Next Article in Journal
A Data-Driven Intelligent Methodology for Developing Explainable Diagnostic Model for Febrile Diseases
Previous Article in Journal
A Hybrid Dimensionality Reduction Procedure Integrating Clustering with KNN-Based Feature Selection for Unsupervised Data
Previous Article in Special Issue
Pediatric Pneumonia Recognition Using an Improved DenseNet201 Model with Multi-Scale Convolutions and Mish Activation Function
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Structural Analysis of AI Implementation Challenges in Healthcare

1
School of Mechanical Engineering, KIIT University, Bhubaneswar 751024, India
2
Institute of Digital Anti-Aging Healthcare, Inje University, Gimhae 50834, Republic of Korea
*
Authors to whom correspondence should be addressed.
Algorithms 2025, 18(4), 189; https://doi.org/10.3390/a18040189
Submission received: 18 February 2025 / Revised: 14 March 2025 / Accepted: 24 March 2025 / Published: 26 March 2025
(This article belongs to the Special Issue Machine Learning in Medical Signal and Image Processing (3rd Edition))

Abstract

:
The incorporation of artificial intelligence (AI) into the healthcare system has been revolutionized, promising key advancements in diagnosis, treatment, patient care, administrative tasks, and operational efficiency. Using an in-depth analysis of the extensive amount of research on artificial intelligence and how it could help the medical industry, this study identified eleven barriers and challenges. Interpretive structural modeling (ISM) was used as a methodological approach to determine the relationship between the difficulties extracted and their dependency and driving powers. It resulted in a five-tiered model, with the introduction of innovative and new-generation tools topping the chart as the most dependent challenge. Similarly, Insufficient Data, Data Acquisition, Data Misuse, and Missing Compassion were the key drivers. Therefore, during the implementation of artificial intelligence in medicine, these challenges should be considered. Although artificial intelligence (AI) possesses the groundbreaking power to enhance patient care and operational efficiency in the healthcare sector, there are several key problems that must be addressed for implementation to be fruitful. The order of these challenges was ascertained through interpretive structural modeling, underlining the significance of innovation and data-related issues. Health systems can optimize AI’s benefits and enhance diagnosis, patient care, and overall hospital management by aggressively eliminating its deterrents.

1. Introduction

AI, or artificial intelligence, can be explained as the creation of systems of computers that can perform activities that usually require intelligence matching human levels. This requires a broad range of skills, including understanding natural language, identifying patterns, solving challenging issues, and learning from experience (machine learning) [1]. The applications of AI are found in multiple areas such as healthcare, finance, manufacturing, and services. The objective is to develop systems capable of conducting tasks independently, adapting to new, different environments, and continually improving their performance. AI is a crucial innovation in the twenty-first century, affecting various aspects of our everyday lives and revolutionizing businesses via automation, data analysis, and advanced decision-making capabilities.
Artificial intelligence (AI) is a ubiquitous and transformational force in many sectors in the twenty-first century. AI spurs innovation, changes industries, and impacts the development of society. Particularly evident are its effects on automation, data processing, and decision-making. Data analysis is one of AI’s key functions. AI algorithms can sift through enormous databases, extracting crucial insights and supporting well-informed decision-making in the face of the exponential growth of data. In sectors like finance, marketing, and research, this is crucial. Another significant part of AI’s participation is automation. Robotics driven by AI enhances production procedures, boosting accuracy and efficiency. Chatbots for customer service are an example of automation; they expedite communication and improve user experiences.
AI is radically changing patient care, diagnosis, and research, among other sectors of the healthcare industry. AI-powered image analysis has had a notable influence on diagnostic efficiency and accuracy. Clinical professionals may now detect minute abnormalities with remarkable precision because of machine learning algorithms’ quick interpretation of complicated medical images, including MRIs and X-rays [2]. Moreover, AI makes customized medicine possible by utilizing extensive patient data, such as genetic and medical history, to customize treatment regimens [3]. Healthcare can take proactive measures to interrupt and perhaps halt the spread of illnesses when diseases are detected early on thanks to predictive analytics and machine learning.
Additionally, it lessens medical mistakes and revises existing research [4]. Chatbots and virtual health aides [5], supported by artificial intelligence, improve patient involvement and assistance by offering data, responding to inquiries, and encouraging medication adherence and lifestyle management [3]. AI is also transforming the findings and development of new drugs by identifying possible medication candidates much more quickly and estimating their effectiveness [6].
With its continued development, AI has the potential to improve patient outcomes, streamline healthcare delivery, and spur new ideas to solve persistent problems in the medical industry. There are, however, several obstacles to the integration of AI into healthcare systems, including the requirement for legal frameworks, significant implementation costs, and data privacy issues [7]. Notable challenges include ensuring the ethical application of AI [5], eliminating algorithmic biases, and earning the confidence of medical practitioners. Further obstacles include the interpretability of AI judgments and the difficulty in smoothly incorporating AI into current healthcare operations. Succeeding in integrating AI into healthcare requires striking a balance between ethical issues and technological progress.
Although research has been conducted on the promise and difficulties of artificial intelligence in healthcare globally, little is known about the particular challenges faced by the Indian healthcare industry. By concentrating on specific complexities impeding AI integration in the Indian context, this study seeks to close this gap and offer a significant, regional viewpoint on this quickly developing topic.

2. Literature Review

An overview of the literature on the challenges of implementing AI in healthcare systems revealed pertinent barriers. Table 1 highlights the key challenges of AI in healthcare.

2.1. Insufficient Data

Insufficient data is a significant problem in the incorporation of AI into healthcare since it compromises the quality and generalizability of AI models [8]. Inaccuracies may occur owing to a lack of varied patient groups, leading to biased predictions and reduced diagnostic accuracy. The lack of extensive data sets [6] covering unusual medical disorders or recent advances in therapy limits AI systems’ capacity to adapt to changing healthcare environments. Data aggregation is hindered by privacy concerns and interoperability challenges, limiting AI’s capacity to provide pertinent solutions [9]. To overcome the challenge of limited data, it is crucial to prioritize data sharing, standardization, and ethical concerns while developing AI [10].

2.2. Social Issues

Social considerations provide a substantial hurdle to the integration of AI into healthcare. Concerns about equity, transparency [5], and bias in AI algorithms may exacerbate current healthcare disparities, as these systems may inadvertently favor certain demographic groups over others. Patients and healthcare practitioners disagree on the reliability and ethical implications of AI-generated advice, raising trust and accountability problems. Concerns concerning a lack of emotional availability that would aid in patient therapy have also been raised [11]. Furthermore, the potential employment displacement induced by automation in healthcare contexts poses ethical and financial concerns. To ensure that AI applications in healthcare address social challenges, it is necessary to carefully examine algorithmic biases, communicate openly about the technology’s limits, and implement regulations that prioritize inclusion.

2.3. Clinical Implementations

Clinical application is a significant hurdle in the introduction of AI into healthcare. Despite promising advances in AI technology, successfully integrating these technologies into established healthcare procedures involves overcoming several challenges [4]. Clinicians may struggle to comprehend and trust AI-generated outputs, limiting their use in decision-making processes. Furthermore, integrating AI systems with electronic health records [12] and other aspects of healthcare infrastructure may be difficult, frequently necessitating extensive customization and interoperability concerns. Keeping AI systems compliant with regulatory norms, protecting patient privacy [4], and addressing ethical concerns complicate the deployment process.

2.4. High Costs

Its high cost provides a significant barrier to the adoption of AI into healthcare. Creating, deploying, and sustaining AI systems in healthcare contexts sometimes necessitates significant financial expenditures [4,17]. The costs include purchasing modern gear, hiring specialized workers for development and maintenance, and assuring compliance with strict data security and privacy requirements. Training healthcare workers in AI technologies and integrating these systems into existing infrastructures might add to the economic burden. This cost barrier [3] may limit the availability of AI-powered healthcare solutions, aggravating existing discrepancies and preventing universal adoption. Addressing the issue of high prices necessitates intelligent investment planning, collaboration between the public and commercial sectors, and the creation of cost-effective models that assure the long-term integration of AI into healthcare procedures.

2.5. Black-Box Scenario

A black-box schema occurs when the operations and actions undertaken in a system are not visible or accessible to the user or other interested parties. Because machine learning algorithms are opaque, the black-box situation makes applying AI in healthcare systems tough [13]. While these algorithms can successfully process massive quantities of data to generate predictions or aid in medical diagnosis and planning, the lack of openness [14] about how these judgments are made raises questions about responsibility, trust, and possible biases. In medical care, where decisions may have life-or-death effects, doctors must comprehend the logic behind AI suggestions before securely incorporating AI technologies into their practice. Furthermore, legal and ethical norms need openness [12] and explainability in the decision-making process, hindering the use of AI in healthcare. To address the black-box dilemma, interpretable AI models and rigorous validation procedures must be constructed and explicit rules must be set for transparency and accountability in artificial intelligence-driven healthcare systems.

2.6. Data Acquisition

Data collection is a substantial difficulty in the incorporation of AI into the healthcare sector for several reasons. First, healthcare data are frequently dispersed across several systems, in different forms, and with varied degrees of standardization, making them difficult to combine and harmonize for AI applications [4]. Second, concerns with interoperability across various healthcare platforms impede the seamless sharing of information. Third, concerns about patient privacy and data security result in severe rules that limit the sharing and use of healthcare data for AI research. Furthermore, data quality, accuracy [15], and completeness may vary, affecting the performance and generalizability of artificial intelligence models. To address these difficulties, efforts must be made to improve data interoperability, implement strong privacy rules, and encourage collaboration across healthcare organizations to construct complete and representative data sets.

2.7. Introduction of Innovative and New-Generation Tools

There are various obstacles to overcome for innovative and next-generation technology implementation [13] for the incorporation of AI into the healthcare sector.
First, there is sometimes opposition to change inside existing healthcare systems, including personnel used to old techniques. The deployment of new instruments and technologies necessitates extensive training and a cultural shift in the way healthcare personnel approach their jobs and also in how well they are incorporated into the system [16]. Second, implementing and integrating cutting-edge technology might be prohibitively expensive, creating financial issues [18] for healthcare organizations. Third, guaranteeing these technologies’ compatibility with current systems and infrastructure is a challenging undertaking. Furthermore, regulatory frameworks may not always keep up with quickly emerging technologies, resulting in questions about compliance and standards.

2.8. Missing Compassion

One of the obstacles to the incorporation of AI into the healthcare sector is that computer interactions may lack compassion compared to human interactions. While AI systems excel at analyzing vast amounts of data and drawing unbiased findings, they might not be able to convey empathy or emotional understanding, which is crucial in the healthcare industry [11]. Patients frequently expect healthcare practitioners to communicate compassionately and empathetically [4] alongside accurate diagnosis and treatment strategies. In healthcare, the human touch entails interpreting emotional cues, offering comfort, and tailoring communication to individual requirements. The difficulty is in striking a balance between the efficiency and objectivity of AI technologies and the compassionate, human-centered treatment that patients demand. Healthcare practitioners must develop methods to smoothly integrate AI while preserving the critical human factors that lead to a pleasant patient experience and therapeutic connections. This entails carefully planning the design and implementation of AI systems to supplement, rather than replace, the sympathetic components of healthcare delivery.

2.9. Data Misuse

Data misuse is a big concern when incorporating AI into the healthcare industry. The large amount of sensitive patient information in healthcare makes it a tempting target for hostile activity if not handled with extreme caution. Unauthorized access, data breaches [9], and the incorrect sharing of health data are all examples of serious privacy infractions that can hurt individuals. AI integration necessitates the collection, storage, and processing of large data sets; hence, strong security measures must be implemented. Furthermore, the possibility of biases and biased outcomes in AI systems might worsen current healthcare inequities [4]. Ensuring ethical procedures, robust security standards, and respect for privacy legislation is vital in limiting the danger of data exploitation and creating confidence between patients and healthcare.

2.10. Data Privacy and Security

Data privacy and security are crucial considerations when incorporating AI into the medical sector [9]. Healthcare systems handle extremely sensitive and private patient data, making them prime candidates for cyberattacks and unauthorized access. The sheer extent and complexity of healthcare data, which frequently include electronic health records, diagnostic pictures, and genetic information, heighten the danger of data breaches. Inadequate security measures can have serious repercussions, such as identity theft, financial fraud, and breached patient confidentiality [10]. Furthermore, AI integration may entail sharing and analyzing data across several platforms, which increase the potential attack surface [7]. It is critical to maintain uniformity while making data more accessible for AI applications and providing strong privacy measures. Stringent adherence to data protection legislation, the use of encryption technology, and the establishment of standardized security standards are critical components in resolving these issues and establishing confidence among patients, medical professionals, and regulators.

2.11. Technology Development

For several reasons, technological advancements make incorporating AI into the healthcare industry challenging. First, the rapid rate of technological progress necessitates that healthcare personnel constantly upgrade their abilities and adjust to new equipment and procedures. Training and instruction are required to guarantee that medical professionals can use AI technologies correctly. Second, the creation and implementation of AI systems sometimes require enormous financial commitments, making it difficult for some healthcare organizations, particularly the smaller ones, to purchase and apply cutting-edge technology. In addition, guaranteeing the interoperability of artificial intelligence solutions with the current health IT infrastructure is a daunting undertaking since disparate systems may not connect properly. Finally, the healthcare business is highly regulated, and the development of AI systems must adhere to stringent ethical and legal norms, necessitating the careful navigation of regulatory structures. Addressing these problems will require continuous education, monetary considerations, cooperation among technology developers and medical professionals, and compliance with regulatory rules to ensure the proper integration of AI in healthcare. Table 2 represents the notations of eleven identified factors.

3. Methodology

To achieve the aims outlined in this work, the writers took a series of actions. First, the essential components were determined through thorough literature research and expert comments. After that, a model based on ISM was developed to examine the connections between the various parts. Finally, MICMAC analysis was employed to classify them. This is the study framework shown in Figure 1.

3.1. Data Collection

Initially, the authors conducted extensive literature research to learn about artificial intelligence. This enabled them to determine how and where AI could be introduced in the healthcare sector.
The authors then examined research databases using key terms like artificial intelligence, the healthcare sector, medicine, automation, and ISM. Then, the abstracts from all retrieved papers were evaluated to identify appropriate research for the study at hand. Further, the specialists were tasked with assessing a list of crucial factors that elucidate the advantages and challenges that AI poses in the medical industry. We had the honor of interacting with 12 experts, all from various medical backgrounds, from 4 government and 3 private institutions. These experts provided us with valuable insights that served as the foundation of our study (for the key factors) and of our ISM model.
Our twelve experts included seven doctors cum academicians from different departments like oncology, radiology, pediatrics, neurology, gynecology, ophthalmology, and cardiology, with two having pharmaceutical backgrounds, two medical technicians, and one head nurse. The selected experts had various medical field expertise from day-to-day handling and research knowledge. They helped point out where all automation could be enabled to facilitate daily processes and the common challenges that would come up while establishing automation in healthcare. ISM does not impose any restrictions on the sample size of experts; hence, the number of experts chosen for this investigation was adequate. Furthermore, other studies have considered a comparable sample size. Expert participation in this study was sufficient according to previous research [19,20].

3.2. Interpretive Structural Modeling (ISM)

ISM is a process for determining and summarizing the connections between various elements that comprise a problem or issue. An interactive method that takes expert advice into account is used to build connections between the parts of a complex system [21,22]. In contrast to other methods like Delphi and structural equation modeling, ISM has the benefit of requiring fewer professionals and yielding a structured model from an unstructured and ambiguous raw data collection.
The ISM technique helps establish strong relationships between the many aspects of a problem. ISM transforms an ambiguous and unstructured model into one that is well defined and structured [23]. It offers an open perspective on the problem and concentrates on the factors’ direct and indirect relationships [24]. Additionally, it highlights the aspects of a problem that are both short- and long-term-oriented [22]. To create a managerial vision quickly, ISM efficiently arranges expert perspectives and in-depth knowledge [25,26]. Table 3 represents the significance of ISM in different research domain.
This research establishes the key variables that pose challenges in implementing artificial intelligence in the healthcare sector, making ISM one of the most fitting methods for this study. The framework is an appropriate choice since this methodology gives a group a way to impose an order on its items’ complexity. ISM has many benefits, including merging expert opinions and knowledge bases methodically while allowing for frequent judgment changes. For targets (items) with ten to fifteen numbers, the computational effort required is quite low, and it can be a convenient tool in practical applications. Additionally, the ISM model’s outcomes could be impacted by the experts’ biased assessments, mitigated by the experts’ breadth and depth of expertise in providing objective comments. As a strategy for the implementation of automation in an ever-changing medical archetype, ISM is a dependable technique for our investigation in determining and creating a framework to illustrate the difficulties and solutions faced by industries in generating revenue.
This procedure is followed by the ISM method. Expert judgment from the factors found was first used to create a structural self-interaction matrix (SSIM). After that, a reachability matrix was generated to ascertain the inter-dependencies and driving forces between the chosen factors. Third, the factors were divided into levels using a reachability matrix. Fourth, an interpretive structural model (ISM) was created. Lastly, the factors were classified into four different groups based on their driving forces and dependencies, autonomous, dependent, linkage, and driver, using MICMAC analysis.

3.2.1. Structural Self-Interaction Matrix (SSIM)

The writers constructed the contextual interaction among the components with the valuable input of fifteen professionals in business and academics. The SSIM was developed when the pair-wise comparison data amongst the different factors were received. The relationship nature between two factors (x and y) is represented by the four symbols below. The following scenarios occur: (i) V happens when x affects y; (ii) A happens when y affects x; (iii) X happens when x and y interact; and (iv) O happens when x and y are unrelated. Table 4 presents the SSIM for the primary factors (namely Insufficient Data (ID), Social Issues (SI), Clinical Implementation (CI), High Cost (HC), Black-Box Scenario (BS), Data Acquisition (DA), Introduction of Innovative and New-Generation Tools (IIN), Missing Compassion (MC), Data Misuse (DM), Data Privacy and Security (DPS), and Technology Development (TD)) that hinder the adoption of artificial intelligence in the medical sector based on the opinions of the experts using the four symbols mentioned above. This SSIM is necessary for the reachability matrix to function.

3.2.2. Reachability Matrix

The structural self-interaction matrix, also known as the SSIM, further yields the reachability matrix. This matrix is achieved by substituting 1 and 0 for V, A, X, and O based on the factors in each of the rows and columns. The norms that control the replacement of the 1s and 0s are as follows, letting (x, y) be the respective row and column entries in the reachability matrix:
(i)
if the (x, y) entry in the SSIM is V, the reachability matrix entry becomes 1, and the (y, x) entry is 0;
(ii)
if the (x, y) entry in SSIM is A, the reachability matrix entry becomes 0, and the (y, x) entry becomes 1;
(iii)
if the (x, y) entry is X in the SSIM, the reachability matrix entry becomes 1, and the (y, x) entry is also 1;
(iv)
if the (x, y) entry is O in the SSIM, the reachability matrix entry becomes 0, and the (y, x) entry is 0.
This shows that it respects transitivity as well. The resulting reachability matrix is shown in Table 4. The relationships between and motivations for each component are also depicted in this reachability matrix. For example, in Table 4 (SSIM table), the notation A is used to determine the influence of CI on DA. Then, in the reachability matrix (Table 5) the notation ‘A’ is substituted to ‘0’ from CI (factor-3) to DA (factor-6) {row-wise} and to ‘1’ from CI (factor-3) to DA (factor-6) {column-wise}.
In this analysis, factors 7 and 1, 6, and 9 show the highest levels of dependency and driving power, respectively. The driving force and dependency of each factor (Insufficient Data (ID), Social Issues (SI), Clinical Implementation (CI), High Cost (HC), Black-Box Scenario (BS), Data Acquisition (DA), Introduction of Innovative and New-Generation Tools (IIN), Missing Compassion (MC), Data Misuse (DM), Data Privacy and Security (DPS), and Technology Development (TD)) help with the MICMAC analysis, as represented in Table 5.

3.2.3. Level Partition

Using the reachability matrix, step three entails determining the reachability set and antecedent set for each factor. These are the two sets that define when a selected factor influences the other factors (reachability set) and when the remaining factors affect or influence the selected factor (antecedent set). A third set represents the shared components or factors between the antecedent set and the reachability set and is referred to as the ‘intersection set’. For example, High Cost affects the Introduction of Innovative and New-Generation Tools and Technology Development, and it is affected by all the other nine factors.
Level one is established when the values of the intersection set and reachability set coincide. In this study, the intersection and reachability sets have the same value for factor 5. Factor IIN is therefore viewed as level one and is shown in Figure 2. After obtaining the level one, or top-level, factor, it will be segregated from the other components. Similar outcomes will be reported for the remaining component levels. The final ISM model shall be created with the help of the specified level. Table 6 displays the reachability set, antecedent set, intersection set, and level for each factor. This study identified five levels, with component F7 relating to level 1 and factors F1, F6, F8, and F9 belonging to level 5.

3.2.4. Formation of Interpretive Structural Model

The ISM-based model, created using a level partition table and digraph, is displayed in Figure 3. The interconnectedness of the factors is represented by the digraph’s nodes and edges. The ISM model is generated by the lines of the edges and vertices, or nodes, of the final reachability matrix [33]. The relation between factors x and y is represented as an arrow pointing from x to y. Finally, the digraph is converted to the ISM model by removing the transitivities. The ISM model presents the structural relationship between the key components [34].

3.2.5. MICMAC Analysis

The driving powers and dependence diagram are also known as MICMAC (Cross-Impact Matrix Multiplication Applied to Classification) analysis. The determined driving powers and dependencies serve as the infix for the MICMAC analysis [35]. As illustrated in Figure 3, the identified critical elements that were previously defined are categorized into four quadrants:(i) the factors in the first quadrant are mentioned as autonomous factors because they have low driving powers and low dependencies; (ii) the factors of the second quadrant are cited as dependent factors as they have low driving powers and high dependencies; (iii) third-quadrant factors are named as linkage factors because they have high dependencies and high driving powers; and (iv) fourth-quadrant factors are mentioned as drivers because they have high dependencies and low driving powers [36].

4. Results and Discussion

We found eleven significant obstacles, namely Insufficient Data (ID), Social Issues (SI), Clinical Implementation (CI), High Cost (HC), Black-Box Scenario (BS), Data Acquisition (DA), Introduction of Innovative and New-Generation Tools (IIN), Missing Compassion (MC), Data Misuse (DM), Data Privacy and Security (DPS), and Technology Development (TD), in our study of the successful application of AI in healthcare during our investigation. Our research explores the interconnectedness of these issues, even though previous researchers like Aung et al. [4] and Kasula [37] have recognized some barriers, such as data constraints, privacy concerns, ethical considerations, and development hurdles. The potential advantages of AI are well highlighted in the literature, including the development of individualized health programs, addressed by Jimma [3] and Gille et al. [5], improved decision-making in healthcare management [4], and quicker and more accurate disease detection through comprehensive medical data analysis [5,38]. There is still a lot to learn about the interactions and combined effects of these obstacles on the healthcare system and how the medical community perceives them, as analyzed and researched by Dumitrașcu et al. [39].
Our paper sets itself apart by going beyond a simple list of difficulties. We explored the complex interrelationships among these elements, clarifying how they impact one another and mold the short- and long-term uses of AI in healthcare. This research offers a more thorough and nuanced viewpoint, highlighting several ways these difficulties function in concert and providing a stronger foundation for negotiating the hindrances of implementing AI in this crucial industry.
In our study, ISM was adopted to construct and analyze these limitations. Insufficient Data (IN), Data Acquisition (DA), Data Misuse (DM), and Missing Compassion (MC) were the independent driving factors. The introduction of innovative and new-generation tools (IIN) topped the chart by being the most dependent and driven factor among the ones under study.
This study revealed five levels of hindrances from the pre-identified ones. Level-five obstacles included Insufficient Data (IN), Data Acquisition (DA), and Data Misuse (DM). This implies that the foundational issues in creating an artificial intelligence-integrated healthcare sector system include data availability and collection. It also includes Missing Compassion (MC) as one of the factors that cause the reduction in human emotions with one another when AI is introduced.
Next, in the fourth tier, there are Data Privacy and Security (DPS), Black-Box Scenario (BS), and Social Issues (SI). Data Privacy and Security (DPS) and Black-Box Scenario (BS) pose a huge threat, which Bartoletti [9] and Stewart et al. [2] also addressed in their studies, as they could lead to the disruption of the entire healthcare system by challenging its authenticity. Social Issues (SI), on the other hand, not only create hindrances in terms of emotional understanding but also disable the ability of understanding and comprehending the pains and troubles of an ailing patient.
The Clinical Implementation (CI) challenge follows in tier three, wherein the installation of the complete automation system in practice when using AI in the medical field is very different from the theory. With all the infrastructural changes, software requirements, and data necessities, shifting from our current modes of working and usage to new automations is a daunting task, as analyzed by Aung et al. [4] as well. Moreover, for doctors and other the healthcare sector workers, trusting the diagnosis and results of a system-generated response in matters of life and death can also be very difficult.
In tier two, the challenges recorded are Technological Development (TD) and High Cost (HC). While the development progress of technology is rapidly increasing and easing things, it creates hurdles because of users. In this case, the doctors, nurses, and medical technicians, have to relearn how to use the technology every time when there are advancements. High Cost (HC), on the other hand, determines the feasibility of the switch. While switching the existing system to its more advanced versions is quite necessary, the possibility of turning it into reality depends on the affordability criteria, as agrees Bertsimas et al. [17].
Lastly, the Introduction of Innovative and New-Generation Tools (IIN), identified in level one, causes troubles in changing the existing healthcare system, as addressed by Rebelo et al. [16] in their study. Aside from the requirements to change the infrastructure, the financial requirements, and the technological advancements, one of the most significant factors needed to make the integration of artificial intelligence into the existing healthcare system successful is government policies and frameworks.
Finally, the entire research model can therefore be divided into four different categories, which include autonomous, dependence, driver, and linkage groups. The autonomous factors include Social Issues (SI), Black-Box Scenario (BS), and Data Privacy and Security (DPS). These variables exhibit poor drive and dependence power. They are detached from the system and have few strong connections. The Clinical Implementation (CI), High Cost (HC), Introduction of Innovative and New-Generation Tools (IIN) and Technological Development (TD) factors fall into the dependent category as they are weak drivers but have strong dependence power. The Insufficient Data (ID), Data Acquisition (DA), Missing Compassion (MC), and Data Misuse (DM) factors lie in the driver category. These factors have strong driving power and weak dependence power. The last category of linkage factors includes factors that are both strong drivers and have strong dependence characteristics. However, none of the aforementioned factors belong to this category in our study.

5. Conclusions

Finally, the operation of AI in medical care holds both great potential and severe hurdles. While AI technologies can potentially improve diagnosis, treatment planning, and patient care, data privacy concerns, costs, clinical implementation, and ethical considerations constitute their top priorities. The complexity of healthcare systems, along with their requirement for the rigorous validation and interoperability of AI solutions, emphasizes the need for collaborative attempts between researchers, medical practitioners, policymakers, and technology developers. Addressing these issues necessitates a multifaceted strategy prioritizing openness, accountability, and patient-centeredness. Despite its challenges, the ongoing progress and appropriate use of AI have the potential to transform healthcare delivery and enhance patient outcomes on a worldwide scale.
Expanding on our structural analysis, future studies might examine the difficulties in using AI in particular healthcare applications, consider the effects of new technologies, and carry out cross-cultural investigations to comprehend various adoption trends. Importantly, including human-centered designs to address compassion concerns is crucial, as is creating strong ethical frameworks, data governance plans, and explainable AI methodologies. Our findings will be validated and expanded through workforce and economic effect studies, improved methodological techniques like simulations and real-world case studies, and targeted policy research on legislative frameworks like GDPR and HIPAA. For a thorough understanding, it is essential to address issues with hardware, software, and interoperability through a comprehensive analysis of AI governance and privacy. Our comprehension of AI integration in healthcare will also be further refined by using additional MCDM methodologies, such as AHP, which will offer an alternate prioritization of the aspects that have been identified. Lastly, all future initiatives must be directed by interdisciplinary cooperation, patient-centered strategies, and fair access.

Author Contributions

Conceptualization, Q.A. and S.T.; methodology, Q.A. and D.S.; validation, K.B., H.-C.K. and S.S.; formal analysis, Q.A. and S.S.; investigation, K.B., S.T. and D.S.; resources, K.B., H.-C.K. and S.S.; data curation, Q.A., K.B. and H.-C.K.; writing—original draft preparation, Q.A.; writing—review and editing, S.T., D.S. and S.S.; visualization, Q.A. and D.S.; supervision, S.T., H.-C.K. and S.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

All the participants gave their consent to participate in this study.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Zahlan, A.; Ranjan, R.P.; Hayes, D. Artificial intelligence innovation in healthcare: Literature review, exploratory analysis, and future research. Technol. Soc. 2023, 74, 102321. [Google Scholar] [CrossRef]
  2. Stewart, J.; Sprivulis, P.; Dwivedi, G. Artificial intelligence and machine learning in emergency medicine. Emerg. Med. Australas. 2018, 30, 870–874. [Google Scholar] [CrossRef]
  3. Jimma, B.L. Artificial intelligence in healthcare: A bibliometric analysis. Telemat. Inform. Rep. 2023, 9, 100041. [Google Scholar] [CrossRef]
  4. Aung, Y.Y.M.; Wong, D.C.S.; Ting, D.S.W. The promise of artificial intelligence: A review of the opportunities and challenges of artificial intelligence in healthcare. Br. Med. Bull. 2021, 139, 4–15. [Google Scholar] [CrossRef]
  5. Gille, F.; Jobin, A.; Ienca, M. What we talk about when we talk about trust: Theory of trust for AI in healthcare. Intell.-Based Med. 2020, 1, 100001. [Google Scholar]
  6. Paul, D.; Sanap, G.; Shenoy, S.; Kalyane, D.; Kalia, K.; Tekade, R.K. Artificial intelligence in drug discovery and development. Drug Discov. Today 2021, 26, 80–93. [Google Scholar]
  7. Sun, T.Q.; Medaglia, R. Mapping the challenges of Artificial Intelligence in the public sector: Evidence from public healthcare. Gov. Inf. Q. 2019, 36, 368–383. [Google Scholar] [CrossRef]
  8. Neyigapula, B.S. AI in Healthcare: Enhancing Diagnosis, Treatment, and Healthcare Systems for a Smarter Future in India. Adv. Bioeng. Biomed. Sci. Res. 2023, 6, 171–177. [Google Scholar]
  9. Bartoletti, I. AI in healthcare: Ethical and privacy challenges. In Artificial Intelligence in Medicine: 17th Conference on Artificial Intelligence in Medicine, AIME 2019, Poznan, Poland, 26–29 June 2019; Proceedings 17; Springer International Publishing: Cham, Switzerland, 2019; pp. 7–10. [Google Scholar]
  10. Shah, R.; Chircu, A. IoT and AI in healthcare: A systematic literature review. Issues Inf. Syst. 2018, 19, 33–41. [Google Scholar]
  11. Khaled, N.; Turki, A.; Aidalina, M. Implications of artificial intelligence in healthcare delivery in the hospital settings. Int. J. Public Health Clin. Sci. 2019, 6, 22–38. [Google Scholar]
  12. Reddy, S.; Fox, J.; Purohit, M.P. Artificial intelligence-enabled healthcare delivery. J. R. Soc. Med. 2018, 112, 22–28. [Google Scholar] [CrossRef] [PubMed]
  13. Wang, Y.-C.; Chen, T.-C.T.; Chiu, M.-C. An improved explainable artificial intelligence tool in healthcare for hospital recommendation. Healthc. Anal. 2023, 3, 100147. [Google Scholar] [CrossRef]
  14. Srinivasu, P.N.; Sandhya, N.; Jhaveri, R.H.; Raut, R. From blackbox to explainable AI in healthcare: Existing tools and case studies. Mob. Inf. Syst. 2022, 2022, 167821. [Google Scholar]
  15. Mueller, B.; Kinoshita, T.; Peebles, A.; Graber, M.A.; Lee, S. Artificial intelligence and machine learning in emergency medicine: A narrative review. Acute Med. Surg. 2022, 9, e740. [Google Scholar] [CrossRef]
  16. Rebelo, A.D.; Verboom, D.E.; dos Santos, N.R.; de Graaf, J.W. The impact of artificial intelligence on the tasks of mental healthcare workers: A scoping review. Comput. Hum. Behav. Artif. Hum. 2023, 1, 100008. [Google Scholar]
  17. Bertsimas, D.; Bjarnadóttir, M.V.; Kane, M.A.; Kryder, J.C.; Pandey, R.; Vempala, S.; Wang, G. Algorithmic prediction of health-care costs. Oper. Res. 2008, 56, 1382–1392. [Google Scholar] [CrossRef]
  18. van Mens, K.; Kwakernaak, S.; Janssen, R.; Cahn, W.; Lokkerbol, J.; Tiemens, B. Predicting future service use in Dutch mental healthcare: A machine learning approach. Adm. Policy Ment. Health Ment. Health Serv. Res. 2021, 49, 116–124. [Google Scholar]
  19. Mangla, S.K.; Luthra, S.; Mishra, N.; Singh, A.; Rana, N.P.; Dora, M.; Dwivedi, Y. Barriers to effective circular supply chain management in a developing country context. Prod. Plan. Control. 2018, 29, 551–569. [Google Scholar]
  20. Mishra, A.; Dutta, P.; Kakoti, S. Exploring key success factors of Indian pharmaceutical supply chain using interpretive structural modeling. In Multi-Criteria Decision Analysis in Management; IGI Global: Hershey, PA, USA, 2020; pp. 46–61. [Google Scholar]
  21. Diabat, A.; Govindan, K. An analysis of the drivers affecting the implementation of green supply chain management. Resour. Conserv. Recycl. 2011, 55, 659–667. [Google Scholar] [CrossRef]
  22. Tripathy, S.; Sahu, S.; Ray, P.K. Interpretive structural modelling for critical success factors of R&D performance in Indian manufacturing firms. J. Model. Manag. 2013, 8, 212–240. [Google Scholar] [CrossRef]
  23. Chakraborty, K.; Mondal, S.; Mukherjee, K. Critical analysis of enablers and barriers in extension of useful life of automotive products through remanufacturing. J. Clean. Prod. 2019, 227, 1117–1135. [Google Scholar]
  24. Sushil. Interpreting the interpretive structural model. Glob. J. Flex. Syst. Manag. 2012, 13, 87–106. [Google Scholar]
  25. Thakkar, J.; Deshmukh, S.G.; Gupta, A.D.; Shankar, R. Selection of third-party logistics (3PL): A hybrid approach using interpretive structural modeling (ISM) and analytic network process (ANP). Supply Chain Forum Int. J. 2005, 6, 32–46. [Google Scholar] [CrossRef]
  26. Thakkar, J.; Deshmukh, S.G.; Gupta, A.D.; Shankar, R. Development of balanced scorecard: An integrated approach of interpretive structural modeling (ISM) and analytic network process (ANP). Int. J. Product. Perform. Manag. 2007, 56, 25–59. [Google Scholar]
  27. Iqbal, M.; Ma, J.; Ahmad, N.; Ullah, Z.; Hassan, A. Energy-Efficient supply chains in construction industry: An analysis of critical success factors using ISM-MICMAC approach. Int. J. Green Energy 2022, 20, 265–283. [Google Scholar]
  28. Akpinar, H.; Caylan, D.O. Modeling organizational resilience in maritime business: An ISM and MICMAC approach. Bus. Process. Manag. J. 2023, 29, 597–629. [Google Scholar] [CrossRef]
  29. Agarwal, R.; Bhadauria, A.; Kaushik, H.; Swami, S.; Rajwanshi, R. ISM-MICMAC-based study on key enablers in the adoption of solar renewable energy products in India. Technol. Soc. 2023, 75, 102375. [Google Scholar] [CrossRef]
  30. Gadekar, R.; Sarkar, B.; Gadekar, A. Model development for assessing inhibitors impacting Industry 4.0 implementation in Indian manufacturing industries: An integrated ISM-Fuzzy MICMAC approach. Int. J. Syst. Assur. Eng. Manag. 2022, 15, 646–671. [Google Scholar] [CrossRef]
  31. Asif, M.; Sarim, M.; Khan, W.; Khan, S. ISM- and MICMAC-based modelling of dairy supply chain: A study of enablers in Indian economy perspective. Br. Food J. 2023, 126, 578–594. [Google Scholar] [CrossRef]
  32. Feng, X.; Li, E.; Li, J. Critical factors identification of digital innovation in manufacturing enterprises: Three stage hybrid DEMATEL–ISM–MICMAC approach. Soft Comput. 2024, 28, 7341–7361. [Google Scholar] [CrossRef]
  33. Govindan, K.; Palaniappan, M.; Zhu, Q.; Kannan, D. Analysis of third party reverse logistics provider using interpretive structural modeling. Int. J. Prod. Econ. 2012, 140, 204–211. [Google Scholar] [CrossRef]
  34. Attri, R.; Dev, N.; Sharma, V. Interpretive structural modeling (ISM) approach: An overview. Res. J. Manag. Sci. 2013, 2319, 1171. [Google Scholar]
  35. Sivaprakasam, R.; Selladurai, V.; Sasikumar, P. Implementation of interpretive structural modelling methodology as a strategic decision making tool in a Green Supply Chain Context. Ann. Oper. Res. 2014, 233, 423–448. [Google Scholar] [CrossRef]
  36. Raj, T.; Shankar, R.; Suhaib, M. An ISM approach for modelling the enablers of flexible manufacturing system: The case for India. Int. J. Prod. Res. 2008, 46, 6883–6912. [Google Scholar] [CrossRef]
  37. .Kasula, B.Y. AI applications in healthcare a comprehensive review of advancements and challenges. Int. J. Manag. Educ. Sustain. Dev. 2023, 6, 2023. [Google Scholar]
  38. Iliashenko, O.; Bikkulova, Z.; Dubgorn, A. Opportunities and challenges of artificial intelligence in healthcare. E3S Web Conf. 2019, 110, 02028. [Google Scholar] [CrossRef]
  39. Dumitrașcu, L.M.; Lespezeanu, D.A.; Zugravu, C.A.; Constantin, C. Perceptions of the Impact of Artificial Intelligence among Internal Medicine Physicians as a Step in Social Responsibility Implementation: A Cross-Sectional Study. Healthcare 2024, 12, 1502. [Google Scholar] [CrossRef]
Figure 1. The different steps followed as the framework of this study.
Figure 1. The different steps followed as the framework of this study.
Algorithms 18 00189 g001
Figure 2. ISM model.
Figure 2. ISM model.
Algorithms 18 00189 g002
Figure 3. Micmac analysis.
Figure 3. Micmac analysis.
Algorithms 18 00189 g003
Table 1. Key challenges of AI in health sectors identified by previous researchers.
Table 1. Key challenges of AI in health sectors identified by previous researchers.
Identified ChallengesNotationsKey Resources
Insufficient dataIDNeyigapula, 2023 [8], Paul et al., 2020 [6], Bartoletti, 2019 [9], Shah and Chircu, 2018 [10]
Social issues SIGille et al., 2020 [5], Khaled et al., 2019 [11]
Clinical implementationCIAung et al., 2021 [4], Reddy et al., 2019 [12], Shah and Chircu, 2018 [10]
High cost HCAung et al., 2021 [4], Jimma, 2023 [3]
Black-box scenarioBSWang et al., 2023 [13], Srinivasu et al., 2022 [14], Reddy et al., 2019 [12]
Data acquisition DAAung et al., 2021 [4], Mueller et al., 2022 [15]
Introduction of innovative and new-generation tools IINWang et al., 2023 [13], Rebelo et al., 2023 [16], Van Mens et al., 2022 [17]
Missing compassion MCAung et al., 2021 [4], Khaled et al., 2019 [11]
Data misuse DMBartoletti, 2019 [9], Aung et al., 2021 [4]
Data privacy and security DPSShah and Chircu, 2018 [10], Bartoletti, 2019 [9], Sun et al., 2019 [7]
Technology development TDWang et al., 2023, Rebelo et al., 2023 [16]
Table 2. Notations of the different challenges identified in this study.
Table 2. Notations of the different challenges identified in this study.
Factor No.NotationsDifferent Challenges
F1IDINSUFFICIENT DATA
F2SISOCIAL ISSUES
F3CICLINICAL IMPLEMENTATION
F4HCHIGH COST
F5BSBLACK-BOX SCENARIO
F6DADATA ACQUISITION
F7IININTRODUCTION OF INNOVATIVE AND NEW-GENERATION TOOLS
F8MCMISSING COMPASSION
F9DMDATA MISUSE
F10DPSDATA PRIVACY AND SECURITY
F11TDTECHNOLOGY DEVELOPMENT
Table 3. Demonstration of how modern scholars have applied ISM to decision-making in several domains.
Table 3. Demonstration of how modern scholars have applied ISM to decision-making in several domains.
ResourcesObjectives
Iqbal et al., 2023 [27]Energy efficient supply chain in the construction industry
Akpinar et al., 2023 [28]Resilience in maritime business
Agarwal et al., 2023 [29]Adoption of solar renewable energy products in India
Gadekar et al., 2024 [30]Study of the inhibitors that affect Industry 4.0 implementation in manufacturing industries of India
Asif et al., 2024 [31]Dairy supply chain
Feng et al., 2024 [32]Digital innovation in manufacturing enterprises
Table 4. Structural self-interaction matrix (SSIM).
Table 4. Structural self-interaction matrix (SSIM).
NotationF1-
ID
F2-
SI
F3-
CI
F4-
HC
F5-
BS
F6-
DA
F7-
IIN
F8-
MC
F9-
DM
F10-
DPS
F11-
TD
F1-ID OVVVOVOOVV
F2-SI VVOOVOOOV
F3-CI VAAVOOAV
F4-HC AVOOAO
F5-BS AVOOOV
F6-DA VOOVV
F7-IIN AAAA
F8-MC OOV
F9-DM VV
F10-DPS V
F11-TD
Table 5. Reachability matrix.
Table 5. Reachability matrix.
F1-
ID
F2-
SI
F3-
CI
F4-
HC
F5-
BS
F6-
DA
F7-
IIN
F8-
MC
F9-
DM
F10-
DPS
F11-
TD
Driving Power
F1-ID101110100117
F2-SI011100100015
F3-CI001100100014
F4-HC000100100002
F5-BS001110100015
F6-DA001111100117
F7-IIN000000100001
F8-MC011100110016
F9-DM001110101117
F10-DPS001100100115
F11-TD000000100012
Total dependencies1289411111491
Table 6. Level partition.
Table 6. Level partition.
Factor No.Reachability SetAntecedent SetInteraction SetLevel
F1-ID1,3,4,5,7,10,1111One
F2-SI2,3,4,7,112,82Two
F3-CI3,4,7,111,2,3,5,6,8,9,103Three
F4-HC4,71,2,3,4,5,6,8,9,104Four
F5-BS3,4,5,7,111,5,6,95Two
F6-DA3,4,5,6,7,10,1166One
F7-IIN71,2,3,4,5,6,7,8,9,10,117Five
F8-MC2,3,4,7,8,1188One
F9-DM3,4,5,7,9,10,1199One
F10-DPS3,4,7,10,111,6,9,1010Two
F11-TD7,111,2,3,5,6,8,9,10,1111Four
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Angelina, Q.; Begum, K.; Kim, H.-C.; Tripathy, S.; Singhal, D.; Singh, S. A Structural Analysis of AI Implementation Challenges in Healthcare. Algorithms 2025, 18, 189. https://doi.org/10.3390/a18040189

AMA Style

Angelina Q, Begum K, Kim H-C, Tripathy S, Singhal D, Singh S. A Structural Analysis of AI Implementation Challenges in Healthcare. Algorithms. 2025; 18(4):189. https://doi.org/10.3390/a18040189

Chicago/Turabian Style

Angelina, Q, Khadija Begum, Hee-Cheol Kim, Sushanta Tripathy, Deepak Singhal, and Saranjit Singh. 2025. "A Structural Analysis of AI Implementation Challenges in Healthcare" Algorithms 18, no. 4: 189. https://doi.org/10.3390/a18040189

APA Style

Angelina, Q., Begum, K., Kim, H.-C., Tripathy, S., Singhal, D., & Singh, S. (2025). A Structural Analysis of AI Implementation Challenges in Healthcare. Algorithms, 18(4), 189. https://doi.org/10.3390/a18040189

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop