today's increasingly complex social and economic environment, at hand is the vital issue of improving quality of offered healthcare services while lowering prices. This is largely what health informatics has attempted to solve. The major purpose of health informatics is to increase our understanding of medicine and medical practice by using real-world medical data. In the scope of healthcare, health informatics is practically a blend of information science and computer science. [15]

At the core of health informatics has historically been a collection of computerized systems for assisting patient analysis and diagnosis. More recent technologies have emerged that make it even easier for clinicians to make better healthcare decisions. [11, 12] As health informatics continues to evolve, it promises to improve public health activities through the advanced application of information and communication technologies (ICT). [13] ICTs have been shown to help healthcare systems increase productivity, which has resulted in significant cost savings in operations and service delivery. For administrative and healthcare objectives, ICTs have already proven to be quite effective. Additionally, new prospects for new medical equipment and systems are opening up as ICTs become smaller, quicker, wireless, and remotely controlled.

The internet and web have recently brought up new possibilities for increasing the response time of healthcare services, while also lowering costs. It is clear that we are in the early stages of a new era that will fundamentally alter the way healthcare services are provided. This will help us acquire the public's trust in using high-quality healthcare services. However, new e-Health services and technology must still be researched, developed, promoted, and disseminated with significant effort. With the COVID-19 pandemic presently sweeping the globe, increasing ICT use has demonstrated that healthcare can and will become more contactless in the future, with fresh means of treating patients and providing healthcare services emerging. This is a popular yet difficult research subject since it necessitates interdisciplinary competence. [14]

Additionally, as big data continues to increasingly find its way into healthcare, additional challenges exist in the effective use of big data within ICT frameworks. For example, big data in healthcare is intimidating not only because of its sheer magnitude, but also due to the variety of data types and the pace with which it must be managed. To gain people' trust and give quality healthcare services, all health service providers are now putting in extra effort to use the most up-to-date technologies to effectively use big data to provide quality health services and advanced treatments.

Various requirements drive innovation in this industry, such as finding appropriate accommodation with standardization and coordinating the acquisition and implementation of newer healthcare systems and services on a national/international level. With COVID-19 still threatening disruption in the healthcare sector, investments in this sector are gaining steam with new-age healthcare units in many nations, and growing economies such as India and China will continue to play a vital role in providing quality healthcare services to its citizens in the future. At the same time, those new-age healthcare units and systems will aid in dramatically lowering costs, making public healthcare systems more dependable, and instilling citizens' confidence in using inexpensive, high-quality healthcare.

Related work

"Big data" is a term used to describe a significant volume of data that is collected and stored yet has outgrown standard data management and analysis solutions. Solutions like Hadoop and Spark, according to Roger Fyre and Mark McKenney, have arisen to solve some of these big data concerns. [16] For example, researchers have used Hadoop to implement a variety of parallel processing algorithms to efficiently handle geographical data. [17, 18] Multistage map and reduce algorithms, which generate on-demand indexes and retain persistent indexes, are the end result of these techniques. [19]

Other techniques such as predictive analytics and data mining have also been employed. Much of the current work on predictive analytics, particularly in clinical contexts, is aimed at improving health and financial outcomes, which will aid in making better decisions. [20] Data mining, which is defined as the processing and modeling of huge amounts of medical/health data to identify previously unknown patterns or associations, is another important machine learning approach. [21, 22] Data mining has been used, for example, in the collection of data for diseases such as cancer and neurological disorders in order to improve disease prognosis. [23, 24] Cancer detection and diagnosis, as well as other health-related issues, have been made possible because to these breakthroughs. [25] Machine learning is also crucial in the testing and development of various models that take into account clinical and other important medical characteristics for decision making.

Deep learning is now also being used to solve more difficult problems in the arena of health informatics. [26, 27] For example, advances in medical imaging and its data management have made positive contributions to decision making. Today, medical imaging incorporates capabilities such as image segmentation, image registration, annotation, and database retrieval, holding greater promise for decision makers. As such, new deep learning and machine learning models can be employed with medical imaging for speedier decision making. [26] However, this means researchers in the fields of data science, machine learning, and deep learning remain in high demand for developing effective algorithms that adapt to changing data.

Holzinger et al. [28] examined many approaches to developing an explainable prediction model for the medical domain. Prediction explanations can be useful in a variety of situations, including teaching, learning, research, and even the courtroom. Similarly, the demand for interpretable and explainable models is growing in the medical field. However, these models must be able to re-enact the decision-making and knowledge-extraction processes. Ribeiro et al. (29) have emphasized this requirement, discussing how machine learning models are essentially black boxes. Understanding the reasons for predictions can help to build trust, better assess model performance, and construct better, more accurate, and correct models by providing insights into the model. They propose the LIME algorithm [29] for explaining predictions of any model. Similarly, though dealing with neural machine translation, the proposed model of Bahdanau et al. [30] can be used in a variety of other applications such as healthcare.

Introduction to HL7 FHIR architecture

In the last two decades, EHRs have been widely implemented in the United States to improve healthcare quality, increase patient happiness, and reduce healthcare costs. [31–33] As growing countries such as India, China, and Bangladesh experiment with innovative ways to establish EHR systems, they will significantly aid in the development of effective public health systems in those countries. In all cases, at the core of most effective EHRs is Health Level 7's (HL7's) Fast Health Interoperability Resources (FHIR) architecture.

The basic idea behind HL7's FHIR (pronounced “fire”) was to create a set of resources and then create HTTP-based REST application programming interfaces (APIs) to access and use these resources. FHIR uses components called "resources" to access and perform operations on patient's health data at the granular level. This feature distinguishes FHIR from all other standards because it was not present in any earlier version of HL7 (v2, v3) or the HL7 clinical document architecture (CDA). [35]

The fundamental building blocks of FHIR are the so-called resources, which are generic definitions of common health care categories (e.g., patient, observation, practitioner, device, condition). For data interchange and resource serialization, FHIR employs JavaScript object syntax and XML structures. FHIR not only supports RESTful resource exchange but also manages and documents an interoperability paradigm.

FHIR has grown in popularity and is being increasingly used by the healthcare industry since its inception. In 2018, six major technology companies—including Microsoft, IBM, Amazon, and Google—vowed to remove barriers to healthcare interoperability and signed a statement mentioning FHIR as an emerging standard for the interchange of health data. With the incorporation of Substitutable Medical Applications Reusable Technologies (SMART), a platform for interoperable applications [34], FHIR can be expected to attract even more attention to digital health tools in the future. As is, the use of FHIR for medical data transmission has the potential to deliver benefits in a wide range of disciplines, including mobile health apps, EHRs, precision medicine, wearable devices, big data analytics, and clinical decision support.

The primary goal of FHIR is to reduce implementation complexity while maintaining information integrity. Furthermore, this new standard integrates the benefits of existing HL7 standards (v2, v3, and CDA) and is projected to overcome their drawbacks. FHIR enables developers to create standardized browser applications that allow users to access clinical data from any healthcare system, regardless of the operating systems and devices used. Figure 1 represents the general architecture of FHIR. [35]

Figure 1. General architecture of the Fast Health Interoperability Resources (FHIR) standard [35]

FHIR for patient access to medical records

FHIR is an HL7 standard for electronically transferring healthcare information. The Centers for Medicare and Medicaid Services (CMS) Interoperability and Patient Access final regulation, announced in 2020, mandates all CMS-regulated payers to use FHIR version 4. Unlike earlier releases, the fourth iteration is backward compatible, ensuring that software suppliers' solutions will not become obsolete when a new FHIR version is released.

The FHIR standard defines a collection of HTTP-based RESTful APIs that allow healthcare platforms to exchange and share data in XML or JSON format. FHIR offers mobile apps, which users can obtain from the Apple App Store or Google Play in order to access their medical records and claims data.

FHIR's basic exchangeable data piece is known as a resource. Each resource is formatted similarly and contains roughly the same amount of data. Each resource offers information about patient demographics, diagnosis, prescriptions, allergies, care plans, family history, claims, and so on, depending on the kind. They span the complete healthcare workflow and can be used independently or as part of a larger document.

Each resource is given a unique ID, and many health systems, insurers, patients, and software developers can access the underlying data element using an API. Figure 2 represents the data layers and resources of FHIR. [35]

Figure 2. FHIR data layers and resources [35]

FHIR resources

A resource is the smallest discrete concept that can be independently maintained and is the lowest feasible unit of a FHIR-based transaction. [36] As a result, a resource is a known identity that provides useful data. Each resource has distinct bounds and differs from all others. A resource should be provided in sufficient depth to specify and enable the process's medical data interchange. The FHIR community has specified over 150 resources to date, according to the most recent FHIR version (R4). [37]

There are five key categories in which these resources can be found:

1. Administrative: location, organization, device, patient, and group
2. Clinical: CarePan, diagnostics, medication, allergy, and family history
3. Financial: billing, payment, and support
4. Infrastructure: conformance, document, and message profile
5. Workflow: encounter, scheduling, and order

FHIR is fast gaining popularity due to its dynamic properties. FHIR is projected to quickly become a symbol for clinical data interchange in the healthcare industry.

Workflow description

Workflow is a critical component of healthcare; orders, care regimens, and referrals drive the majority of activity in inpatient settings, as well as a significant amount of activity in community care. FHIR is concerned with workflow when it is necessary to share information about workflow state or relationships, when it is necessary to coordinate or drive the execution of workflow across systems, and when it is necessary to specify permissible actions, dependencies, and behavior requirements.

Workflow state and relationships

FHIR does not have to be used for workflow execution. Orders, care plans, test findings, hospital admissions, claim payments, and other documents can all be exchanged utilizing FHIR resources without the need for a FHIR transaction to solicit fulfillment of those orders or request payment of those claims. Because it necessitates a greater level of standardization, interoperable support for workflow execution is a more advanced FHIR activity. Interoperable workflow execution necessitates the standardization of processes, roles, and activities across multiple systems, rather than just the data to be exchanged.

Even if FHIR is not used for workflow execution, there is still a requirement to standardize workflow data elements: how does an event or a result point to the order that allowed it? How are parent and child steps tied together? How does a care plan know which protocol it is following?

FHIR distinguishes three types of resources engaged in activities: requests, events, and definitions. Each of these categories is associated with a “pattern.” Resources in that category are encouraged to follow their specific pattern. These patterns provide conventional elements that are common to the majority of resources in each category. Work groups are anticipated to align with common domain behavior, and requirements as more authoritative than “desired” architectural patterns, and as such, strict conformance is not necessary. When a pattern capability is assessed to be “not common, but nonetheless relevant” for a given resource, it may be supplied through extensions rather than core parts. Figure 3 represents the workflow relations of the FHIR standard. [38]

Figure 3. Workflow relations with the FHIR standard [38]

Overview of health informatics

Health informatics involves more than merely automating routine tasks. With contemporary technology developments in machine learning and deep learning, it is possible to redesign systems using methodologies that were previously impossible or not even considered. [26] Machine learning and deep learning are computationally expensive, however, though they can now be handled by the latest IBM POWER9 processors with GPU capabilities, which was previously impossible because the data was not available in electronic form and the number of possible symptoms/incident patterns was too big to manage. Early detection of patterns that can anticipate what kind of treatment or diagnosis can be offered to such patients has improved dramatically. [39]

In the future, modern healthcare units will make use of such a framework for successful delivery of treatments for society as a whole, making effective use of data obtained from such systems by extracting insights that assist decision makers such as doctors, hospital owners, and health policymakers. However, classification and regression tree (CART) components are required for successful clinical decision support. For example, vital signs must be clearly classified to be applied to prediction. [40, 41] From there, the prediction of a patient's breathing rate obtained from sensors acts as an example of regression. [42] In the future, this sort of CART-based Bayesian inference will be used to make better predictions in the domain of health informatics.

However, several not-so-commonly employed principles will have to first be implemented in informatics systems in order to answer some the most pressing future difficulties in health informatics:

• Multi-task learning: Traditional machine learning frameworks consider only one learner attempting to solve a single task. However, in many applications, there are multiple tasks that label the same data instances differently. When the tasks are related, the information learned from each task can be used to improve learning of other tasks. Learning relevant tasks concurrently, rather than learning each task independently, is thus advantageous. Multi-task learning makes use of the intrinsic relationships between multiple tasks to improve generalization performance. It benefits all tasks by leveraging task relatedness and shared information across relevant tasks. [43]
• Transfer learning: Traditional machine learning technology has had a lot of success and has been used in many practical applications, but it still has certain limits in some real-world settings. Machine learning works best when there are a lot of labeled training cases with the same distribution as the test data. In many cases, however, gathering sufficient training data is costly, time-consuming, or even impossible. Semi-supervised learning can help to alleviate this difficulty by removing the requirement for large amounts of labeled data. A semi-supervised approach typically requires a small amount of labeled data and a large amount of unlabeled data to improve learning accuracy. However, in many cases, unlabeled instances are difficult to collect, making the resulting traditional models unsatisfactory. [44]
• Multi-agent hybrid systems: Multi-agent systems are networks of interconnected autonomous agents in which the behavior of neighboring agents influences the dynamics of each agent. Because of the increasing importance of multi-agent systems, there is a growing interest in coordination control to ensure consensus, flocking, containment, formation, rendezvous, and so on. To better understand multi-agent coordination, a variety of dynamic models of agents have been developed over the last two decades. Furthermore, many mathematical methods are used in the analysis and control of multi-agent systems. For more information, see research by Qin et al. [45] and the references within the research of Zheng et al. [46]
• Representation learning: Patient-specific data such as vital signs, medications, laboratory measures, observations, clinical notes, fluid balance, procedure codes, diagnostic codes, and so on are all included in modern EHR systems. Clinicians originally employed the codes and their hierarchies, as well as their associated ontologies, for internal administrative and invoicing functions. Recent deep learning algorithms, on the other hand, have attempted to project discrete codes into vector space, identify intrinsic commonalities between medical concepts, more accurately depict patients' health, and perform more precise predicting tasks. Word embedding and unsupervised learning have been used to examine medical concepts and patient representations in general. [47]

Health informatics has a number of long-term benefits in terms of research and healthcare delivery that can be used to create a sustainable ecosystem. ICTs aid in the enrichment of relevant data for analysis and decision making by health professionals. Following the COVID-19 pandemic, new-age healthcare units will arise, with increased investment and research spending, making public healthcare more accessible. As a result, solutions are required to manage the massive amounts of data created by medical equipment and healthcare systems, allowing for effective storage and retrieval in real-time data analysis and decision making.

Informatics approaches

To make electronic health data more easily usable for research, recent publications have identified the need for effective adoption and use of standards, essential data and research services, clear and consistent policies regarding data access and use, and transparent and effective governance structures. [48, 49]

To achieve data quality criteria, electronic health data utilized in research frequently require standardized ontologies, additional contextual information, field transformations, and missing or contradictory data to be handled. [50] For research-related data or functions, such as cohort identification and repeated extracts of source data over time, system development is frequently required. [49] Organizations with expertise utilizing and enhancing their health IT infrastructure for research have shared their lessons learned in these areas, adding value to organizations with similar goals but less experience or resources. [51]

For example, when preparing data for research use, organizations must understand the clinical context and structure of electronic health data, just as they do for other data uses such as decision support or population health. Individual researchers and data analysts can be relieved of their load by informatics support that spans research and operational usage of data. It is vital to evaluate and develop informatics tools and approaches by establishing processes that allow for coordinated governance and decisions informed by research users.

Investing in infrastructure to enable the use of electronic health data for research has also been shown to be beneficial to researchers by providing them with the necessary tools and expertise, to patients by providing clinical trial participation opportunities, to clinicians by enabling more rapid translation of research into practice, and to population health analysts by facilitating patient cohort views. In order to reduce project-specific IT costs, using health IT to assist research necessitates greater flexibility, increasing use of standards, and reusable ways for getting, preparing, and evaluating data. [52]

Any use of operational data in research necessitates the establishment of a privacy and security framework, as well as data governance monitoring. Two initiatives—Informatics for Integrating Biology and the Bedside (i2b2) and Observational Health Data Sciences and Informatics (OHDSI)—have developed informatics tools and approaches that allow researchers to query organizational participants and support transformation or analytics of relevant data to facilitate research. [53, 54]

Specifically, i2b2 has standardized data models and distributed computational tools that enable for the anonymous identification of potential genomic study participants at the institution level. OHDSI also employs a single data model, which incorporates information such as health economics and health systems. The methodologies utilized in these programs demonstrate the kind of functionality that may be required in health IT systems to better support research, as well as the types of concerns with the quality of electronic health data that regularly arise.

Status QUO in health informatics: An Indian perspective

Health care delivery systems

Despite having a solid telecommunication infrastructure, many existing healthcare systems are based on manual record keeping. It would seem the value of health informatics in healthcare delivery has yet to be recognized by many policymakers. [55]

In countries such as India, health informatics is a new and emerging discipline. Its future prospects are very bright, thanks to the development of excellent infrastructure here. However, in order to implement a robust framework, this necessitates a multidisciplinary interaction with various stakeholders.

Information systems development could be another area where research is being conducted to improve the way data flows from various sources such as devices and medical equipment, allowing doctors and decision makers to make rational decisions on critical cases or equipment purchases in the future.

Digitizing all medical data also aids in the creation of a structure for patient-related data in a hospital that can be easily retrieved and searched. Finally, the development of some kind of EHR can be accomplished through the development of information systems. [56]

Applications of health informatics

Health informatics in India can become cost-effective and ensure proper service delivery, which aids in beneficial behavioral change through the use of ICTs. [57] We can create novel applications that can be used effectively by utilizing local talent and effective use of ICT in remote parts of India. [58] Various governance issues can be addressed in the future with certain checks and balances in the data collection and analysis process. [59] The following are some of the areas where health informatics can be used:

• Epidemiological disease prediction
• Disaster management
• Awareness in healthcare processes
• Healthcare in remote areas
• EHRs and their linkages with health systems
• Health statistics
• Education and training
• Development of clinical decision support systems (CDSS)
• Public health research
• Visualization tools for doctors
• Recommendation systems for health informatics
• Precision drug prediction

The potential of this emerging area has far-reaching benefits over a long period of time, and new and novel solutions can be built using various machine/deep learning models. There is a lot of work to be done in this area where we can use cutting-edge technology to aid/assist in the development of robust products and frameworks for public health policy. [60–63]

Future trends and directions in health informatics

There are many current trends in the field of health informatics that can be used to develop sustainable products, services, and health-related policies for effective implementation across the country and internationally. A few of them are listed below, and many more trends may emerge in the near future as a result of discussions with multiple stakeholders.

• Data standards and Interoperability
• Processes to transform medical/clinical data
• Toolkits and pipelines for data management
• Standardized reporting methodologies
• Appropriate use of informatics expertise

More trends may emerge in the future, taking into account the most recent technological advancements. Because health informatics is a new and emerging field, more research challenges may emerge, bringing forth newer perspectives. To solve more difficult problems in this area, future researchers will likely prefer machine/deep learning methods/models. [26] There are numerous other research directions being pursued in relation to various aspects of healthcare data such as quality, veracity, privacy, and timeliness. The following are some of the most notable data characteristics of healthcare data [9, 64]:

1. Complexity and noise: Because healthcare data is multisource and multimodal, it has a high level of complexity and noise. Furthermore, there are issues with impurity and missing values in high-volume data. It is difficult to deal with all of these issues, both in terms of scale and accuracy, despite the fact that a number of methods have been developed to improve data accuracy and usability. [65] Because the quality of data dictates the quality of information, which in turn affects decision-making, it is vital to develop efficient big data cleansing ways to improve data quality in order to make effective and correct decisions. [66]

2. Heterogeneity: Traditional healthcare data is frequently fragmented into multiple forms due to a lack of standardization. As a result, it is both reasonable and important to investigate and adopt universal data standards. However, due to the complexity of developing universal data standards, it is a difficult undertaking. Not only is healthcare data diverse, but there are also numerous technical challenges to integrating that data for specific purposes. [67] Even with standardized data formats, the multi-modal character of data makes efficient fusion difficult [68], necessitating the development of advanced analytics that cope with vast amounts of multi-modal data. The integration and synthesis of multi-source and multi-modal healthcare data on a larger scale would be a significant issue.

3. Longitudinal analysis:

Notes

This presentation is faithful to the original, with only a few minor changes to presentation. In some cases important information was missing from the references, and that information was added. Some grammar and sentence placement was cleaned up for better readability.