1. Introduction
Conceptually, the term “Artificial Intelligence” typically refers to a computerized system consisting of hardware, software, and IT infrastructure that aims to perform real-time commercial and non-commercial applications and cognitive functions with structured human inputs [
1]. A typical AI-based machine or process runs through mathematical logic and computing programs. A wider variety of methods and claims come under the broader scope of AI, including advanced algorithms, machine learning, deep learning, and pattern recognition [
2]. With high productivity and performance, AI has the potential to replace human-oriented work in a wider variety of industrial and social applications. With intensified non-human computational intelligence activities, AI overcomes human limitations [
3]. Thus, AI becomes a solution to real-time industrial and managerial problems, ranging from procurement to after-sales services, through various personalized recommendations to customers through advanced data-driven technologies [
3]. Thus, AI tools present a trade-off between potential benefits and risks, as higher risk and greater value through the usage of perceived technologies are preferred over human-centric solutions [
3,
4]. AI uses supervised and unsupervised machine learning techniques for autonomous decision making for multiple industrial solutions, ranging from BFSI, manufacturing, and retail management to supply chain and logistics management [
4,
5].
Data-driven services are becoming more AI-driven, and thus, AI has become an important element of business strategies for sustainable competitive advantage [
6].
Continuous innovation is bringing new opportunities to various industries, ranging from manufacturing, retail, supply chains, and logistics to transportation and healthcare. A variety of areas in public healthcare leverage the use of artificial intelligence-based technology. Implementing AI in the healthcare system primarily deals with the assessment of the challenges of AI implementation, which aims to manage or alleviate complications and provide ideal treatment for a disease. In public healthcare in particular, AI supports clinical decision support systems for patient-specific diagnosis, treatment decisions, and health analytics [
7]. Thus, for the healthcare industry, these AI-driven services are becoming key elements for creating competitive advantages in the ecosystem. On the risk side, public health has several concerns, including potential bias in the data usage for artificial intelligence algorithms, the prevention and protection of patients’ privacy, and the healthcare practitioners’ distrust of digital tools [
7,
8].
AI has transformed the delivery of public healthcare in emerging countries for specialty treatments, including radiology and pathology. The digital reform in public healthcare is largely supported by the availability of datasets and the novel methods for assessing these datasets. Despite this, it faces a number of challenges in achieving the SDG public health goals, including the lack of a trained workforce and inadequate public health surveillance systems [
9].
Recent advancements in AI have encouraged public enterprises to identify and analyze the risks caused by uncertainty while supporting planning and policy formulation. However, AI interventions requires support from regulators, and practitioners to provide public benefits. AI has been making a lot of progress recently, which is helping public health organizations figure out how bad future outbreaks will be. This is a big improvement over traditional methods, and it will help policymakers and practitioners make better plans and save many lives. However, AI intervention requires that the implementation barriers related to ethics, legality, behavior, and operation are addressed before deployment to developing countries [
10].
In this direction, efforts are being made by countries to achieve health-related SDGs among emerging economies. Past research gives evidence to support a variety of health issues addressed by AI, although it also shows the immense requirement to formulate country-specific guidelines and policies for developing an AI implementation roadmap for low-income and emerging economies [
11].
The recent developments in the area of AI encourage researchers to investigate and evaluate its adoption and implementation. This study discusses the need for and the significance of AI in public services, especially in technological education, vaccine trials, and data informatics. In particular, in the domain of public health systems, the key usage of AI is in the gathering, diagnosis, and interpretation of medical data. However, in addition to its global adoption and implementation in a variety of industries, there are a few important concerns related to ethical and social issues, including trust and reliability. These issues become prominent as AI-driven healthcare systems carry highly sensitive health information and high-end customer vulnerabilities [
11].
In past studies, academic researchers raised key concerns about the implementation of AI-based tools. In developing countries, with the rapid advancements in the area of AI technologies, public organizations are in the process of deploying AI applications to build their productivity and to generate sustainable competitive advantage and value for the beneficiaries.
Leading countries such as India and China are showing a sharp upward trend towards readiness with regard to the adoption of AI applications [
12]. The rapid growth of highlighted technologies, such as 3D printing, big data analytics, and ML/DL, brings a collective emergence of industry game changers [
13]. AI is envisioned as a critical enabler technology in areas such as public finance, labor markets, marketing, public service advertising, public distribution management, road and transportation, and public information systems [
14]. Therefore, AI has become a tool for enhancing the digital capabilities of government in public services and provides a comprehensive socio-economic measure to meet the challenges related to public firms [
15]. AI leverages data integration across both inter-organizational and intra-organizational sources and firms to build more customer-oriented, low-cost business solutions [
16].
In developing countries, the usage of AI is empowering public services, including enhancing public delivery, precision planning and production, and direct benefit transfer. Technology in its new avatar has the potential to boost economic growth and reduce poverty. The cost of managing public health is very high in developing countries, and it rises during emergency times [
17]. In developing economies such as China and India, public health spending has increased by up to 5% of GDP [
18]. At the same time, there are opportunities to create value and incentives in health care systems, with the majority of spending on digitalization and automation. Considering the non-availability of a responsive physical health eco-system, in the recent health strategy draft by the WHO a digital health initiative with the effective usage of digital technologies and health informatics was proposed, with the aim of achieving an economic, equitable, and sustainable health system [
19]. Thus, the adoption of digital technologies, including smart automation and artificial intelligence, can support the WHO’s efforts to make public health more affordable and improve public health. In addition, the constant usage of digital technologies brings a significant impetus to the economic development of the nation [
13]. The United Nations (UN) has shown its commitment to aligning multiple stakeholders to evaluate the role and benefits of digital technologies, including AI, to achieve the Sustainable Development Goals (SDGs) [
20].
In high-income countries, AI is gradually improving public health services. In the USA, AI applications are saving up to USD 150 billion in healthcare costs [
21]. In the context of resource-poor developing countries, the potential of AI in public health needs to be assessed and unleashed [
22]. However, developing countries are struggling with two fundamental issues related to the adoption of digital technologies. Firstly, there is the issue of the lack of public health infrastructure and the dearth of trained human resources. It is believed that the new digital wave is creating a psychological fear of unemployment due to the system-level automation of higher cognitive tasks. Such fears provoke mistrust in institutions and rising populist sentiments among the masses [
23]. Conversely, due to recent government initiatives and successful public-private partnerships, the Indian healthcare industry is projected to reach to USD 372 billion [
24]. On the other hand, the dearth of trained human resources in the healthcare system is facing a global challenge in the imparting of quality health services [
25]. The use of digital technologies, including blockchain, BDA, IoT, AR/VR, and artificial intelligence, is helping clinical practitioners to make precision decisions related to the health industry [
26]. AI’s potential in the healthcare industry is broadly applied in biomedical research, translational research, and medical practice. Thus, AI in health systems amplifies its capabilities [
27]. The usage of AI in public healthcare increases operational efficiency and accuracy during diagnosis, in monitoring health conditions, and in reducing surgical complications [
28] (
Table 1).
The multi-industrial applications and the expanded growth of BDA and AI has prompted the healthcare industry to preview their potential and their risks in public healthcare research [
36]. The success rate from other industries has brought growth potential in the healthcare industry as well [
33]. Globally, the COVID-19 pandemic has transformed healthcare delivery platforms from conventional face-to-face set-ups to online care using digital tools [
37]. While mitigating the risk of exposure to the COVID-19 infection, the healthcare industry has quickly adopted digital collaboration tools for remote clinical aids to patients [
38]. According to Accenture (2017), the AI-enabled global healthcare operations have the potential to bring cost optimization to the level of USD 150 billion by 2030 [
33]. Various precision health apps are creating clinical–community linkage and establishing dialogue between healthcare providers and patients and catering to the health needs of the masses [
39]. In addition, AI also demonstrates the possibility of reducing healthcare costs, providing preventive healthcare to the masses, and increasing the accuracy of diagnoses [
40]. Considering this trend, the majority of healthcare solutions companies are adopting scientifically validated AI methods in their R&D projects [
41].
Motivation of the Study
AI adoption in public sectors and its implementation risks are quite low in some developed countries [
42]. In the majority of South Asian emerging economies, the implementation of AI is at an embryonic phase due to a low clarity regarding digital technologies, a lack of AI regulation and laws, and relevant issues pertaining to the data privacy and trustworthiness among stakeholders [
43]. Thus, digitization in public services can be driven by good governance and robust legislation [
44,
45]. Currently, developing nations are transforming their public systems towards digitization to address the upcoming challenges arising due to the pandemic situation in recent times [
46]. In particular, public health systems impose various challenges for emerging economies that are different to those of the developed countries [
47]. AI implementation can bring together healthcare solutions for various sections of society, irrespective of the socio-economic status [
47]. Thus, there is role for AI in developing countries, where digitally equipped resources and human expertise are very limited and untested. Previous studies have assessed various aspects, including clinical and consumer need identification [
48,
49] and innovation [
50]. AI readiness has challenges [
51,
52]. A few studies also discussed a framework for the real-time health systems [
53], resource optimization [
54], and mass usage [
55]. There is a need for strong and affective governance and a strategic plan for implementing AI applications in public healthcare, education, and other public sectors [
56]. Similarly, recent research has discussed the significance of ethics and policy challenges in the effective governance of AI [
57,
58]. In the developed countries, AI adoption has witnessed research initiatives, and the efforts increase every day. However, developing countries such as India lack research efforts in the area of AI adoption and its practices. The previous literature suggests that the quality of public health using digital technologies in developing countries has been investigated [
55,
58]. However, the conceptual frameworks that divulge the inter-relationships among the barriers and enhance AI adoption by effective and appropriate strategies to reduce the implementation barriers are inchoate. In the past few years, digital developments in public domains are highly acknowledge, AI has emerged as the strategic domain for public services [
59]. Therefore, to address the societal problems related to public health in emerging economies where the public health systems face a lot of constraints related to capacity planning and operational effectiveness, AI may help local governments to develop ample opportunities [
60]. There is an immense need to explore, implement, and expand the usage of AI for public healthcare for the present and future needs of the masses. In a highly populated country such as India, the public healthcare organizations are facing multiple barriers while implementing AI. The past literature shows that the public healthcare industry still expresses few practical concerns about the implementation of artificial intelligence [
61]. There is a need for constant research to be carried out with the aim of understanding and evaluating the various implementation challenges of the AI technologies [
62,
63]. Furthermore, there is a dearth of quality research on the design, development, and implementation of AI-enabled tools to address public health issues [
64]. From the AI application perspective, more practical linkages are required to demonstrate the relationship between AI in medicine and consumer care. There is the need to create a strong theoretical foundation for future research on AI implementation for public healthcare. The present study attempts to bridge that gap and aims to address the research inquiries, including the following:
RQ1: What are the key implementation barriers to artificial intelligence in public healthcare in the developing countries?
RQ2: What is the inter-relationship among the artificial intelligence (AI) implementation barriers in the healthcare industry in the context of developing countries?
RQ3: What is the roadmap to reduce the AI implementation barriers in the healthcare industry?
Thus, this study aims to explore the possible solutions to the research questions in the context of AI implementation in developing countries such as India. To enable this purpose, the paper sets the following research objectives:
To investigate the implementation barriers of AI in public healthcare in developing countries, viz., the Indian context;
To understand the linkage barriers and the dependent, driving, and autonomous barriers among the selected barriers derived from the systematic literature review (SLR);
To provide strategic commendations to smoothen the AI implementation in the public health systems.
The paper has the following organization:
Section 2 elaborates on the literature on AI technology and public health systems and the theoretical groundwork of the study. The next section explains the methodology and methods used for the present study.
Section 4 discusses the detailed research framework on AI implementation in public health.
Section 5 explains the key results of the research study. In
Section 6, a strategic blueprint is developed to reduce the AI implementation bottleneck in the public healthcare domain. The last section covers the conclusion, limitations, and open research challenges for the future.
5. Findings and Discussion
Based on the SLR, the study initially attempted to address RQ1.
Table 4 depicts the key implementation barriers of AI in public healthcare systems, in the context of the developing countries, in order to answer RQ2 and to measure the inter-relationships among the AI implementation barriers in the healthcare industry in developing countries. Based on the expert survey of the subject experts and professionals to assess the hierarchical levels of AI implementation in public healthcare, the integrated ISM Fuzzy MICMAC approach was conducted to formulate the hierarchical level of AI implementation in public healthcare. The ISM results are shown in
Figure 2. Low awareness about AI (
AI-
9), lack of awareness of the legal aspects of AI (
AI-
8), low envisioned future planning towards technological projects (
AI-
11), lack of awareness of the legal aspects of AI (
AI-
10), and low commitment from top management (
AI-
12) are the key drivers for all the other AI implementation barriers. Lack of awareness of legal aspects of AI (
AI-
10) and low commitment from top management (
AI-
12) are declared as the implementation barriers that have the maximum driving power and collectively answer RQ1. The results are validated by the fuzzy MICMAC method, which shows the classification of the AI implementation barriers into three main clusters.
Cluster I does not have any implementation barrier. This implies that there are no weak implementation barriers in the study.
Cluster II represents the dependent barriers; Cluster III demonstrates the linkage barriers; Cluster IV demonstrates the driving barriers. On the basis of dependence and driving power, the implementation barriers with high dependence and weak driving power are included in Cluster II. This cluster includes implementation barriers, low level of coordination among parties, lack of trust (AI-1), limited data repository facilities (AI-2), upscaling of data (AI-3), data ownership (AI-4), high cost of maintenance (AI-5), data risk (AI-7), absence of health informatics standards (AI-6), lack of proper infrastructure to support AI implementation (AI-14), and data security and privacy (AI-15). These implementation barriers require all the other barriers to minimize the impact of the dependent values on the overall performance. There is a lack of know-how and technical expertise among executives (AI-3). The linkage implementation barrier is included in Cluster III with high driving and dependence power. The driving implementation barriers are included in Cluster IV. These barriers have the highest driving and the weakest dependent power. AI-8, low investments in R&D, low awareness about AI (AI-9), low envisioned future planning towards technological projects (AI-11), lack of awareness of the legal aspects of AI (AI-10), and lack of commitment from top-level management (AI-12) are the driving barriers. These barriers are obtained on the lowest of the ISM levels.
Practical Implications
The study outcomes not only contribute to the existing research literature but also give an in-depth understanding of the AI implementation barriers in public healthcare settings in the context of a developing country. The insights from the study can help future researchers and decision makers to understand the significance of the AI implementation barriers and focus on the most critical driving and dependent implementation barriers. From the study, it is clear that the policymakers need to understand the key benefits of AI implementation in the public healthcare. The summarized implications are:
- i.
Securing medical and clinical data
AI can provide the integration of a variety of partners, suppliers in the healthcare supply chain with shared information from existing information databases, and infrastructure and relevant digital records related to patients, their medical history, and their feedback. The removal of implementation barriers related to data security will be required for the smooth flow of information.
- ii.
Trusted collaboration
AI can be helpful in reducing data counterfeit and other threats related to healthcare operations. The tracking of a vaccine supply to healthcare systems using AI can be performed for ensuring quality and timely delivery.
- iii.
Holistic quality management
AI can ensure the holistic quality in the healthcare system. There are various applications of AI, including medical imaging to ensure quality and timely delivery.
As discussed in
Section 6, RQ 3 is responded to by the development of a roadmap to reduce the AI implementation barriers to the healthcare industry.
6. Strategic Roadmap
The main research aim of the study is to determine the potential of AI in public healthcare and also to evaluate the barriers towards AI implementation in the industry. The research findings can benefit policymakers to develop a strategic roadmap for the implementation of digital technologies (such as BDA, AR/VR, and blockchain technology) in public healthcare. The key outcome of the research is the knowledge about inter-relationship among the barriers related to AI implementation and also the basis of the causality and prominence. Due to various implementation barriers that have an impact on AI implementation, it remains low in emerging economies. In this context, the implementation of AI, particularly for public health, can be enhanced if health systems and policy makers are aware of the barriers that contribute to its successful deployment and have an understanding of the relationships among the implementation barriers. Thus, the research is significant in evaluating the importance of the variety of AI implementation barriers in public healthcare systems. Furthermore, the strategic roadmap consists of the following steps:
The development of industrial symbiosis leads to a digital ecosystem for resource sharing among parties;
The development of a centralized AI-enabled system is for the co-creation of new and open healthcare systems;
The support from the top-level management of key sustainable practices will enhance the focus of the health organizations to collaborate in AI implementation.
6.1. Conclusions, Limitations and Open Research Challenges
AI exhibits a great potential to transform the public healthcare sector. If adopted effectively, various operations issues such as public hospital record maintenance costs, inefficient healthcare practices, and data breaching can be easily handled. The overall ability acquired by AI in public healthcare can help hospital and public healthcare centers to fully secure patient data and trails and manage the outbreak of a harmful situation such as that generated by the COVID-19 pandemic. AI in public health is an important area of research to explore collaboration and inter-dependencies. Electronic resource sharing, and public services with support from the government, private organizations, and NGOs can take more initiatives to resolve the ongoing societal issues and provide electronic health services to the public. Very rarely, studies have confirmed the ISM outcomes through fuzzy MICMAC analysis to determine the inter-relationships among the implementation barriers. Our modeling results show the different implementation barriers, according to their significance on scale of dependence, autonomy, linkage, and independence. It has been observed in the interaction with experts that they are still watching the top players utilize the technology to see its impact on their performance and the implications of their day-to-day operations. As the implementation of AI is a costly affair, therefore they do not want to risk it. There is a dearth of quantitative studies about the potential benefits of AI in public health in developing countries. In addition, there is a larger concern related to infrastructural support, functional skills requirement, and implementation readiness among the top-level management. The scarcity of professionals specifically in the combination of healthcare systems and AI is another concern in developing countries such as India. Top-level management includes either doctors or those with other qualifications.
This technology experiences difficulty in the Indian healthcare sector due to the existing IT act (IT Act, 2000). According to the IT Act, any kind of breach of personal data should compensate the victim. If the healthcare ecosystem can overcome the driving variables, those that are recognized as the barriers to implementing AI in healthcare, then the formulation of a real-time public healthcare system can be realized. Currently, the public healthcare system is having many implementation challenges from data generation to secure and reduce the cost of operations for the two main stakeholders, namely the patients and the healthcare providers. Moreover, AI implementation can be helpful in the identification of low quality and counterfeit drugs and other medical commodities.
6.2. Limitations and Future Research Directions
The authors have deployed a multi-criteria decision approach to determine the implementation barriers towards AI adoption in public healthcare. The research carried out in the near future may record the effect of the barriers on AI implementation in the public healthcare sector in the context of developing countries such as India. Furthermore, empirical shreds of evidence can be collected to validate the results of the study, and selected case studies can be prepared. Based on the study and from the public healthcare system perspective, interested parties can develop a collaborative implementation roadmap while designing, managing, and implementing AI across public hospitals and healthcare systems. In addition to the above, various allied technologies, including public clouds, IIoT, and blockchain, are used for data collection, gathering, storage, and access to ensure a secured and decentralized access for effective public healthcare. Future studies can undergo empirical studies for evaluating the AI implementation barriers and their impact on the healthcare industry. In addition to the above, allied technologies, including public clouds, IIoT, and blockchain technology can be useful in data generation and their gathering and data access to ensure operational excellence in public healthcare systems. Further studies can undergo empirical studies for evaluating the AI implementation barriers and their overall impact on the public healthcare industry.
The research limitations can be further improvised in future research works. The identification of implementation challenges and their identification as common entities for both developing and developed countries is difficult. This study evaluated 15 implementation barriers from a single country; thus, for the purpose of generalization, more cross-country data are required. Moreover, more empirical evaluation of the research problem shall be required. Furthermore, various perspectives on the design and development of the conceptual framework can be further expanded and empirically developed from the viewpoint of sustainable public healthcare systems.